⚛️┃physics

[drifting sleet]

if the player choses a narrow opening on the ceiling

that is too narrow to fit the player, but tall enough

navmesh will not help you

the correct answer is to simply not design narrow openings like that

unity cannot help you with that in the general case, you can only heuristically solve that

A* Pathfinding Project will let you do one such heuristic, which is to voxelize the space

and query that grid

no

navmeshagent cooperates with character controllers / is a substitute for character controllers

it really depends on what kind of gameplay you want to achieve

you can always move around without navmesh

it was just, if you want to teleport

yes

[golden star]

So Object A has a 2d circle collider, and object B has a 3d box collider. Object B has IsTrigger checked on, will collider events in a script on B activate if Object A collides with Object B?

[formal trail]

hi, how do i check if a collider is convex?

I tried

MeshCollider mcder;
mcder = otherObject.GetComponent<MeshCollider>();
cvex = mcder.convex;```
but I get `The name 'convex' does not exist in the current context`

[formal trail]

nevermind, found a workaround

[uncut mantle]

Hey guys, I'm having some issues with the rigidbody rotation constraints. I've got an object with a rotation constraint on the Y axis, but when it collides with something, it'll still rotate on that axis. I've made sure it isn't happening cause of a script, but it even happens with 2 default cubes colliding. Was wondering if anyone else had this issue before?

[autumn jetty]

Anyone know if there are any performance implications of moving static entities in dots?

[drifting sleet]

@golden star the 2D and 3D physics system do not interact with each other, so no collision methods will be called

@uncut mantle i'm not sure, but maybe try a different constraint (like X or Z) and see what happens. try checking them all

@autumn jetty usually with the way DOTS is engineered, you have to explicitly ask for anything that has a cost. so my expectation is that moving static entities is free on DOTS, but actually using them in their updated location requires a call to something that has a cost proportional to the number of static entities in the entire scene and the number moved

[golden star]

Thank you, one of the objects in my game uses a cube for the visual model, however I would like to use a 2d collider. I have rotated the cube's transform to get it to look a certain way, is there a way to rotate the collider so it is aligned with the other 2d colliders in the scene? (aligned to X/Y axis)

[uncut mantle]

@drifting sleet The constraints only seem to properly work when all of them are active. Once I try to constrain 1 rotation axis, all of them keep rotating

I guess I can just overwrite it with a script 😦

[drifting sleet]

i think it's probably a physics simulation quality setting @uncut mantle

@golden star you can use a 2d collider with a cube that rotates, simply attach the 2d collider to a parent game object, and make the cube the child

[golden star]

Thank you

[silent mural]

Hello people. i got myself a nice 3d cockpit amatur. now the problem is that i want its button to be pressable, but it turns out onmouseover doesnt work if any of its parent has a rigidbody

but my cockpit needs to be part of the ship and the ship needs a rigidbody. suggestions?

Solved it by myself. The button needs its own rigidbody, set to kinematic so it doesnt move away

google says the parents rigidbody overwrites all colliders childed to it that dont have own rigidbodys. Kinda helicopter mom. would have been nice to have a docu entry there saving me lifetime

actually, is there a way to leave any kind of comment on the docu? theres so often moments where literally 1 sentence would have saved me half a day of debugging.

[viral ginkgo]

@silent mural button doesn't need rigidbody for raycast hits

[silent mural]

Yeah im switching to that tomorrow. Neither on mouseOver nor onmousedown are reliable

[viral ginkgo]

Just raycast

@silent mural

Have like a player controller that figures out input and feeds it in the ship
And raycasts on to ship panels n stuff

[silent mural]

Yeah still figuring out the structure of that. I have 4 ( maybe more later) ingame screens with buttons that i want full control over

also, can anyone think of a way of placing a whole bunch of moving objects (asteroids in my case) while making sure they will not hit eachother on their path? I thought about spawning in big cylinders the diameter of the asteroid and length of their precession range and have them fall with gravity onto a plane, stacking voer eachother, then freezing them and using that as the paths. like the game mikado.

i know thats a super dirty solution but honestly anything else just is above my intellect, at least what i can think of

[viral ginkgo]

I didn't understand really

[silent mural]

well you know i got these asteroids that move forwards and backwards on a random axis with a sinus function

i want the asteroid fields to move but the asteroids to be on paths that dont crash eachother

[drifting sleet]

@silent mural there are a lot of ways to do that

i think a lot of them are pretty mathematical

and you should probably try designing it by hand, for insight into the problem

it'll become obvious what you need to do after you try to solve the problem in the design tool

[silent mural]

yeah ill try the mikado way first though

if it works then im fine with keeping it.

[viral ginkgo]

@silent mural
well in reallife asteroids shouldnt move like that
they should be in same orbit

isnt it better you did it that way?

[silent mural]

well technically they COULD move like that. same orbit time, different orbit shape causes that

nah, i dont want locked asteroids.

[viral ginkgo]

@silent mural cant be like that
look at saturn

its the stable state

[silent mural]

the moving ones right now are set to 60 seconds period and its not really noticable that they "swing" like a pendulum

and it is cool to fly through them

[viral ginkgo]

@silent mural that will be awkward though

it will be arcadey

not realistic

you want arcady?

[silent mural]

at the moment it feels more realistic than any stationary locked asteroid fileds i have seen in other games though 🙂

and ill increase period time to match with planet orbit time so its as realistic as possible

[viral ginkgo]

asteroids should move in same driection with same speed when you observe a limited space with asteroids

like in saturn

[silent mural]

well but those fields in themselves move too

the thing is, i dont own nasa so i cant calculate 5000 asteroids ramming into eachother building rings

[viral ginkgo]

asteroids shouldnt be ramming into eachother regularly

because the ones that are in weird orbits have already crashed and fell off

[silent mural]

im 100% sure the fields move quite a lot

[viral ginkgo]

reallife asteroid fields should be stable

wdym fields move?

[silent mural]

in themselves

like water moves in a swimming pool

[viral ginkgo]

if you put yourself in saturns orbit, all asteroids will be stationary relative to you
@silent mural

[silent mural]

