⚛️┃physics

[elder lintel]

fixed it on my own

[elfin latch]

what happened lol?

[timid dove]

I would guess a script setting the vertical velocity to 0 each frame, that's the usual cause

[elder lintel]

#⚛️┃physics message

Basically what Praetor said

[signal wyvern]

Anyone able to help me debug my Client Prediction setup for my game, Im currently predicting the entire scene, this includes water, bouyancy and the boats, Meaning that almost everything you see as a client is ahead of the server, the clients then receive updates from my server, client then resets to this update and re-simulates back up to its current frame, continuously realligning the client with the server. im specifically stuggling with the boat destruction and im getting some jitter when splitting boats which indicates a difference in state, Im reappling joints and resimulating the destruction for clients which i believe is potentially causing this jitter, wondering if theres any experts out there which know everything there is to know about adding/destroying joints specifically fixed joints.

[wicked phoenix]

Hi, does anyone know of a clean way to move a non-kinematic rigidbody character on a moving kinematic-rigidbody platform? I currently have the platform using a spline so that it loops around smoothly in FixedUpdate.

Here is the script for the moving platform currently:

public class MovingObjectParenter : MonoBehaviour
{
     [SerializeField] private Rigidbody rb;
    [SerializeField] private List<Rigidbody> collected;
    private Vector3 prevPos;
    private Vector3 _velocity;


    void Start()
    {
        prevPos = rb.position;
    }

    void FixedUpdate()
    {
        Vector3 _distance = rb.position - prevPos;
        _velocity = _distance;


        for (int i = 0; i < collected.Count; i++)
        {
            collected[i].MovePosition(collected[i].position + _velocity);
        }

        for (int i = 0; i < agents.Count; i++)
        {
            agents[i].Move(_velocity);
        }

        prevPos = rb.position;
    }

    private void OnTriggerEnter(Collider other)
    {
        if (other.TryGetComponent(out Rigidbody _rb))
        {
            if (!collected.Contains(_rb)) collected.Add(_rb);
        }
    }

    private void OnTriggerExit(Collider other)
    {
        if (other.TryGetComponent(out Rigidbody _rb))
        {
            if (collected.Contains(_rb)) collected.Remove(_rb);
        }

    }
}

It "somewhat" works on the player, but the player slides around too much during the movement. Any help is appreciated. Note: I move the player rigidbody using AddForce.

Here's how it looks like

[signal wyvern]

How much does your game require you to stick to platforms like that?

[wicked phoenix]

None so far because I hadn't figured out a good way to do so haha. But if I can get this to work, I had an idea for a level designed to be entirely moving

Unless I can figure out an alternative way to "fake" that movement I guess

[signal wyvern]

my game is above we move entirely on moving boats

[signal wyvern]

What are you struggling with?

[wicked phoenix]

It's mainly keeping the rigidbody player on the moving platform. Either the player just doesn't move fast enough with the platform or he slides too much in the direction of the platform's movement

[signal wyvern]

How are you moving the platform? does it have physical movement with a RB? or are you just moving its transform.position?

[wicked phoenix]

It's using an asset called Dreamteck Splines which moves the platform's kinematic rigidbody with MovePosition along the spline in FixedUpdate

[signal wyvern]

Sounds strange, gets a bit more difficult when you use RBs on your platforms also, i dont think you can match velocity here because not sure if MovePosition updates velocity

[unique cave]

It does update the velocity and is the way moving kinematic platforms should be implemented as far as I know (if one wants to use the physics system)

[signal wyvern]

If i were you i would make a child object of the platform, with no physics & i would set the players rb.MovePosition(ChildObject) each frame

you can then apply player movement to the child object which will automatically consider the movement of the platform

[pulsar canopy]

Is it possible to add IK to joints in any meaningful way? Or, joints to ik.

ideally, they would behave the same like normal IK but be reactive with the environment?

Or do I have to forego the animator and write some kind of movement system for the joint that would emulate ik movement?

[bleak umbra]

You may want to have a look at the animation rigging package and see if it can do what you want https://docs.unity3d.com/Packages/com.unity.animation.rigging@1.4/manual/index.html

[pulsar canopy]

I have already used anim rigging extensively. I'm trying to figure out if and how I might combine it with joint behavior or some other kind of rigidbody type physics behavior because rigidbodies DO NOT like animations and vice versa

[bleak umbra]

Well, they are fundamentally opposed concepts. To make them cooperate you need some sort of negotiation mechanism like Ubisoft‘s IK Rig discussed here https://youtu.be/KLjTU0yKS00?si=ORFMo95zwygaAhmh

[pulsar canopy]

from a glance, it seems like that isn't even exactly what I'm talking about, it wouldn't achieve the same goal

the closest thing to what I'm trying to do is puppet master but I've also tested that in a number of ways and it didn't work in my case

maybe I should remake puppet master with my own rules and joint types and logic etc.

[valid cave]

hey i'm making a 2d platformer game and just started making the prototype and almost finshed the movement script done but i have a bug where my rig or player is sticking to the edge of a box collider 2d here my code using UnityEngine;
using UnityEngine.InputSystem;
you will not get credit when i realse the game
because i will forget your name or username https://paste.mod.gg/cuvicopulcbs/0

[silver moss]

What is this video showing about the issue? All I see is a player moving forward (or the environment moving backwards)

If your character was moved by recalculating the velocity each frame, it'd be easy to just add the current platform's velocity to the final velocity before applying

[wicked phoenix]

The platform is moving forward and I'm applying its velocity to the player, but the player slides forward without me touching the controls

[wicked phoenix]

I'm using rigidbody.MovePosition to apply the platform's movement change to the player

[signal wyvern]

My physics is in simulation mode: script for networking purposes, and ive realised that interpolation does not work when in this mode for some reason, does anyone have any fixes for this or know of any workarounds to get some interpolation, otherwise my objects just move at the fixedUpdate rate which is only 50fps

[pulsar canopy]

I have an ik chain for my player's arm which grabs the environment and pulls the player forward when moving (or at least that's what it should look like).

It works well but it doesn't have reactivity with the environment that I would want.

I turned off the animator and rig weight and tried building a chain of configurable joints. It gets me the reactivity that I want (i.e. player extends hand forward but is blocked by a terrain piece jutting out, so he bends the arms inward) but I find this hard to control.

It also only works if it gets stuck there for a bit, otherwise it just moves forward through parenting.

I have problems transitioning the animation to the next iteration of it because my hand just gets stuck behind the player.

Bare in mind, I'm talking about crawling through tight tunnels and pulling myself forward with the hands. Even if I manage go get the hand stuck on terrain (purpose being to make it appear like he's grabbed the terrain, it gets left behind the player when he moves forward and there is no space to bring it back.

I've been thinking of some kind of system that combines joints and ik but it's tough to make that combo work I think. Does anyone have any ideas on what I could do here?

Making a dual rig is another options without joints, just rigidbody colliders but I'll have the same problem pulling it back

[late hound]

Working with an fps movement system i created based on dani's old movement tutorial but heavily adapted to my needs. Everything works great the only problem is the fundamental issue of applying external impulses (such as an explosion force) or a knockback.

The system doesn't use unity's friction (friction on all objects is zero) and instead uses a custom countermovment logic, the problem is that when a player recieves a horizontal knockback while grounded they barely move, compared to while airborne they go flying.

This seems to me like a fundamental flaw in the movement system I created, but pivoting to a whole new system is kind of impossble right now unless absolutely neccessary.

Any other advice and feedback regarding the code is also very welcome!

*Main movement lines are 700-800
This is for an fps parkour game
For additional reference, the player should not be able to surpass their maxspeed with their own input (walking, sprinting, etc), but things like external impulses (explosions, and such) should allow the player to go far beyond their max speed

[timid dove]

I haven't read your code (on mobile) but it sounds like you're probably doing something like setting velocity explicitly while grounded which is overwriting the changes made by any external forces

[late hound]

im almost sure the problem is coming from the countermovement block which is used to keep the movement snappy and also stop the player when no inputs are pressed and they are grounded, the problem is how the countermovement is programmed is contradictory to letting external forces be applied and i don't really know how to still keep the movemeent snappy but also allow things like explosions and impulses to affect the player

perhaps i could have an external script handle the forces applied to the player applied by explosions, but i don't even know where i would get started

[twin bramble]

From also not reading the code (on mobile too) it just sounds like it works as it's supposed to. On ground the friction is gonna be much higher than in air

You can modify the counter force in air to make it a bit better

[noble prawn]

hey so I'm turning gravity on in a rigidbody via a script like so:

public void Interact() {
    if (Player.Instance.HasBeerCup()) {
        //Player is carrying a beer cup
        BeerCup playerBeerCup = Player.Instance.GetBeerCup();
        playerBeerCup.SetBeerCupParent(this);

        int beerCupTableLayer = 7;
        beerCup.SetBeerCupLayer(beerCupTableLayer);

        beerCup.transform.GetComponent<Rigidbody>().useGravity = true;

        float beerCupPushForce = beerCup.transform.GetComponent<Rigidbody>().mass * 2 * (tableLength / Mathf.Pow(timeForBeerToGoToTheEnd, 2));
        beerCup.Push(transform.right, beerCupPushForce);
    }
}

but for some reason, despite me turning gravity on, the rigidbody still won't fall

I already checked if it is kinematic, which it wasn't

has anyone encountered this before?

this interact function is also only called once; when the player clicks the interact button

but for one, it seems like the rb moves forever

and two, the y value never changes--so the gravity isn't working

I've also tried adding a friction physics material to the surface but that doesn't seem to stop it either

[twin bramble]

Have you checked if it actually isnt kinematic and IS using gravity in the editor?

After u cause the script to change them ofc

And change it manually to see if it's a problem with the script or with the rigidbody

[worldly burrow]

Hey, I am trying to make a pong clone but I am a bit unsure on how to handle the physics. I made the two player paddles by locking the X and Z axis, and moving them up and down with rb.MovePosition(). Still, when the ball hits one of the paddles from up or down, that paddle it's pushed because of the applied force.
I tried using Kinematic body type but now learned that it doesn't interact with other objects. Is there a way to create a body that cannot be pushed by other objects but still be stopped by static ones or push other objects?

[pulsar canopy]

You can increase the mass of the rigidobies and they shouldn't be pushed by the ball

But really, the best way would be to use kinematic rigidbodies and manually code the ball movement as it would only require using reflection vectors and no other calculations.

[worldly burrow]

I see, I like that approach. Just one thing, that's I limit the paddles movement by having a static object with two edge colliders on top and bottom to stop the paddles from going offscreen. Is there any way to detect this thing by code?

My first guess would be to establish a max and min y in a variable and not allowing the paddles to move further if they are already there, but I wonder if there is a different aproach

[pulsar canopy]

That's the best approach

[worldly burrow]

I see, thanks!

[kind mauve]

@tender gulch #💻┃unity-talk message
that does not seem to work
I double checked all colliders, main one have a layer not included in mask, all other colliders have layers included in a mask, and everything have rigidbodies
code samples in case

Physics.Raycast(shooting_barrel.position, mob_ai.target.coords - shooting_barrel.position, out RaycastHit hit, test_mask_of_shooting);
RaycastHit[] hits = Physics.RaycastAll(shooting_barrel.position, mob_ai.target.coords - shooting_barrel.position, test_mask_of_shooting);
foreach (var ht in hits)
    {
         Debug.Log("RaycastAll hit " + ht.transform);
    }
Debug.Log("Raycast hit " + hit.transform);

Maybe Unity don't allow parent collider exclusion?

Raycasting all hits everything, Raycasting one hits only first biggest collider (parent)

[tender gulch]

Try logging hit.rigidbody.gameobject.

As well as hit.collider.gameobject

[kind mauve]

they are the same

all that looks right

I mean I can just use RaycastAll and pick what I want but I don't understand why Raycasts hitting everything

[tender gulch]

Just to be sure, what type is test_mask_of_shooting?
And can you log it's value along the raycast as well?

As well as the layer of the hit object.

Also....

Can you put the normal raycast log right after the raycast and not after raycast all?

[kind mauve]

it's a LayerMask

seems like logging out LayerMask doesn't work

[tender vapor]

try casting it to int and logging

[kind mauve]

640... whatever that means

I ve heard of bitwise operators or something but it looked like LayerMask type allowed me to bypass that

that might be it!

I mean issues with a mask

[tender gulch]

It's 8th and 10th layers I think.

or rather index 7 and 9

[tender gulch]

So, RaycastAll showing that it hit layer 8 must not be using the layermask🤔

Yeah, it's probably used as the max distance:

public static RaycastHit[] RaycastAll(Vector3 origin, Vector3 direction, float maxDistance = Mathf.Infinity, int layerMask = DefaultRaycastLayers, QueryTriggerInteraction queryTriggerInteraction = QueryTriggerInteraction.UseGlobal);

And probably same in the single cast as well:

public static bool Raycast(Vector3 origin, Vector3 direction, float maxDistance = Mathf.Infinity, int layerMask = DefaultRaycastLayers, QueryTriggerInteraction queryTriggerInteraction = QueryTriggerInteraction.UseGlobal);

So you're just not using the layermask as a layermask

[zenith parcel]

guys im trying to use character controller for my player but its not applying collisions so it just phases through the floor and falls, but when I try to put a BoxCollider it says it conflicts with the character controller

[silver moss]

CharacterController does not support extra colliders. It acts as a capsule collider for itself

So the issue is in your code or object setup

Does the floor have a proper collider?

[zenith parcel]

it has a boxcollider

[silver moss]

Have you confirmed that it is the correct size

If yes, then it's probably a code issue

Your character movement code

[zenith parcel]

the movement code is literally just this

btw after messing up some properties, now the player just get stuck in the same y position instead of falling

[kind mauve]

oh god
that's pretty embarrassing
thanks for saving me

[timid dove]

If this is your movement code I wouldn't expect any vertical motion at all. Are you also using a Rigidbody?

[tender vapor]

you can't use CC together with RB, they will fight each other

praetor's question wasn't a suggestion

with a CC, you need to apply gravity yourself

[zenith parcel]

and how I do that

[tender vapor]

well, normally, you'd store the vertical velocity yourself, apply acceleration, and then combine that with the horizontal velocity in the move (1 move per frame)
(with extra logic - if grounded, apply a fixed grounded vertical velocity)

i don't know how much, if any, of that changes with the context of network

[tender vapor]

though, on that note - i do notice that you're using FixedUpdateNetwork there? if that's analogous to FixedUpdate, then that'd be the wrong method - CC expects to work on Update

[zenith parcel]

and how I apply collision with anything I want, like walls, and the other players from the room as well

[tender vapor]

i believe that's handled by CC, assuming your Move calls are set up correctly

[barren gorge]

PhysX docs says that you should call Move only once per frame

[pulsar canopy]

Can anybody explain articulation bodies to me?

Loke, I tried chaining joints to one another for something, but it was unstable and jittery when constantly colliding with stuff.

All I understand is that art. bodies are supposed to help with that. Maybe? What is the feature that makes them more enticing that joints?

[noble crown]

the rigidbody's velocity stays at zero, even though it's explicitly marked as nonkinematic and has a velocity change of 250 being applied to it

[tender vapor]

which object is supposed to move?

make sure the code is being called as well

[noble crown]

the fragments that come off. they're visibly moving, but they should be going extremely fast. currently they just fall

[noble crown]

it is. debug messages, and even changes to the rigidbody besides velocity all get correctly applied

[tender vapor]

show the code and the rigidbody setup

!code

[gloomy lodge]

Posting code

📃 Large Code Blocks

Use links to services like:
https://paste.mod.gg/, https://hastebin.skyra.pw/, https://paste.ofcode.org/, https://paste.myst.rs/

📃 Inline Code

Surround code with three backquotes. Not quotation marks.
To format as C#, add cs to the first line:
```cs
// Your code here
```
Add a comment with a line number if there is an error message.

[noble crown]

rb setup:

[noble crown]

code: https://hastebin.com/share/asizesopit.csharp

[tender vapor]

waiiit 3d rigidbodies do have info..

[noble crown]

indeed they do

[tender vapor]

why was it missing here lmao #💻┃code-beginner message

[mellow cradle]

use force impulse on your rigidbody 24/7

[tender vapor]

don't do that.

[mellow cradle]

ok fine add an input command first then do it and become superman

[tender vapor]

impulse isn't a continuous force

[mellow cradle]

multiply it by 100

[tender vapor]

that's not at all relevant to what i said

[tender vapor]

where's force coming from?

[mellow cradle]

its a rigidbody command

[tender vapor]

have you tried debugging the overall force you're applying, eg force * mult * 100

[noble crown]

variable. i've already debug logged this and it's set properly to 250

[tender vapor]

though i guess you'd just have to debug force, it's the only possibly-zero one there

and you've debugged that it's indeed velocitychange?

[noble crown]

yes

[tender vapor]

are you doing linearVelocity = ... anywhere?

[noble crown]

nope

nor do i set the velocity manually in any other way

[tender vapor]

well, that's all my ideas

[timid dove]

Are you sure this code is running? Maybe hitRB is coming back as null

[noble crown]

yes, the code's running. no null errors

ok, it looks like destroying the fixedjoint before adding the force fixes it. the only problem is that i don't want to guarantee that damaging the thing will kill the joint...

[timid dove]

I didn't say errors. There wouldn't be any errors if that variable was null

[noble crown]

i’d assume there would be a null reference exception, which i would definitely consider an error

[timid dove]

Why would you assume that? Your code is explicitly checking the variable with an if statement. There wouldn't be an exception, it would just silently do nothing.

[noble crown]

oh right, i forgot i had a check in there. either way, yes, the code runs

[halcyon sentinel]

funny bug i came across while trying to sort out some movement

[dapper arch]

what's the best way to check how much distance a raycast/linecast "spent" in a collider? i have an idea to just raycast in two directions (first raycast from A to B, another one from B to A), check their raycast hits and calculate the distance between their points, but maybe there's a better way to do that?

[crisp owl]

hit.fraction

tells you the fraction of the ray to the hit point

the other part is past the hit point, which you could assume is inside the collider

it gives an answer from 0.0 - 1.0 so youd have to multiply by ray length to get the actual length

if you want something more complex, like also being able to tell when the ray leaves, I would use RaycastAll

itll give you all the points, and you can feed that into some system to figure out how much distance the ray was inside the collider. Its probably kinda complicated tho

since there will be multiple points

im also unsure if you would get the exit point actually

you may have to raycast in the opposite direction

[dapper arch]

alright, i'll try that. though if there's no other way to get the exit point, then i'll probably stick to my 2 ray solution :/

or i could also raycastall from both points (A-B and B-A), collect all the hits, sort them by dist from A and then create pairs from that (if a linecast between two points finds an obstacle, it means that these two points are probably start/end of a line IN the collider) and check the dist between two points in each pair.

[crisp owl]

that is the solution

there may be edge cases, but im not 100% sure

[dapper arch]

well, i think it should be quite accurate but there may be some physx quirks that i dont know. but it should be ok

[umbral minnow]

if i have a question about colliders and how the Collider.bounds.Contains works, do I ask it here?

[barren gorge]

yep

[umbral minnow]

okay so, super simple question, and the docs are all telling me that I'm right here but just checking, if I have a mesh collider with two cubes stacked on top of eachother, with the second cube on top being much bigger, and im checking the meshcollider's bounds.Contains, what is the actual bounds of this? is it gonna extend the small cube's bounds when calculating to the bounds of the bigger cube? i dont know if this makes much sense lol

basically, if the mesh collider is the black part, is the bounds.Contains simulating the inside as the green part?

[tender vapor]

yes, the bounds of something is just some rectangular cuboid that encompasses the entire thing.
in the case of unity, Bounds is an AABB, an axis-aligned bounding box

collider.bounds.Contains doesn't check for something being in the collider. it gets the bounds of the collider, and it checks if something is in those bounds

[umbral minnow]

thanks for the clarifications :p

[deft belfry]

Can I calculate a rigidbody's velocity so it moves around a point in an orbit?

[bleak umbra]

yes

[deft belfry]

How exactly?

I tried some methods, but they all involve also rotating the rigidbody, which I dont want

[tender vapor]

compute it tangential to the vector between the rigidbody and the center and then keep updating it

that's basically how circular motion works physically

[deft belfry]

As in, the perpendicular vector of (rigidbody-center)?

[tender vapor]

https://en.wikipedia.org/wiki/Circular_motion

[tender vapor]

mb, i combined "tangential to the circle of orbit" and "perpendicular to the vector between the rigidbody and the center" lmao

your interpretation is correct

[deft belfry]

@tender vapor thank you

[bleak umbra]

if you don't want to rotate it, then you can't or shouldn't use velocity, make it kinematic and just calculate an orbit position (which is simple trigonometry) without using force simulation.

[tender vapor]

how are those 2 things related

you can just set the linear velocity for a dynamic rigidbody just fine

[deft belfry]

Maybe they meant angular velocity?

Which I dont want to use

[tender vapor]

that wouldn't be related to the position, so im not sure

a kinematic rigidbody (or no rb at all) might be easier to work with depending on what the overall goal is, but i'm not seeing how that decision relates to rotating it or not

like a kinematic rb for an obstacle that shouldn't stop for anything, as if it had arbitrarily high mass

[bleak umbra]

maybe you can give a bit more context. Do you actually want a physics simulation regarding motion of that orbiting object or do you you just need that object to have a trigger? why is it a rigidbody? this would illuminate which approach would be the most practical for your project.

[worn mantle]

Ello, I'm currently trying to figure out how to test if a box intersects with a Concave mesh repersented by an SDF. If you need to figure out if a box overlaps other things without direct physics methods, do you just raycast 12 times(1 along each edge of a box) to figure out if something is intersecting without containing the box? (ps: if a sdf contains 1 vertex of a box, but not all of them, then it has to intersect.)

[pulsar canopy]

How much of a wizard must one be to get rigidbodies not to shake and jitter with precise movements and frequent collision?

[unique cave]

There are a lot of cases where the physics might appear unstable and jittery. Do you have a specific example in mind? In order to be fast and stable in most cases, the physics system can make some assumptions about the "general use case"

[timid dove]

out of the box as long as you're using interpolation, it shouldn't be jittery

[pulsar canopy]

well it is for me

[pulsar canopy]

I am trying to make dual rig: animation + IK => rigidbody + joints

it even works with addtorque and addforce, it tries to get there, but there is just a whole lot of shaking

the issue is the hands (what I'm trying to rig like this) scrape along terrain etc. and I can't get them to be stable

[timid dove]

you've done something wrong then. Hard to say what without seeing your project.

animation + IK + physics is almost certainly the issue

It's not something I've ever worked with so I don't know what steps you need to take to make that work, but it's a complicated setup

[pulsar canopy]

how

the animation ik and rb are completely unrelated in this context

[timid dove]

those are two completely different systems competing over setting the pose of the objects

[timid dove]

well how would I know that?

[pulsar canopy]

Point is, I do the logic for the IK. The animations play and all that.

Then I have a completely separate hand model that just gets the transforms and rotations of each of the components (arm, forearm, etc) and use AddForce and AddTorque to match that transform, but with physical collisions and reactions

so the fact that animations and ik are used is completely unrelated to the rigidbodies

But even before this, I tried a simple test. I didn't use animations or IK or anything. I literally just stuck a hand with configurable joints that was anchored to the player, that's it. And even that would jitter and shake.

Like, I just cannot get rigidbodies to not shake when they are doing constant collisions and trying to move (in that instance, the joints were trying to spring back to the default position)

[silver moss]

@pulsar canopy Do you mean really fast jittering (direction changing ~each frame) or slower oscillations/wobbliness?

You might just need to add some damping to the forces to avoid overshoot

Joints and physics also have a lot of settings that can help with stability

[pulsar canopy]

that might help, I will try

[pulsar canopy]

joints yeah, I fiddled with a lot of settings

what about "physics"? Any settings I should know about?

[silver moss]

defaultSolverIterations, defaultSolverVelocityIterations come to mind
https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Physics.html

And ofc Time.fixedDeltaTime

[pulsar canopy]

ah, that. To my recollection, I did increase solver iterations

[silver moss]

But I'm still unsure if you're having an instability issue or a damping/overshoot issue (or both)

A video could help

[pulsar canopy]

but I actually didn't know about velocity iterations. However, I use force, so that might not come into play?

[silver moss]

Forces are just another way of modifying velocity

[pulsar canopy]

It can't really be an overshoot issue because it jitters back and forth even before it reaches the target (the force direction being the target)

[silver moss]

Just to rule one thing off, are you making rigidbodies children of other rigidbodies?

[pulsar canopy]

also, I made it so that forces stop being added when it reaches .1 of the position, and at that point my hand (last in the chain), just turns off like a circular saw, spins a couple of revolutions and stops lol

[pulsar canopy]

yes

[silver moss]

And are you moving any of the rigidbodies via its transform (i.e. transform.position = , or an animator controlling it directly)?

[pulsar canopy]

no, ofc not

[silver moss]

That's problematic

[pulsar canopy]

strictly addForce and addTorque

[silver moss]

A rigidbody being a child of another rigidbody (unless kinematic i guess) counts as its transform being modified

Because a child tries to follow its parent transform

[pulsar canopy]

Are they all supposed to a separate game object that just follow via code?

[silver moss]

So avoid rigidbody parenting. Have them as "siblings" instead

[silver moss]

Yeah

[pulsar canopy]

ok I will try that

[silver moss]

That's how i've done my active ragdolls

[pulsar canopy]

because

so a little bit of a backstory

[silver moss]

You can still have them under an empty parent "container" object so they aren't just loose in your scene

(And don't move that parent)

[pulsar canopy]

a little while back, I tried using puppet master for this very thing and it had the same issue of jittering and shaking. Couldn't get it to be stable.

That also automatically set it up so that the rigidbodies are in the default armature chain

and used character joints, but that's unrealted

[pulsar canopy]

yeah ok I understand that much...

[silver moss]

Yeah I'm a bit confused because IIRC unity's built in ragdoll wizard also creates joints that are parented to each other. But I've never had good experiences with rigidbody parenting

[pulsar canopy]

yeah. So the implication is that you've also had issues with parented rbs?

[silver moss]

Definitely, I'd say you should never do that

[pulsar canopy]

well then, I'm confused as to why they're done that way but ok

also, I find the character joint to be the most stable. Other ones behave weirdly for me.

Maybe character joints are good too but they don't have a force driving them back to the default position, so you use them as ragdolls (they are limp like fish)

But I suppose if I'm just adding forces towards positions and rotations, they should work ok for me?

[silver moss]

Idk, I opted for ConfigurableJoint because it exposes all the properties and can do everything that other joints can do

It's obviously more complex

But I'd say you can do most of it with forces/torques too

Here's some good helper code for setting target rotations of a ConfigurableJoint if you'll ever use that
https://gist.github.com/mstevenson/4958837

That's how I make the ragdoll bone follow the animated target bone's rotation

[pulsar canopy]

So have made dual rigs before?

I figured out the code, I just didn't add any damping

should I just damp the force based on proximity or something?

[silver moss]

I think you'd usually do damping by taking the current velocity into account when calculating the force

I just used builtin damping of ConfigurableJoint springs

[pulsar canopy]

oh

yeah I do use damping there tbh

I guess I'll have to play around there. But I'm curious whether unparenting will solve anything for me

[silver moss]

Yes and it was a lot of work, especially if you need somewhat realistic motion

For goofy games it isnt that bad

[pulsar canopy]

Unparenting seems to have solved the jitter, which is great

I need to solve other problems. Probably change the physics material to low friction and some logic, then modify a lot of the values

[proud trellis]

Just saw 6.4 beta launched. Does anyone know when this is coming 👀

[pulsar canopy]

did they elaborate on what they mean with the physics improvements

[torpid hollow]

pretty sure it was talked about more in the video where this was announced

[barren gorge]

improving their DOTS Physics quality

I guess PhysX (game object physics backend) still more stable

[bleak umbra]

physx and unity physics also solve different problems, the latter being a deterministic (non-iterative) solution that is, because of that, simpler and easier to sync/parallelize than iterative solvers like physx, it also likely means that its not as complete on the solver side and can't solve complex constraints as robustly as physx.

[tame python]

Hey everyone. I just encounter an issue can anyone help me?

when i shoot or fire shells at my tanks it flew far that's shouldn't happen but it happened..can anyone tell me what i did wrong?

or how can i fix it?

[tender vapor]

that's nowhere near enough info to help

[tame python]

something like this?

and sometimes it flew far like.... alot far out of the arena.

anyone?

[silver moss]

Still not enough info. What objects have rigidbodies and what type of colliders do they have? Like maybe the shell's rigidbody and collider is pushing the tank. What does your code do, etc.

[twin bramble]

if you're not doing it already, you can try applying the forces in FixedUpdate

that can fix some inconsistencies with physics forces

[vague sentinel]

not sure if this is the right channel but how can i possibly fix this

[timid dove]

#1390346776804069396 or #🔀┃art-asset-workflow

[vague sentinel]

okkkay

[tame python]

ok i'll try that

[tame python]

it's prefabs from the tutorials i'm just doing as they instructs us to do

you can see that on unity hub learn

[tame python]

the tanks have rigidbody and box collider

shell have rigidbody and capsule collider?

is this an issue ?

[deft belfry]

            //add groundSpeed, positionVelocity and the platform's velocity
            rb.linearVelocity = groundSpeed * (Vector2)getTransformVec(Vector2.right) + positionVelocity + platformRBVelocity;``` This is my movement function. groundSpeed is the player's regular speed, and it has acceleration and deceleration. platformRBVelocity is the velocity of a moving platform, and it is added to the overall velocity to keep the player moving with the platform. As you can see in the video, it works normally when the player is moving in the same direction as the platform, but because groundSpeed is larger than rb.linearVelocity.x when moving the opposite direction, the deceleration makes the player move slightly to the right.

How do I fix this?

[arctic echo]

ah so you have applied the acc/dec to the combined velocity not just the player

couple of options - make sure you subtract the platform velocity before applying the acc/dec calculation

or you decouple the groundspeed entirely so it's never influenced by the platform

[deft belfry]

Technically, this is what happens

groundSpeed and acc/dec happens before applying every other needed velocity

[arctic echo]

ah no, groundspeed is being updated by the final velocity which will include the platform

[deft belfry]

So, what exactly do I do?

I still need to make the player move accordingly with the platform

[arctic echo]

i'd separate them out

do the vector 2 calc: groundspeed * etc

then add the platform in on the next line down

rb.linearvel = player + platform

[deft belfry]

Same thing happens

[arctic echo]

what's the new code?

[deft belfry]

            //add groundSpeed, positionVelocity and the platform's velocity
            rb.linearVelocity = groundSpeed * (Vector2)getTransformVec(Vector2.right) + positionVelocity;
            rb.linearVelocity+=platformRBVelocity;```

@arctic echo Again, groundSpeed is unrelated to the platform's velocity, its just that the linearVelocity is composed of groundSpeed+platformVelocity

[arctic echo]

yeah so replace groundspeed with relative speed and then add a float first step reset relativespeed = groundspeed

[deft belfry]

relative speed?

[arctic echo]

then the prior value doesn't stack and re-enter the loop

[arctic echo]

yeah just a placeholder variable that ticks up each time

float relspeed = ground speed

relspeed = { xxxx}

vector 2 playervel = { yy}

and then the linervel

[deft belfry]

Where am I supposed to to use relative speed then?

[arctic echo]

everywhere except the first line

[deft belfry]

Even in the accelDecel function?

[arctic echo]

yes

[deft belfry]

Wouldnt this render groundSpeed completely useless

[arctic echo]

lol no you need to replace that with the calculation for it - ie. linearvel - platformvel

placeholder variable

[deft belfry]

Im just confused on what xxxx and yy supposed to mean

like is xxxx just my accel

[arctic echo]

the content of your existing code

i'm trying to avoid doing a copy paste solve for you :p

[deft belfry]

Honestly a copy and paste might be better than this guessing game

[arctic echo]

so long as you get the idea - make the two vel calculations separately

float relativeSpeed = rb.linearVelocity.x - platformRBVelocity.x;

relativeSpeed = generalStaticFuncs.accelDecel(
    relativeSpeed,
    maxMoveSpeed,
    groundAccel,
    groundDecel,
    groundFriction,
    dir,
    Time.fixedDeltaTime,
    isGrounded
);

// 2. Convert to world velocity
Vector2 playerVelocity = relativeSpeed * (Vector2)getTransformVec(Vector2.right) + positionVelocity;

// 3. Add platform velocity LAST
rb.linearVelocity = playerVelocity + platformRBVelocity;```

[deft belfry]

where is groundSpeed then

[arctic echo]

its been replaced by the explicit calculation in the first line

groundspeed = linearvel - platformvel

[deft belfry]

Again, groundSpeed is already UNRELATED to the platform's velocity

it's its own variable, with its own calculation

[arctic echo]

that's in theory but you need to decouple them in practice to avoid these artifacts

[prime lagoon]

does anyone have an idea as to how I can keep the player momentum during a jump without them having control over it? this is my current code and the problem right now is that the momentum is based on where you looking

if(isGrounded && velocity.y < 0f)
{
    velocity.y = -2f;
}


if(isGrounded)
{
    x = moveA.action.ReadValue<Vector2>().x;
    z = moveA.action.ReadValue<Vector2>().y;
}
mouseX = Mouse.current.delta.ReadValue().x;
mouseY = Mouse.current.delta.ReadValue().y;


xRotation -= mouseY * settings.Instance.mouseSensitivity * 0.5f;
xRotation = Mathf.Clamp(xRotation, -90f, 90f);

cam.transform.localRotation = Quaternion.Euler(xRotation, 0f, 0f);
Player.Rotate(Vector3.up * mouseX * settings.Instance.mouseSensitivity * 0.5f);

Vector3 move = transform.right * x + transform.forward * z;


            // x = horizontal movement, z is vertical movement


rb.Move(move * speed * Time.deltaTime);


if(isGrounded && jump.action.WasPressedThisFrame())
{
    velocity.y = MathF.Sqrt(jumpheight * -4f * gravity);
}

velocity.y += gravity * Time.deltaTime;
rb.Move(velocity * Time.deltaTime);```
if possible I would like the character model to still change the direction it's looking without it impacting the trajectory of the character

[twin bramble]

i mean cant you just disable the player movement control while in air?

if thats what u mean

[prime lagoon]

it's what I have right now, it was the first solution I thought of.

[tender vapor]

wouldn't a dynamic rigidbody handle this pretty easily

[prime lagoon]

but now you can still controll where you're going with your camera

[tender vapor]

might want to consider that if it's viable for your usecase

[twin bramble]

ah thats cuz ur adding the transform.forward to the equasion probably

[tender vapor]

otherwise, you'd want to affect the velocity rather than recomputing it just from the input every tick

so there's an effect of momentum

wait, is this a CC?

why is it named rb

[prime lagoon]

it's C# and it's called rb because I was lazy and had already named the character controller variable rb

[tender vapor]

that's not what Rigidbody.Move takes, so assuming it is a CC - make sure Move is called once per frame for consistent results, you have it being called twice here

[tender vapor]

fun fact, IDEs have multiple features to remedy this
there's the obvious find & replace, but there's also commonly rename symbol that's more intelligent, commonly F2

(also, i'm aware it's c#, cc is short for character controller.)

[prime lagoon]

ah okay, didn't know that

[twin bramble]

ngl i have never used cc's i dont even know what they are to be exact

i always just used a rigidbody

[prime lagoon]

for this I'm guessing I need to fill every vector of the velocity before the second time I'm calling the rb.Move and can remove the first?

[twin bramble]

yeah just add the gravity calculation to the movement one and call Move once

[prime lagoon]

from what I know it should be easier declaring what angles it can walk upon and control some other stuff, but it's also a first for me

yeah something else is happening now, but now it's always walking the same direction and the camera doesn't do anything as to what direction I'm going in

[tender vapor]

you do need the transform.forward bit

you might consider, for example, applying it as an acceleration instead assigning it as the velocity

if that feels too slippery you could add drag or counterforces to oppose momentum when you try to go a different direction

[sharp carbon]

Hey, I’m making a VR sailing game in Unity using the XR Interaction Toolkit and I’m stuck with a physics/joint issue.
I have a rudder GameObject with: Rigidbody, HingeJoint (limits set, connectedBody = boat Rigidbody), XRGrabInteractable set to Velocity Tracking
When I grab and move the rudder by itself, it works fine. But when it’s connected to the boat, grabbing it makes the boat accelerate insanely fast and it starts spinning/torquing like crazy, even on axes that are frozen in the boat Rigidbody constraints. I also tried unparenting the rudder outside the boat and connecting via HingeJoint connectedBody and it still happens. It feels like the grab velocity/forces are getting transferred through the hinge into the boat and injecting huge impulses.
Has anyone dealt with this?

[prime lagoon]

can you tell me how? trying it a few different ways and it is constantly changing something for the worse

[tender vapor]

might want to step back from the code and design/analyze it in terms of physics first

[timid dove]

Are you using a Rigidbody? That Move function signature looks suspiciously like a CharacterController

[tender vapor]

it's a CC

[timid dove]

If it's a CC please tell me you didn't name the variable rb 😢

[tender vapor]

yeah #⚛️┃physics message

[timid dove]

you need to actually keep a variable around keeping track of the current velocity and use that to move when there's no input.

Right now you have a "velocity" variable but it seems to only be handling vertical velocity. You're also calling Move twice per frame which is going to cause bad jittering when you touch objects while jumping

but yeah currently your code is always directly applying the input to the velocity, so there will be no momentum

[prime lagoon]

will be changed, but it made sense when I started with that script but was to laze/busy with other stuff to actually change it

[tender vapor]

it's 4 keypresses

[prime lagoon]

CC.Move(velocity * Time.deltaTime); here it's changed

[twin bramble]

if (Physics.Raycast(LFrontPoint.position, -LFrontPoint.up, out RaycastHit hit2, 0.1f))
{
    Vector3 PointLeft = BoardRB.transform.forward;

    Debug.DrawRay(LFrontPoint.position, hit2.normal * 2, Color.yellow);

    //cross products
    Vector3 left = Vector3.Cross(hit2.normal, Vector3.up);
    Vector3 down = Vector3.Cross(hit2.normal, left);

    Debug.DrawRay(LFrontPoint.position, left * 2, Color.green);
    Debug.DrawRay(LFrontPoint.position, down * 2, Color.blue);
    //flatten
    Vector3 planeNormal = Vector3.Cross(left, down).normalized;
    Vector3 flattenedEdge = PointLeft - Vector3.Dot(PointLeft, planeNormal) * planeNormal;
    Debug.DrawRay(LFrontPoint.position, flattenedEdge * 2, Color.red);

    Vector3 LFrontV = BoardRB.GetPointVelocity(LFrontPoint.position);

    float sideSpeed = Vector3.Dot(LFrontV, flattenedEdge);
    Vector3 sideV = flattenedEdge * sideSpeed;
    Debug.DrawRay(LFrontPoint.position, sideV * 2, Color.magenta);
    Debug.DrawRay(LFrontPoint.position, LFrontV * 2, Color.cyan);

    //calculate the force
    Vector3 CalculatedForce = (sideV * TEMP_FrictionFunction * (1f * 9.81f * 0.4f));
    float forceX = Mathf.Clamp(CalculatedForce.x, -3, 3f);
    float forceY = Mathf.Clamp(CalculatedForce.y, -3, 3f);
    float forceZ = Mathf.Clamp(CalculatedForce.z, -3, 3f);
    Vector3 fullForce = new Vector3(forceX, forceY, forceZ) * -1;
    Debug.DrawRay(LFrontPoint.position, (fullForce) * 2, Color.black);

    BoardRB.AddForceAtPosition(fullForce, LFrontPoint.position, ForceMode.Force);
}

I have this script which rotates the board to face down the hill. problem is that the front point (which is this script) rotates the board backwards and goes down the hill backwards. its not accidentally rotating backwards cuz it stabilizes itself after it has done so. all of the axis are correct since i checked with the 1 million debug ray's. basically this front point is acting like the rear.

this script might be confusing but any help will be appreciated since im not good at math and stuff

[timid dove]

it's hard to understand what you mean by "the board" and "the hill" etc without any context

I'm realizing now you probably mean like... a snowboard or something?

Anyway this code doesn't seem to rotate anything

it only adds a force

unless the rotation is coming from the torque generated by adding the force off center

[twin bramble]

Yeah probably should've given more context: yeah it's a snowboard and the force is added on the front of it in order to turn

I can give SS's and more details tmr if needed

Oh and btw BoardRB.transform.forward is the left side of the snowboard not the front. Just cuz of how it's rotated

[timid dove]

I'm going to guess there's probably a bunch of confusion and bugginess happening due to this

I think the best course of action is to go in and fix your 3D model so that forward is actually forward

then your code can be simpler and easier to think about

[twin bramble]

alr i'll fix the rotation and add some notes in the code to be easier to understand what it does

[short sable]

When it comes to physics I know when using addforce.force gives you better physics, does the same apply when using addforce.acceleration?

[tender vapor]

force is acceleration times mass. applying force takes mass into account. applying acceleration does not.

that's all they differ by

[unique cave]

Better compared to what? I don't think there is much if any difference whether you use .velocity or .AddForce with a force that instantly adds up to the same velocity. It is just that objects generally don't change their velocity instantly but rather over time which is easier to do with .AddForce but I don't think there is anything in there that makes it inherently better or worse

[old abyss]

hi, has anyone ever tried the limits of determinism in unity?
And by determinism i mean only the physics. Considering a simulation that requires no user input, and considering that it only needs to be deterministic on one platform/architecture, how much can we expect from Unity?

[pulsar canopy]

what is the "main axis" in a joint? If it's allowed to move in 3 axes, what makes the main axis the main axis? I don't understand what changes there.

Furthermore, if the joint is locked in 1 axis, does that just mean that it's rotation/movement is locked while trying to move/rotate directly in that axis?
It has got to be able to move via translation of Eulers, right?

[low breach]

unity uses physx, which is not deterministic

[vagrant ingot]

Hello
I want to check if there is a collider in front of my object, so can be my player or an enemy.
This method is :

public static bool CanMove(Vector3 direction, Vector3 currentPos, bool isCheckObject = false)
    {
        // True if no obstacle in the direction, otherwise false
        
        int layerMask;
        //Get the bitmask and not the index of the layer (so 0010 0000 instead of 6)
        if (isCheckObject)
        {
            layerMask = LayerMask.GetMask("Obstacle", "Object");
        }
        else
        {
            layerMask = LayerMask.GetMask("Obstacle");
        }
        
        //Start the raycast at the center of the object 
        Vector3 rayOrigin = currentPos + Vector3.up * 0.5f;
        
        Debug.DrawRay(rayOrigin, direction, Color.green, 1f);
        
        //If nothing is hit, can move
        return !Physics.Raycast(rayOrigin, direction, 1f, layerMask);
    }

And its call in the player class here :

private void CheckToMove(Vector3Int targetGridPos, Vector3Int moveDirection)
    {
        if (!Toolbox.CanMove(moveDirection, transform.position))
        {
            Debug.Log(moveDirection);
            StartCoroutine(BumpAnimation(moveDirection));
            return;
        }

        StartCoroutine(MoveCoroutine(targetGridPos));
    }

When I go in front of a wall or not, the CanMove method return always true, meaning the raycast return always false. Why ?
There is the screen of my drawRay going through my wall with the properties show

[old abyss]

deterministic has a multitude of definitions tho. im aware that its not cross platform deterministic but it is algorithmically deterministic

ive read the docs and on a single platform and architecture say 64-bit intel processor running windows 11, every run should be deterministic

ive been testing and playing around with this stuff for months and i just wanna see if anyone else has some findings related to determinism so i can compare my results to theirs

[timid dove]

one kinda dumb question - can you show your physics settings?

Especially these two

[pulsar canopy]

How can I make it so that the x and y axes switch? I have tried EVERY possible combination of axis values but none of them can do it

I can actually get close to it, but it seems impossible to get there

ofc I get it after right after I posted, even though I've been trying for a while...

I had to switch the secondary axis

[vagrant ingot]

I forgot to put the GameObject SDK to PhysX 🫠

But in simulation mode, what is the difference between fixed update and script ?

[timid dove]

FixedUpdate is the default and most likely what you want
Script means you manually call Physics.Simulate in your code

[vagrant ingot]

In which context you want to simulate physics only on code ?

[timid dove]

When you want full control over it. Could be for a simulation-style game, or a non-realtime game. It is also useful for networking frameworks when the physics engine needs to be syn ed up with networking stuff.

In your game it's potentially useful if your game is not realtime.

it would mean simulating physics only manually when you want to instead of it happening automatically all the time

[vagrant ingot]

Okay I see, thanks for the explanation 👍

[edgy kite]

Hey guys. I cant seem to work this out even though it looks like it should work. Maybe a second pair of eyes will help.

I want to have a knockback on the player when they get hit from a moving object.

Scripted on moving object

private void OnTriggerEnter2D(Collider2D collision) {
    if (collision.CompareTag("Player")) {
        hitDirection = (transform.position - collision.transform.position).normalized;
        StartCoroutine(KnockBack(collision.gameObject));
    }
}

IEnumerator KnockBack(GameObject player) {
    player.GetComponent<Movement>().canMove = false;
    player.GetComponent<Rigidbody2D>().linearVelocity = Vector2.zero;
    player.GetComponent<Rigidbody2D>().AddForce(hitDirection * hitForce, ForceMode2D.Impulse);
    yield return new WaitForSeconds(0.2f);
    player.GetComponent<Movement>().canMove = true;
}

And this is my movement script

void Update() {
    if (canMove) {
        moveDir = new Vector2(Input.GetAxis("Horizontal"), Input.GetAxis("Vertical")).normalized;
        if (moveDir != Vector2.zero) {
            Quaternion toRotation = Quaternion.LookRotation(Vector3.forward, moveDir);
            transform.rotation = Quaternion.RotateTowards(transform.rotation, toRotation, rotationSpeed * Time.deltaTime);
        }
    }
    else moveDir = Vector2.zero;
}

void FixedUpdate() {
    rb.linearVelocity = moveDir * moveSpeed * Time.deltaTime;
}

As you can see this should work? shouldn't it?

[inner thistle]

The movement script sets the velocity to zero every frame that the knockback is in effect

[edgy kite]

of course cause that is there to stop the player still moving when a shop is open. Thank you

That actually still doesn't work.

But in saying that adding the condition to the fixedUpdate to only run while im allowed to move did work.
Told you a second pair of eyes always works

[inner thistle]

Unrelated but velocity shouldn't be multiplied by deltatime

[edgy kite]

Is that because its physics based?

[inner thistle]

Velocity is already a frame-independent value (units per second) so it doesn't need to be converted

[edgy kite]

Ah no worries. Thank for that too.

[granite geyser]

Is there a way to make collision sounds that works for objects that are already colliding ?
Like for example a very tall plank that falls over wouldnt trigger OnCollisionEnter because it is already toutching the floor, but if i use OnCollisionStay it just plays sound every frame, is there a proper way to make this work or should i just use a cooldown or something like that between sounds?

[timid dove]

Use OnCollisionEnter and OnCollisionStay but actually check the relative velocities and impulses in the contact points

Play the sound when they exceed a threshold, and/or modulate the volume by that threshold

[granite geyser]

yeah thats what im doing but it does still play every frame if the object is dragged/ is sliding across a surface

[timid dove]

Well first off I would expect that to make a sound

Second, like I said you need to vary or modulate the sound based on the impulses and velocities of the contact

You can see if the motion is parallel to the contact surface and play a dragging sound instead of an impact sound

All of the information you need is in the Collision object

[granite geyser]

oohh thats a good idea, ill try

thanks!

[bold yew]

Could anybody help me figure out why my Physics Debugger won't show collision meshes? Unity 6000.3 Neither Scene or Game view show them, Gizmos are on, "Collision Geometry" is checked.

All of the colliders are Default layer convex mesh colliders.

but not even cube colliders or the ground (quad with mesh collider) are showing.

[timid dove]

where are the box colliders supposed to be?

The quad is a concave mesh collider IIRC and those don't ever render AFAIK.

[bold yew]

Cube offscreen. Same deal. The quad is set to convex which seems to work but let me know if there's something wrong with that

Cube is here just to see if the collision geometry just wasn't rendering for my vehicle. It's not rendering for anything.

But it does have a cube collider.

Box collider rather

[timid dove]

I can see the collider on the left

[timid dove]

And I see the tank colliders on the right here

[bold yew]

Thats because I'm selecting them

Previously when I have collision geometry selected with the physics debugger it renders them all the time, and with a translucent shader instead of wireframe

I don't think I've changed anything but the debugger has stopped working on even archived versions of the old project where I use it.

Drawing contacts/forces does still work, just not collision hulls.

The shot with them shown is just to show that the colliders are there.

[timid dove]

hmm...

It's not working for me either... trying to see what I'm missing

[bold yew]

Thats bizarre. On 2022.3 (old revision of project) it worked for me, but even on that archived backup it doesn't work anymore.

I was worried it was my machine. I'd think it's a unity bug but since it stopped working? I really have no clue.

[golden island]

trying to do a char controller with movement a bit like octodad

I've got a rigged character with iks on legs

was going to try and use those to move him around and use ragdoll generator for the rest of the rig

dunno if that will work but regardless of that the physics is super whack for some reason

everything just kinda spazzes and slides about

no intersecting colliders

any ideas?

[twin bramble]

Would have to see what that means

And the setup of the character

[glad cargo]

yo, I'm trying to get this sort of floppy driven-by-IK locomotion working in 2D, anyone got any pointers? https://youtu.be/ZRLibDF-JGQ

[twin bramble]

just make like a 2d active ragdoll with ik legs

shouldnt be too hard to replicate i think

[upper helm]

is there an inbuilt way to have strong gravity, but after the player jumps around the peak of their jump for a bit longer and then accelerate downwards

because my movement is server authoritative it wouldnt be good for players to calculate on script by themselves

[timid dove]

"inbuilt"? Now, this is a highly bespoke behavior you need to write the code for yourself.. You can certainly do it using the tools the physics engine gives you.

The physics engine provides basic newtonian physics.

[inner thistle]

Also the premise doesn't make sense, it makes no difference to server authoritative if it's custom or built-in physics

[upper helm]

yes that is what i meant sorry

[upper helm]

o yeah

nevermind then thanks

[granite geyser]

Dumb question but why is there include and exclude layers on colliders ? why not just include or exclude ? what happens if something is on both include and exclude ? or neither ?

[unique cave]

The Layer Collision Matrix in the physics settings is what defines which layers interact with each other. The Layer Overrides can be used to override that behavior for specific collider. Includes are layers that the given collider can interact with even if the collision matrix says they shouldn't interact and exclude layers are ones that should not interact with the given collider regardless of the collision matrix. You can check this page https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Collider-layerOverridePriority.html for more information on how conflicts are resolved if two colliders collide where one should and the other shouldn't interact with each other.

I did not find any information of the situation where you put the same layer on both include and exclude layers. Since that is a contradiction by nature, I would not recommend doing that. The behavior might be undefined and differ between editor versions.

Basically LayersThisColliderInteractsWith = LayersInCollisionMatrix + IncludeLayers - ExcludeLayers

[granite geyser]

Ohh i forgot about the collision matrix ^^'
Thank you, this is all i needed

[floral estuary]

cinemachine camera following zooming out and sometimes zooming in of following target, i just want it to follow the target at a specific area

i have changed the update mode to fixed

the target its following is an aircraft btw so it can get pretty damn fast

[frigid pier]

#🎥┃cinemachine has manual and tutorials pinned, you should find there how to limit camera to a fixed range.

[peak cloak]

I'm trying to make a bike script, everything works well except the steering of the wheels since it just flips over instantly like the bike flips and becomes upside down. https://paste.myst.rs/8i3y4tyg that's the code and I also attached the input asset and the values of all my components.

[silver moss]

You can try setting the rigidbody's center of mass to an (unnaturally) low point so it's more bottom weighted

I mean like way below the rigidbody itself

Been a while since i worked with wheelcolliders tho

Also a joint or corrective forces/torques can be used for balancing

[peak cloak]

I fixed it by making the turning a bit smoother and disabling z rotation

[worn summit]

hi, I have two objects with a rigidbody and they both have 2 colliders, one with a non-trigger collider and one that's a trigger. they belong on 2 separate layers. how can I make them go through each other (non-trigger colliders get ignored) but have their triggers still interact with each other? or something that produces this effect?

what i had done originally was create a child object just for the non-trigger colliders and that child object would live in one of the two layers that i want ignoring each other. the trigger collider would then reside in, say, the parent object that lives on default or something so its ontriggerenter can run. is there a better way of doing this?

example: i want my player and enemy to move through each other (enemy layer and player layer ignore each other) but i still want their triggers to interact with each other so they can do stuff like inflict damage

[timid dove]

yes, put the colliders on different objects which are on different layers and set up the layer collision matrix appropriately

[worn summit]

tyvm!

[peak cloak]

I have this ragdoll human. I'm working on a bicycle rage game that will have 2 people on this bike but the ragdoll sometimes gets loose and ends up under the bicycle which results in the bike clipping through the map. I have fixed joints on his feet and hands, I want him to move a bit with the movement but not too much. Also sometimes the hands start to jitter alot in place which isn't too much what I want. How can I fix these issues?

[peak cloak]

recorded a video to display the problem a bit better

[kind shadow]

Why is the capsule collider shown in physics debug smaller than the character controller capsule?

[craggy raven]

Any idea why the textures look so weird when importing into unity?

[timid dove]

are you sure that's the CC and not a separate capsule?

[timid dove]

#🔀┃art-asset-workflow

[kind shadow]

yes

[inner thistle]

CC might use a "normal" collider internally to interact with the physics system but it's an implementation detail, not really something that's useful to dig into

[kind shadow]

Got it. I was a bit worried I messed up somewhere. Thanks.

[umbral zinc]

Can someone please help me with this when i play the game the character stucks in the ground like this whats the issue @everyone

[timid dove]

You should not try to tag everyone

just post your problem.

Check the positions of your colliders and/or your animations

[umbral zinc]

Sorry

[umbral zinc]

Ok, but can you please give explain further to

[timid dove]

Click on the object that has the collider

make sure gizmos are turned on

and see how the collider is situated relative to the visual elements of your character

[umbral zinc]

Ohk

[glass nebula]

I made geometry dash physics

[tender vapor]

hi, if you're just showcasing your work that would go in #🏆┃daily-win or #1180170818983051344 , this is a help channel

[glass nebula]

whoops

mb

sorry about that

[pulsar canopy]

What is the practical difference between acceleration and velocityChange?

I am currently using acceleration for some rigidbodies to follow and rotate towards some other objects (better than ForceMode.Force).

Would VelocityChange make a practical difference and why?

[twin bramble]

i mean acceleration is the change of velocity so idk if that would make a difference

[inner thistle]

VelocityChange is meant for instant velocity changes and acceleration for continuous change over time

[timid dove]

acceleration multiplies the value you put in by Time.fixedDeltaTime before adjusting the velocity. It makes your input represent an acceleration rate rather than a direct velocity change of that amount

They both effectively do the same thing (adjust the velocity), just by different amounts

[barren gorge]

Force:          body_velocity += (your_force / mass) * Time.fixedDeltaTime
Acceleration:   body_velocity += your_force * Time.fixedDeltaTime
Impulse:        body_velocity += your_force / mass
VelocityChange: body_velocity += your_force

(note that you can change physics update rate from fixed step to dynamic step so in that case instead of the Time.fixedDeltaTime there will be Time.deltaTime)

[timid dove]

(never do that though)

[barren gorge]

(yep)

(btw Time.deltaTime in FixedUpdate method is equal to just Time.fixedDeltaTime, nice unknown feature!)

[pulsar canopy]

ah ok so acceleration is best in this case

[pulsar canopy]

is it wrong to move and rotate something in a coroutine, and the follow it with rigidbody force movement and torque towards it in fixed update?

Should I move everything to fixed update?

[timid dove]

How are you moving and rotating? And at what cadence?

The details matter

If you're modifying the Transform directly that's wrong to do anywhere for a dynamic Rigidbody

[pulsar canopy]

Torque and addforce, every frame

[tender vapor]

you should not addforce/addtorque with continuous forcemodes in the update loop

[timid dove]

It'd need to be at physics update cadence

You can use yield return new WaitForFixedUpdate() to do that

[pulsar canopy]

Oh

Didn't know that

Thanks

There is some jittering and overshooting

Might be out of sync

[pulsar canopy]

It's fixed update

[tender vapor]

you said every frame

[pulsar canopy]

every fixed update tick, whatever

[tender vapor]

these details are kinda important lol

[pulsar canopy]

I did say in the message above that I'm using fixed update

[tender vapor]

and my recommendation was in response to a message where you said it was not in fixedupdate...

[timid dove]

No you asked about a coroutine

Maybe if you just showed the code...

[pulsar canopy]

No, I'm moving something (by transform) via coroutine.

Then I have fixed update where some rigidbodies follow the stuff that moves in the coroutine.

[tender vapor]

ah, i see. you didn't make that particularly clear
but yeah that won't have too much of an issue. generally you'd want to have stuff happen in Update regardless, because that's when frames are rendered - that's what interpolation on rbs handles

with a transform you'd definitely want to be moving it in Update

[pulsar canopy]

Oh, that's useful. I should then move it to update and rewrite the stuff from the coroutine probably

Cause the occasional jitters I get might be from the lack of interpolation? (I use interpolation on the rigidbodies)

[tender vapor]

oh the AddForce should be in FixedUpdate

i meant there should be something in Update, not everything in update, i didn't phrase that well

(in the case of rbs, that's handled for you with interpolation)

[naive ferry]

Bonjour

[pulsar canopy]

oh lol so I can leave it as it?

cause that's how it is now

addForce and addTorque are in FixedUpdate

the source objects that the rigidbodies move to move via transform in coroutines

[timid dove]

it's technically going to have different behavior based on the framerate this way but whether that's a problem or not depends on the details of your game

[pulsar canopy]

doesn't deltaTime mitigate that? Or does it not work for coroutines?

[timid dove]

deltaTime works fine for coroutines but no, multiplying by deltaTime is not enough to mitigate the effects of having a non-fixed timestep when there are second-order physical effects in play

there's a reason the physics engine uses a fixed timestep

[pulsar canopy]

well I don't understand the problem if the rigidbodies move in FixedUpdate. Is it still an interpolation problem?

[timid dove]

think about where they are moving to

you are moving them towards the Transform object

since the Transfrom object is moving with a non-fixed time delta, its exact position per physics frame is going to vary depending on the framerate

if the object was moving with a fixed delta it would move to the same exact series of positions every time you do it

it's not necessarily a problem for your game, but your game will not be deterministic.

[pulsar canopy]

right but I don't think this should make a difference in my case, as it moves with a constant velocity

Basically, I don't see how it might cause issues

[timid dove]

it really depends. It could be the difference between the rigidbody hitting a corner and squeezing past without hitting it, or something

[pulsar canopy]

but if the rigidbody itself is interpolated, isn't that an impossibility? Cause I move them by addForce and rotate with addTorque.

Isn't that taken into account?

[timid dove]

rigidbody interpolation isn't relevant to the topic really. It's just a visual thing. It doesn't affect the motion of the Rigidbody at all

the actual physics happens with a fixed timestep

[pulsar canopy]

so I have a setup when I have an origin arm object that moves via animation + IK.

Then I have a puppet rigidbody object with configurable joints.

I made it so that the puppets move and rotate towards the origins via addForce and addTorque in Fixed update.

I was trying to solve some jittering there. There are always some jitters. I've tried various settings.

However, AI is telling me to simply get the rotations of the origin joints and update target rotation in the puppet joints, rather than adding forces. Is this correct? Could this reduce jitters?

[kind mauve]

is there an API to specifically toggle on/off collision between two colliders?

I think I found one to be able to toggle collisions off but I need the opposite

to describe the whole story I decided to make a jumping mob but then realized that it shouldn't be able to jump through targets

intuitively I could just make it collide with all it's enemies
but there are multiple factions and I am not sure if making a layer for each faction is a good idea
and if they wont' have own layers it would get stuck while jumping from within friendlies (agent without avoidance)

[tender vapor]

are you in 2d or 3d?

[kind mauve]

3d

[tender vapor]

ah, rip. if you were in 2d you could probably use the oncontactfilter thing

i guess you would use this then?
https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Physics.IgnoreCollision.html

[kind mauve]

yeah this one.... can only be set to ignore it seems

[tender vapor]

no?

[kind mauve]

I ll try, tho I read one forum post describing that this API couldn't help them

[tender vapor]

did you read the signature/docs for that method

[kind mauve]

it's ambiguous tbh

[tender vapor]

it really is not

[kind mauve]

"Whether or not the collisions between the two colliders should be ignored or not." like if it's not ignored does it mean that layers matrix still applies or not?

[tender vapor]

they do apply

it's just "is this ignored or not"

if it is ignored, no contact
if not, check further

[kind mauve]

so like
that's what I meant by
can only be set to ignore

there is no option to enable collisions ignoring matrix

[tender vapor]

it can be set to not ignore too

[tender vapor]

sounds like you'd want layer overrides instead

[kind mauve]

if that's the case what you said is wrong

[kind mauve]

no I don't

[tender vapor]

what part lol

[kind mauve]

no wait you weren't
I fail to pinpoint miscommunication spot

full story short I want to enable collisions between two colliders
IgnoreCollision don't seem to do that

unless I manually disable all unwanted collisions... which is kinda pita

[tender vapor]

yeah i think the overall philosophy is to filter out unwanted stuff rather than specify wanted stuff

[tender vapor]

i imagine you could get the control you want with this, but that's for 2d, and given the different engines the 3d engine may just not support that regardless of unity's decision

[kind mauve]

do you know perhaps a good way to imitate collision through code? assuming I don't care about y dimention and good precision
currently I am detecting the intersection through OnTriggerStay and do RigidBody.MovePosition... kinda rough, maybe I could polish that but is there a better way?

basically what I do now looks like a headcrab which only collides with it's target

[tender vapor]

probably a SweepTest? that would kinda mimic what MovePosition does
https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Rigidbody.SweepTest.html

[kind mauve]

no I mean I detect collision with a target and push away the mob which is unsurprisingly not working well

[kind mauve]

I mean borderline acceptable but I don't really know what am I doing and why so much jitter

public void OnTriggerStay(Collider other)
    {
        if(is_in_jump && other == target_collider)
        {
            float y = body.position.y;
            Vector3 direction_away = body.position - target.position;
            direction_away.y = 0;
            if (!is_dropping)
            {
                is_dropping = true;
                body.linearVelocity = new Vector3(0, body.linearVelocity.y, 0);
                distance = direction_away.magnitude;
            }
            distance += Time.fixedDeltaTime * step;
            Vector3 new_pos = target.position + direction_away.normalized * distance;
            new_pos.y = y;
            body.MovePosition(new_pos);
            Debug.Log("Teleported");
        }
    }

still jittery (interpolation set on extrapolate)

right, I am moving it each fixed update afterall

[pulsar canopy]

Instead of applying addForce and addTorque to a desired position and rotation for a configurable joint, I tried just setting target rotation(yes, I used inverse).

I figured it would achieve the same thing but maybe more stable. It doesn't, it just rotates to an unpredictable direction.

Is this because target rotation is aware of joint constraints so it tries to rotate in weird ways to reach thay position? While addTorque brute forces it and somehow manages to do it?

[drifting drift]

Hello! Im trying to make a sausage that behaves "realisticaly" (Basically, it shouldn't be a static capsule, it should be able to bend a little). I tried to make it by joining multiple capsules with configurable joints, but they easily clip through other colliders (apart from other weird artifacts when bending). The rigidbodies are set to interpolate and with a continuous dynamic collision detection, but they still clip. Is there a solution to this with/without joints?

This image is when the pan is static. If I try to move it it easily clips through

[timid dove]

how are you moving them?

[drifting drift]

the capsules in the sausage are rigidbodies, none of them is kinematic. I move the pan with a fixed joint. The pan is not kinematic either

@timid dove

[short sable]

I am using configurable joints and I am trying to make the arms floppy and not rotate too much. I want it to rotate but if you watch the video it rotates too much and looks weird how would I limit the rotation?

[pulsar canopy]

configurable joints are immensely complicated and I can't cover everything, but first you should start by setting the three angular motions to limited, rather than free

and then you define the limits to which they can rotate

[pulsar canopy]

there is a solution with joints too. Hard to tell from a picture. Maybe you can also try articulation bodies; they are better for chains

[drifting drift]

I was able to somehow fix it kinda with configurable joints. I've heard about articulation bodies, but I heard it works better if the root is fixed, which is not the case

[raven edge]

Hello!
I'm trying to figure out how to implement collide and slide (for walls only) for a kinematic character controller.
It mostly(?) works, but there's some strong jittering when I try to move directly against a corner.
The jittering only happens when I push into a spot between wall A and B in such a way that the slide pushes me back from A to B (and vice versa).
It stops as soon as the movement points await from the range that pushes me back into the previous wall.
I'm hoping someone could point out the (maybe obvious) mistake I'm making or give any kind of suggestion

Here's the code https://paste.myst.rs/9pyy3sah and a video of the jittering

[naive badge]

https://discord.com/channels/489222168727519232/1471501146014093387

I have been having this issue for the past few days. I tried thinking about how to solve this, but can't rly come up with a solution.

I genuinly don't know what the problem even is...

[raven edge]

I actually just found a possible solution, even if it doesn't fix 100% of cases (some small jittering still happens if using a gamepad to walk directly into the corner, but it stops after a few iterations)
Section F in this article is the source https://arxiv.org/ftp/arxiv/papers/1211/1211.0059.pdf

this is my implementation of it https://paste.myst.rs/vgmsq2vi

using a very small skin width also helps in this case

[timid dove]

You need to make your raycasting and sliding iterative

you're only doing one raycast

you're not accounting for if that slide brings you into a second wall

Check if your newly calculated trajectory hits any wall and repeat the process

until you run out of movement distance

or until you reach some present number of iterations

[raven edge]

isn't it what I am already doing? (lines 11-21 in the paste)

I capsuleCast from the new position towards the slide calculate in the previous iteration

[timid dove]

there are only 9 lines in the code you pasted

oh sorry

[raven edge]

I mean the paste to the message you replied to

[timid dove]

wrong paste, hold on let me look closer

[raven edge]

sure

actually, I can just paste the new version with what i was talking above

https://paste.myst.rs/oshpfixo

[timid dove]

I'm pretty sure this calculation was just wrong:

Vector3 movementInsideWall = movementDisplacement - movementDirectionToWall;```

actually nvm

[raven edge]

they should both represent the same unit
but it's 100% possible I did mess up some calculation

I'm not that good with doing vector math

[raven edge]

this + projecting on the crease works for the jitter, but this causes the character to sometimes get caught in corners that shouldn't visually stop it until the input direction points away from the second wall

[naive badge]

I checked it out but the thing is, this doesn't rly explain why more than one collisions aren't detected when I am manually moving the ducks. Cus if this jitter was triggered while using spline animate, there's no reason for this to not be triggered when I manually do it

[raven edge]

I didn't check out what your specific problem was before but, looking at it now, this might not apply to your problem

we're probably talking about 2 different things

[naive badge]

OOF

Ya, ur talking about a sphere colliding with a corner, ima talking about triggers being detected between a circle and a box collider

[raven edge]

Kept working on this and my guess is that walking inside the right wall slides the character into the up wall, which then slides it back again towards the first wall on the next frame.
I'd appreciate a lot if someone could point me in the right direction on how to solve this.
The current fix, as mentioned above, is to simply stop in a corner that is (arbitrarily) sharp enough, but it doesn't feel like a good solution since it causes a lot of sticking.
Here's the current code: https://paste.myst.rs/ee0mmmaq

[raven edge]

The sticky corners are most likely caused by the sweep (green lines) hitting the same wall until the input (blue line) is angled in such a way to the sweep completely points away from the left wall and only towards the front wall

[cold rivet]

I have a 2d phyisics material attached to a set of circles connected by a spring joint, yet despite setting the material to max friction, they slide around

https://youtu.be/egCqo8ekwEc

[tall rover]

So I’m trying to make a physics arm like one armed robber where the arm lags bevind and kind of like swerves around and it’s kind of working but I can’t seem to get it stiff I set spring to 125000 and the arm and bones in it still swerve around really hard, any help would be greatly appreciated

[upper helm]

is it okay to do a spherecast on a per-frame basis

my game isnt running that well already so im not sure if i should use it over raycasting, raycasting is buggy for me

[tender vapor]

it's probably not gonna be much different perf wise

compared to other stuff

[pulsar canopy]

What are your settings and what kind of joint are you using?

[tall rover]

Configurable joints

Between every bone

[pulsar canopy]

Ok but configurable joints have 5928582 settings, show them

You need to be a rocket scientist to set them up correctly

[tall rover]

What do you recommend me to do for a physics arm

I want it to kind of lag behind and bend when going into a wall

With some overshoot

[pulsar canopy]

I use mine fir joints but the purpose is a bit different.

So you want it to move towards a default rotations, but have some springiness?

You don't want to move it forward etc?

[tall rover]

Exactly

[pulsar canopy]

ok can you screenshot the settings, so I can see where to begin?

[pulsar canopy]

Ok guys, do you think it is possible to eliminate all rigibody jitter this way:

I have a ghost object matching the rb's rotation and position every frame. I let it detect erratic shifts in velocity direction by, for example, matching the rotation/position only if the rb has had a consistent velocity for the past 3 frames.

Basically something like that. Do you think 3 frames is enough? Does that idea sound legit?

[terse ginkgo]

If you're dealing with jitter, why not just use interpolation on the rigidbody?

[pulsar canopy]

I am

[terse ginkgo]

Imma be real you're probably doing something quite wrong then if you have interpolation on it and still dealing with noticeable jitter

[haughty parrot]

Hey, can anyone tell me how to import the fbx file of a cloth animation from Blender to Unity?

Problem Statement: The Alembic file is working perfectly. In the fbx file, model and animations are getting downloaded, but animations are not working in the scene, they are shown working in the animator, but in scene the cloth is not at all moving.

Can someone please let me know what is happening?

[tender vapor]

this is the physics channel, perhaps ask #🏃┃animation or #🔀┃art-asset-workflow

[pulsar canopy]

that's an insane thing to say

[terse ginkgo]

I'm saying it for the sake of you are potentially handling the rigidbody movement incorrectly if you are noticing jitter. It's not a jab at you, I'm just saying the setup is likely misconfigured

[random palm]

Can it be the camera movement that is causing the jitters?

[terse ginkgo]

If you're doing any transform movement for example, that'd very likely be a large cause of it

Can you show what you're doing perhaps?

[pulsar canopy]

I made a dual rig (like puppet master) where I have a source rig playing animations with ik and a puppet rig where the rigidbody joints try to match the position and rotation of the original rig with addForce and addTorque

It works pretty well, I added damping and everything, but there is jitter when constant collisions happen

there is no jitter when in mid air, but for example, when my origin rig has a wall grabbed, there is jitter in the final position (possibly due to constant collisions with the terrain)

[terse ginkgo]

When the character is animating and you are colliding with any object? Or only certain types of objects

[pulsar canopy]

just the terrain

to be clear, the animated rig doesn't collide (it only raycasts), the rigidbody following rig collides

[terse ginkgo]

Interesting, I mean using heavy forces with joints is likely to cause jitter so I do apologize for the original misinterpretation, I thought you were using a kinematic rigidbody or something of the sort. The most ideal way to work with joints is through direct target rotation / target angular velocity but that goes quite deep into config joint knowledge haha

[pulsar canopy]

I actually tried just that but it didn't work for me

it always tried to rotate to a rotation I didn't specify

and I know joint space is different from local space is different from world space

I basically did double Quat inverse to specify target rotation but it just didn't work. I don't know

[terse ginkgo]

Yeah it goes deep, I've run down this rabbithole and do it for my hobby projects and day job, I would not recommend it lmfao

[pulsar canopy]

but what makes it a rabbit hole? It's a fkn Vector3. What makes it so hard?

[terse ginkgo]

The edge cases you have to account for :P

This may be somewhat of help https://gist.github.com/mstevenson/7b85893e8caf5ca034e6

[pulsar canopy]

I did something exactly like the first method shown there

Wait, could it be because the target rotation is a vector3, rather than a quaternion?

[terse ginkgo]

target rotation is a Quaternion

target angular velocity is a vector3

[pulsar canopy]

Ah yeah, it's just shown as a vector3 in the inspector

[terse ginkgo]

all quaternions show like that yeah

[pulsar canopy]

Yea yeah, true

Idk man

What do you think about my third rig idea?

[terse ginkgo]

Elaborate?

[pulsar canopy]

The third rig would just be a skinned mesh with nothing attached and would match the position and rotation of the puppet rig.

However, it would wait for a few frames to detect erratic changes in velocity of the rb rig. For example, if dot product of frame 1 movement and frame 2 movement is low, don't update position and rotation.

If dot product is high for 3 frames in a row, only then update. Does that make sense?

[terse ginkgo]

Yeah I understand what you mean, but I feel eliminating the jitter entirely would be a better alternative. Trying to drive the joints directly could be good but also difficult ofc

[pulsar canopy]

I mean I've really tried that but couldn't

I wish I could

[terse ginkgo]

Hm got it, then you could attempt what you're talking about, you could also try bumping up your physics project settings solver iterations, or anything else to help with joint slop

[pulsar canopy]

I have increased all of that. I mean, I could try increasing it more and more but there's a point where it becomes unreasonable

it may also just be due to the fact there are many colliders

[terse ginkgo]

Yeah, anything above 32 for a single rigidbody is overkill

[pulsar canopy]

I have joints on each of my fingers (yeah I know)

so it may be impossible to achieve stability with just rb

[terse ginkgo]

That will create joint error for sure

You can try simply boosting your inertia tensors too

[pulsar canopy]

I actually have 3 joints per finger lmao

I considered switching them to an articulation body but they either don't work or idk how to use them

[pulsar canopy]

you mean drag and angular drag?

[terse ginkgo]

That'd prob be a good idea

[terse ginkgo]

No, inertia tensor, in 2022+ it's a setting directly on the RB, else you have to do it in code

You could just simply multiply the whole vector by 2 as a test

[pulsar canopy]

I didn't even know abou that

I will check

[pulsar canopy]

but they can't be parented to RBs

essentially I'd need to make another rig for fingers that follows the rb rig

but I tried doing some articulation body tests and they didn't seem to work, idk

[terse ginkgo]

I mean, if you so want you could build your entire rig out of articulation bodies

[pulsar canopy]

nah, the player himself is a rigidbody. I would need to redo everything and I don't like that

[old abyss]

hey, sorry for the ping out of nowhere. i was skimming through old messages to see if anyone's experimented with the limits of determinism in unity. Seems like you did just that. Did u get anywhere with ur tests?

[void hemlock]

Anyone know which version of physx Unity uses?

[timid dove]

Depends on which version of unity

[void hemlock]

6.3

[old abyss]

I think its 4.1
https://www.reddit.com/r/Unity3D/comments/1q138sx/what_physx_version_is_unity_63_running/

[void hemlock]

Yeah not exactly shouted out by Unity is it! I think they are well behind as maybe license changed for v5

[unique cave]

It seems they changed from physx 3 to 4 already in 2019 and haven't changed since

Back then I think they had a blog post about the changes at least

[void hemlock]

sounds like PhysX has much improved and has 3d wheel colliders now so don't know what happened but maybe is a big leap to change to a different physics setup

seems v5 was a major change in direction to simulation and GPU

[unique cave]

I assume it's quite much to integrate, still bit surprised they haven't upgraded yet. Maybe their physics teams have been busy with DOTS physics stuff etc.

[void hemlock]

Yeah Unity Physics is annoying to me as built for ECS/Dots (do not use) and may be a long project to complete

[dull jungle]

its more about being scared of breaking changes, they have a history of tiptoeing around improvements cause it could break legacy projects on upgrade, on one hand its nice to have a relatively simple upgrade most of the time, on the other hand it stops progress right in its tracks sometimes and led to things having 3 different implementations at a time cause they didnt want to remove the old one

Well beats removing the old one before the next version of it is remotely ready 😄 (looking at you, Unet)

[gritty plank]

ooo i will be reading down this thread

[old abyss]

they're using 4.1 for wheel physics as can be seen stated here explicitly: https://docs.unity3d.com/6000.3/Documentation/Manual/wheel-colliders.html
I'm not sure if they're using just 4.1 vehicle sdk for wheel physics and stuff and the rest of the physics engine runs on some other version, but i just assumed that it's 4.1 overall :/

[terse ginkgo]

6.3 is PhysX 4.1.2

[signal wyvern]

Anyone know the differences here between these 3 ways of setting my boats position & rot to the server Authed state before resim? The top one directly setting transform has worked best for me but it comes with a bit of a performance cost as all colliders need to be re-synced, from reading the other two are maybe what i should be using and seem to be cheaper but they ruin the re-sim somehow as my boat become super jittery and clearly don't align

[tender vapor]

the 1st one is "incorrect" for physics, as it causes a desync
the 3rd one is the "correct" way to do the 1st one, it's a teleport like the 1st but it maintains sync

the 2nd one does a sweep to the new position, simulating collisions in between, rather than teleporting

iirc for one of (2d, 3d) but not the other, MoveX is only interpolated correctly for kinematic rbs (as that's what it's intended for), that may be the cause of the jittering

if this is for "resim", do you even need physics? couldn't you just set the value every frame/tick manually (or maybe with kinematic rbs to get interpolation, if you're using the fixed time step)

also,

!code

[gloomy lodge]

Posting code

📃 Large Code Blocks

Use links to services like:
https://paste.mod.gg/, https://hastebin.skyra.pw/, https://paste.ofcode.org/, https://paste.myst.rs/

📃 Inline Code

Surround code with three backquotes. Not quotation marks.
To format as C#, add cs to the first line:
```cs
// Your code here
```
Add a comment with a line number if there is an error message.

[tender vapor]

that's a very small screenshot that's hard to read

[signal wyvern]

Hmm, okay well it seems like the 3rd one would be the one i want however it produces different results compared to 1st one, any ideas why this would be the case? id love to get some of that performance benefit.

[signal wyvern]

Yeah i kinda need physics.

[tender vapor]

is the jitter just due to network desync of the simulation

is the rb kinematic or dynamic?

[signal wyvern]

Probably better to show you, hold on

Its NOT kinimatic

would breifly setting it to kinimatic first maybe solve it?

[tender vapor]

am i understanding this right that you have some simulation on the server that you want to sync to the client?

[signal wyvern]

Yeah essentially Server and client are running physics simulation for my full client prediction setup, server sends snapshot of world to my client, client then updates everything with these stale values and resims frames to catch back up to updated position

[tender vapor]

are these snapshots every physics tick or less frequent?

[signal wyvern]

1st video uses Transform.SetPos&Rot(Nice & smooth but costs performance), 2nd video uses Rb.position/rot =(Jittery, but better for perfomance)

[signal wyvern]

as you can see the rb.position set causes the boats resulting position to be wrong compared to servers, not every tick, server sends snapshot to client every 5 fixed frames.

Clearly rb.position results in a slightly different outcome compared to directly setting transform, does rb.positon update the position of the object immediately? or does it require another physics step before its truly updated to this position

[tender vapor]

If you change the position of a Rigidbody using Rigidbody.position, the transform will be updated after the next physics simulation step
apparently yeah, it takes a step

if you're doing this in FixedUpdate, the step should be immediately after though

[signal wyvern]

Due to the nature of the simulation im running physics in script mode, that will explain the issue then

is there any way i can force it to update immediately, or does this require running physics.simulate()?

[tender vapor]

im running physics in script mode
what do you mean by that? i don't think ive heard of that

[signal wyvern]

You can set physics.mode to script which means physics does not automatically run after fixed update, physics must be explicitly called using "Physics.Simulate(Time.fixedDeltaTime);" This allows me to preform multiple frames of physics during a single frame

[tender vapor]

huh, interesting

[old abyss]

rb.position != transform.position

im assuming your boat is a rigidbody and you're trying to set it's position and rotation and want that change to be respected by the physics engine, in which case you should be using rb.move https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Rigidbody.Move.html or rb.moveposition if the boat is kinematic https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Rigidbody.MovePosition.html

in your case where your physics is running in script mode, you can use PublishTransform() after modifying the rb pos and rot directly for the rb and transform to sync

[tender vapor]

is Move not just a shortcut for MovePosition+MoveRotation? i assumed it would be since other "move pos + rot" methods say so but i'm just noticing now that Move doesn't say kinematic in its docs

[old abyss]

rb.move is different from transform.moveposition and transform.moverotation

[signal wyvern]

This could be promising, i was just doing Physics.Simulate() after setting the rb.position and that seemed to work but obviously required me to run an extra step with everything kinimatic

[old abyss]

rb.move and rb.moveposition are kinda similar. Ive used both on kinematic rbs and non-kinematic rbs. Didn't see any difference

[signal wyvern]

does publish.transform() need the rb to be kinimatic?

[old abyss]

i dont think so no

PublishTranform() just syncs rigidbody and tranform pos and rot

[tender vapor]

i'm referring to rb.Move being rb.MovePosition+rb.MoveRotation

[old abyss]

the docs dont explicity state that rb.move is to move non-kinematic rbs, but they do state that rb.moveposition is for kinematic rbs

but ive tried rb.move on kinematic and non-kinematic rbs and rb.moveposition on kinematic and non-kinematic rbs and i didnt see any difference

is your interpolation setting set to none or interpolate or extrapolate?

[tender vapor]

i don't have any rbs to test right now, but from what ive heard from helping here, MovePosition breaks interpolation on dynamic rbs (though, that might be for 2d) so i'm curious as to whether Move has the same behaviour

[old abyss]

rb interpolation is frame rate dependent. im not sure how it affects multiplayer setups but interpolation can give different results on the same machine since every run would be at a different frame rate

move and move position respect the interpolation settings btw

says so in the docs

[tender vapor]

empirically, MovePosition does not respect interpolation on dynamic rbs

i'm not referring to this specific multiplayer set up. i'm referring to the behavior of rb.Move in isolation

[signal wyvern]

Thank you both, Think ive made some good performance gains there ❤️

[kind mauve]

What's the good practice when it comes to making colliders for a whole level (low poly but not dumb simple geometry?
Inserting many many boxes?
Using meshes as colliders?
Making and importing simplified meshes of level geometry and using them as colliders?
Also how bad is the idea to make a huge bounding box alike collider as one mesh?

[dapper arch]

mesh colliders are very expensive if they are not convex, because you basically need to check the collision for each triangle. non convex also can't be simulated by rigidbody etc. it's usually better to just use many box colliders, as it's a simple AABB/OBB test

[kind mauve]

so like to make a cupola roof.... I meant to make it from pieces?

[dapper arch]

yeah, just use primitive colliders for that. the complexity will depend on your needs - if it's a parkour object, it will probably require more details in terms of collision because it can be interacted with.

[kind mauve]

tbh this is confusing because it feels like in every game there is many mesh colliders everywhere

[timid dove]

Static mesh colliders are generally fine, as long as there's not too many of them in one place. But you want to generally use a simplified mesh and not the same one you use for rendering.

It will also of course depend on the complexity of the scene and the target hardware

[kind mauve]

what will happen if an object happened to fall off the map

do I expect some crashes or issues due to some overflows or I can just safely ignore that that may happen?

[dapper arch]

as far as i know, unity will just throw an error if you have a rigidbody on an object with non-convex mesh collider and won't simulate properly

it won't crash the editor

[kind mauve]

it was new unrelated question about

indefinite time freefall of a rigidbody

[dapper arch]

oh, sorry. thought you're talking about the mesh collider :P

if an object falls out of the map, it will eventually fall so far from the origin that there will be problems with floating point precision because of how far the object is. generally you'd have some system that would just destroy the object if it falls out, but that depends on your game

[kind mauve]

so I actually have to take care about it, huh

[dapper arch]

yeah, that would be the solution. i guess the object is not that important if you let it fall of the map?

[kind mauve]

yeah I don't mind them being deleted

of course I assume such can only happen during some extreme physic interactions

[dapper arch]

yeah, e.g. if the player has the ability to carry objects (like in R.E.P.O.) and there's a hole in the map

[kind mauve]

but like is there a game where a prop not suddenly getting mach 10 speed and fly through something

I don't trust those terrain/mesh colliders to be 100% impenetrable

[dapper arch]

i mean, if an object is moving extremely fast then it can just phase through another collider. rigidbody's continous collision mode does kinda solve it, but i'm not sure if it's that reliable

you'd usually set the collision detection mode to Continous if an object is very important, e.g. like puzzle elements in amnesia the dark descent (i remember when a pipe just fell through the floor and i had to load last save 💀 )

[kind mauve]

as I read about collision modes it's either fail to detect phasing through or fails to detech phasing through if big angular speed and weird trajectory

sometimes, given high speeds, but anyway, nothing is 100% impenetrable

[dapper arch]

true. i think you can also make a linecast between the last pos and the current pos - to check if the object didn't just penetrate through something

[kind mauve]

gee I guess

maybe worth it for puzzle elements

but in general I discuss about some random minor items

[dapper arch]

then i don't think why it would be worth the struggle. this linecasting thingy can get very expensive for many items

[kind mauve]

I would expect unity to outright delete objects extremely far or something

but I guess I have to do it myself?

[south granite]

i mean yeah

what if you dont want it to do that

[kind mauve]

toggle off some project setting 🤷‍♂️

[south granite]

now imagine how many of these kinds of problems maybe exist

[timid dove]

That's just... such a bespoke thing that it's not worth being part of the engine

[south granite]

and how many toggles would exist

[timid dove]

The engine should not take opinions on most things

[dapper arch]

it's easy to implement, e.g. every N seconds just check the dist from 0,0,0 and destroy the object if it's over some threshold.

[south granite]

respectfully thats an insane solution

just chuck a trigger collider down there

[timid dove]

It's part of why I dislike Unreal engine. Unreal practically begs you to be making a first or third person shooter game, and you have to start changing settings to make any other type of game.

[kind mauve]

I can't, luckly I havent sumbled upon anything like that yet

[kind mauve]

But What If The Thing Phases Through It

[south granite]

dont make things that fast

[kind mauve]

it just sound like the thing that may cause a savebreaking situation one time out of a billion or something which is still bad

[south granite]

dont worry about one time out of a billion

[kind mauve]

anxiety noises

[frigid pier]

For fast things there's a continuous collision mode for the Rigidbody.

Actually several of those

[kind mauve]

I wish I had a link for that Unity's(?) page comparing them
even dynamic one might ignore angular collisions
I suspect that may lead to some object getting stuck in geometry and potentially falling out

[frigid pier]

Long time ago when there was just one continuous collision option and I was making a bricks game it wouldn't be reliable enough at high ball speeds, even with very thick border colliders, so I had to handle it myself placing ball on the last position it would've been when it went through collider.

[kind mauve]

I mean the current one might be good enough yet not impenetrable and inevitably in a usual goofy game with props and stuff setting where rigidbodies even sometimes spawn inside each other or some random stuff like that some thing will get outside

or at least this is what I believe

[silver moss]

Continuous speculative uses rotations:
https://docs.unity3d.com/6000.3/Documentation/Manual/speculative-ccd.html
Continuous and Continuous Dynamic do not use rotations:
https://docs.unity3d.com/6000.3/Documentation/Manual/sweep-based-ccd.html

[kind mauve]

exactly that page, yes

so what if we got a compound collider on a sweep based detection rigidbody flying into a wall with a high angular speed?

I suspect one of colliders might get through and something terrible will happen leading it to phase through by one end or the other

[silver moss]

Probably not robust

The rotation stays the same during the sweep

[runic torrent]

Ok, so let's say I have a camera with a FoV of 60 degrees rotated along the x axis at 30 degrees. I want a given subject's transform point to appear at a viewport point of (0.5, 0.3). How would I go about calculating where to place the camera?

I'm not sure I worded this well, so here's a diagram

[timid dove]

This isn't a physics question, but the answer is to just use cinemachine - the viewport offset you're asking about is something you can just set up with a few clicks

[runic torrent]

I wasn't sure where else to talk about things related purely to math

Also I'm not sure how I would use Cinemachine that way

[timid dove]

Position Composer | Cinemachine | 3.1.5 https://share.google/3sydWqDxktwjncrj7

Literally just by setting the "screen position" property in the inspector for the Position Composer

[runic torrent]

Can it be done in code? Because I'm trying to set this up at runtime

[timid dove]

It could be but there's probably not a reason to

Unless you need to adjust those values at runtime a lot

But yes anything can be done in code

[distant oasis]

so im using rigidbody to make this item can be thrown but on specific scene this object somehow broken and before i think it because mesh collider but i copy the area to testing scene it somehow not broken? is im doing something wrong in my game scene using the same rigidbody setting it just bypasse the collider and fall

[timid dove]

not enough information here

but also - moving it around via the gizmo in scene view at runtime isn't really a test of anything

[dry shell]

Heyy SUPER new noob question. I'm trying to learn about different joints and I'm watching a youtube tutorial for it. In the video the guy sets up a fixed joint with one cube connecting to another one. With the fixed joint, he's able to grab and drag one of the cubes to make them both move.

When I added the fixed joint, I am not able to make the both move, but when I got to play mode I can see the cubes fixed and hovering. How is he able to make these physics active while in Scene view?

[timid dove]

joints are only active in play mode (the whole physics engine is)

So you would enter play mode and look at scene view during play mode

[dry shell]

Ohhh! It's that easy??? thanks man

[timid dove]

you can even set up your editor so that you can see game view and scene view side by side at the same time

just drag the windows around however you want

[upbeat lantern]

i wish unity adds a cylinder based character controller by default instead of making a custom one using rigidbody

[twin bramble]

i mean making a custom one is easy

[frozen timber]

working with a custom 2d physics engine, none of the scripts have any code in OnEnable or OnDisable, but a bug i have gets fixed when i disable and enable the gameobject with the bug. So basically lol im just trying to figure out exactly what does unity do internally during these events.

[timid dove]

You'd have to share more details.

[barren gorge]

why do you need cylinder?

[upbeat lantern]

there is few source projects i was working on that had multiple issues with some jumps not being possible unless you adjust the map or just make a custom collider

[upbeat lantern]

yeah true its very easy to do

[frozen timber]

its ok i did a bandage fix i will never know the cause

will have to live with it

[pulsar canopy]

A collider or controller? Because you can't use anything other than capsule as your collider

[twin bramble]

you cant?

i mean why would u want to anyways

[pulsar canopy]

Not everyone is making the same game

I had to use a horizontal capsule for my game, so I had to switch from char controller to something else

[deft barn]

Hello i am creating a billiards game in unity and i am having some problems seeing realistic break physics. I have tried plenty of friction, bounciness material combinations however it just doesnt look right. Do you think this is a limitation of the physX engine and should i rather try implementing my own simulation physics? I can also provide footage how it looks now if it would be of any help.

[timid dove]

Just don't use physx honestly

Billiards physics is pretty simple you'll be better off with custom physics.

[deft barn]

How should i go about implementing something like that. Because i found an article that seems promising (https://ekiefl.github.io/2020/04/24/pooltool-theory/). I really never did something like that myself so yeah.

[timid dove]

All the objects involved are spheres, which are the simplest possible shape. You can even continue to use physx for the colliders and spherecasts and just do the motion part yourself. It's all just linear motion and vector reflection off collision normals

[deft barn]

so this would mean using unity's On collision events and then manually calculating the movement based on current linear and angular velocities?

[deft barn]

Seems like moving the object while still using the collision system can be achieved with the Rigidbody.MovePosition() method. Any thoughts on using this?

[timid dove]

No

Spherecasts

[deft barn]

I am sorry i dont see how i would use a sphere cast in this way. Would this run in realtime or would i have to precalculate all collisions?

Or is every ball gameobject sphere casting on itself and seeing if it hits another ball?

[timid dove]

you would run it every fixedupdate

to see if there will be any collisions

[deft barn]

so you are checking to see for next frame what happens

sorry my bad every timestep right cuzt this is not the regular Update method

Wouldn't this get really resource intensive calculating every next position with fixedUpdate for every ball in the scene?

[timid dove]

not really. There's only like 15 balls right?

[deft barn]

yes

[timid dove]

it's fine

you can even do them all in a batch with the job system

https://docs.unity3d.com/6000.3/Documentation/ScriptReference/SpherecastCommand.html

[deft barn]

Okay so with this system i calculate a sphereCast that is where the ball will be next time FixedUpdate is called. This means that i should calculate a position, rotation every [timeStep] seconds, having it as a function of time?

[upper helm]

is AddTorque only visual?

if not what does it change on a rigidbody

[timid dove]

AddTorque changes the angular velocity of a Rigidbody

It's the rotational equivalent to AddTorque

Not sure what you mean by "only visual"

[dry shell]

Can someone help me find out why my dude just falls through the map?

I'm following a ragdoll physics tutorial. Each piece of mesh has a mesh renderer, rigidbody, and a joint connected to the relative part of the body. The cube I have stays but not my model. I also put a rigidbody and mesh collider on the full model to try and stop this but it didn't work.

[frigid pier]

With physics based movement you need to use physics methods properly with FixedUpdate frames, so it won't get pushed through colliders and interact correctly. See physics movement tutorials

[timid dove]

You cannot use a concave mesh collider on a dynamic rigidbody

for an animated character especially you should not use a mesh collider at all. Generally a character like this uses a capsule-shaped collider for the basic world movement/interaction

[pulsar canopy]

Let's say that I have a rigidbody sword and a rigidbody axe split into handle and blade part.

I make the axe more top heavy due to the blade having more mass and I make the sword bottom or mid heavy.

Now, if I apply torque with the hilt being the pivot, does physx understand that the axe applies more force on impact, but takes morw torque to spin?

Is this something physx thinks about?

[timid dove]

I make the axe more top heavy due to the blade having more mass
How did you do this exactly?

[dry shell]

Awesome thanks for the heads up, I'll look into FixedUpdate frames as well

[timid dove]

How many Rigidbodies are involved?

If it's a single rigidbody the only way you can distribute mass is with the shapes of colliders. And all colliders are assumed by the system to have uniform mass.

What you can do is set the center of mass manually by disabling "automatic center of mass"

[pulsar canopy]

I reiterate, I split is into two rigidbodies, hilt and blade. Blade has more mass in the case of the axe.

I might need to unparent it and use fixed joint and all that, whatever

[timid dove]

ok you didn't mention multiple rigidbodies before. YOu will need to connect them with joints yes

and the yes the masses will be taken into account when applying torques and forces

[pulsar canopy]

I did mention two parts. I guess I didn't specify they are both rbs, I thought that was imied

[timid dove]

the parenting doesn't make a difference if they're dynamic bodies

[pulsar canopy]

That is cool tho

[pulsar canopy]

Doesn't the transform interfere with force propagation?

[timid dove]

the physics engine basically will ignore the hierarchy. Joints are the only way to connect multiple dynamic RBs

it may interfere/fight with it. It's best to not have them in a parent/child relationship ideally

What's the end goal here though? I expect this is probably more complicated of a setup than you need/require

[pulsar canopy]

Yes, that's what I did (unrelated to this, but also with joints and rbs)

[timid dove]

a single rigidbody with a custom center of mass may be a better solution

[pulsar canopy]

I'm just spinning up a new game idea with actual physical weapons and forces

[pulsar canopy]

No, that's not a proper solution

[timid dove]

I mean it depends on what the game actually needs

[pulsar canopy]

Shifting center of gravity would work in the case of a symmetric staff, but not an axe or mace

[timid dove]

why not

[pulsar canopy]

Oh, I could shift it more upward the heavier the top, right?

That might result in the same thing, yes

[reef stratus]

have anyone ever had this issue? all i do is spawn the car and sometimes wheels dont function

[silver moss]

I agree that center of mass is the way to go here. Having two RBs and a joint would likely not always be stable for something that needs to be firmly connected. Even a FixedJoint isn't always reliable

[timid dove]

make sure you're using continuous collision detection

[reef stratus]

I set it on vehicle's rigidbody but it didnt work

wheels dont have a field for that

[timid dove]

it's a Rigidbody property not part of the wheel collider

[silver moss]

Wheel collider mass 1000 seems really high

Yeah docs recommend 20-80:
https://docs.unity3d.com/6000.3/Documentation/ScriptReference/WheelCollider-mass.html

And the unity car tutorial uses a Rigidbody mass of 1500

This one
https://docs.unity3d.com/2020.1/Documentation/Manual/WheelColliderTutorial.html

Oh the tutorial changed quite a bit for newer versions
https://docs.unity3d.com/6000.5/Documentation/Manual/WheelColliderTutorial.html

[viral beacon]

I built a rigidbody extrapolator and made a camera follower for my car, however Im observing an inconsistent jitter between plays. It seems to be consistently smooth on the first compile, but repeated plays seem to get weird jitter. Is there a common cause for this?

Left video is almost perfectly smooth. Right video is jittery. No changes made to code or scene between both.

[timid dove]

without knowing any detials about your project, the rigidbody configuration, or your custom code, it's hard to say anything. My guess is it's framerate dependent.

Is there a good reason you don't use the built in Rigidbody interpolation/extrapolation?

[viral beacon]

the built in extrapolation drastically altered the simulation and caused massive issues

(my car started doing wheelies)

[timid dove]

What about interpolation?

Afaik neither is supposed to change the simulation at all

[viral beacon]

Its a car game so delay is kinda catastrophic

I think it has to do with the fact that the built in extrapolation/interpolation directly affects the rigidbody position, I would prefer to have these processes be cosmetic only

[timid dove]

I think it doesn't affect the RB position at all.

It affects the Transform position

[viral beacon]

I can show ya what happens when I turn on extrapoliation

[timid dove]

So maybe if you have code or logic depending on the Transform that could be the issue

[viral beacon]

I mean I do raycasts so I guess that qualifies

[timid dove]

You can do the raycasts based on the Rigidbody instead of the transform

[viral beacon]

interesting proposition

ill try it

hm, I am doing these raycasts from a child object, do I need to do rb.transform.position+transform.localposition?

[timid dove]

don't do .transform at all

that's the issue in the first place, remember?

[viral beacon]

right I meant rb.position, because I have a rigidbody on my vehicle object, but I need the offset of the wheel which is where im doing these casts from

[timid dove]

If you want the world space position of a child object you would do this:

Vector3 worldPos = rb.position + rb.rotation * transform.localPosition;``` but this assumes you have 1,1,1 scaling too

[viral beacon]

ill give it a go and let you know whether unitys extrapolation works better or if my issue persists

[viral beacon]

I replaced everything accordingly, but now I cant get the original bug to occur with my own interpolation to compare them

however regular extrapolation works now

so you were right, thanks

[dapper crescent]

Is there anyone who explains realistic Unity car physics? I have a few questions I’d like to ask.

[dapper arch]

it's best to just ask here so that everyone can see what you exactly need. someone with knowledge on the topic might help you

[tender vapor]

please don't crosspost

[dapper crescent]

ok sorry

[rugged kindle]

I’ve got a flick/plunge mechanic working, but it just doesn’t feel right. Weight feels off, launcher strength feels inconsistent, and endless tweaking isn’t fixing it. I even added some overlays to the game where I can tweak the values while playing.

Single dynamic shape, kinematic launchers, static rails. Using PhysicsMaterial2D on the dynamic shape with bounce and friction values.

At what point do you simplify / restart vs keep tuning? What would you check first? I feel like I am missing something important. Could the scale of my objects/scene be the issue? I didn't really consider that when building it out.

[rugged kindle]

Is it true that Kinematic to dynamic energy transfer in Unity 2D is not a stable “power delivery system”? That's what my research has been telling me. Should I instead use a impulse-driven launch using AddForce?

[bold aspen]

I'm in Unity Alpha, seems like 3D physics has a bug: When using any IContactsJob and there's a dynamic body, if there's no AABB contacts, the system floods with errors:

ArgumentException: Actual value False does not equal expected value True.
Unity.Physics.SafetyChecks.CheckAreEqualAndThrow[T] (T expected, T actual) (at ./Library/PackageCache/com.unity.physics@b51efcbbfd90/Unity.Physics/Extensions/SafetyChecks.cs:143)
Unity.Physics.IContactsJobExtensions+ContactsJobIterator..ctor (Unity.Collections.NativeStream+Reader reader, System.Int32 forEachIndexBegin, System.Int32 forEachIndexEnd) (at ./Library/PackageCache/com.unity.physics@b51efcbbfd90/Unity.Physics/Dynamics/Simulation/IContactsJob.cs:406)
Unity.Physics.IContactsJobExtensions+ContactsJobProcess`1[T].Execute (Unity.Physics.IContactsJobExtensions+ContactsJobData`1[T]& jobData, System.IntPtr additionalData, System.IntPtr bufferRangePatchData, Unity.Jobs.LowLevel.Unsafe.JobRanges& ranges, System.Int32 jobIndex) (at ./Library/PackageCache/com.unity.physics@b51efcbbfd90/Unity.Physics/Dynamics/Simulation/IContactsJob.cs:336)

[timid dove]

report it with the bug reporter!

[bold aspen]

IContactsJob has the following code:

                    int forEachIndexBegin = 0;
                    int forEachIndexEnd = jobData.ContactReader.ForEachCount;

                    if (jobData.IsParallel)
                    {
                        if (!JobsUtility.GetWorkStealingRange(ref ranges, jobIndex, out forEachIndexBegin, out forEachIndexEnd))
                            break;

#if ENABLE_UNITY_COLLECTIONS_CHECKS
                        JobsUtility.PatchBufferMinMaxRanges(bufferRangePatchData, UnsafeUtility.AddressOf(ref jobData), forEachIndexBegin, forEachIndexEnd - forEachIndexBegin);
#endif
                    }

                    var iterator = new ContactsJobIterator(jobData.ContactReader, forEachIndexBegin, forEachIndexEnd);

And the ContactsJobIterator has the following cctor:

            public ContactsJobIterator(NativeStream.Reader reader, int forEachIndexBegin, int forEachIndexEnd)
            {
                SafetyChecks.CheckAreEqualAndThrow(true, forEachIndexBegin >= 0 && forEachIndexBegin < forEachIndexEnd && forEachIndexEnd <= reader.ForEachCount);

When there are no broadphase pairs (no AABB overlaps), the contacts NativeStream is allocated with ForEachCount = 0. In the non-parallel path, this means forEachIndexBegin = 0 and forEachIndexEnd = 0. The check 0 < 0 evaluates to false, failing the assertion.

This can easily be checked with the debug system, which uses the Interface (DisplayContactsSystem) and the Kinematic Character Controller, which is how I found the error, in DisableCharacterDynamicContactsSystem. Disabling both systems which schedules jobs makes the log stop happening

[silver moss]

Dunno about others but I personally don't have a clear image of the mechanic you are creating or the issues it has. Maybe a video would help demonstrate it

[rugged kindle]

To simplfy things.... When is a case you would want to use kinetic like a plunger hitting a ball driving the physics, vs using addforce to actually move the ball and just making it LOOK like the plunger is hitting it. Or does everyone just use addforce in these cases?

but yes I can make a video today to show you what I am talking about. I am going to try to implment addforce version now to see if that works the way i think it will.

also this is in regards to 2d physics.

[timid dove]

you mean like pinball? I would 100% just do that with an AddForce call and have the plunger be visual only

[rugged kindle]

pretty close to a pinball plunger! ok i have been trying to use AddForce for a few hours now and things are looking up.

One thing that is not related but still physics.... How in the world do i use a vector file as my collision map? I can't seem to find anything in the docs or any tutorials or anything. Saving my svg collision map as a png and then generating the collisions in unity produces the stair step effect on straight lines. In my vector its already exactly the way I want it

[timid dove]

the 2D physics engine doesn't support colliders defined as vectors

[rugged kindle]

but i can't generate a collider from SVG data?

[timid dove]

you can generate a sprite from SVG using the vector graphics package and generate a collider based on the sprite, but no it does not support SVGs directly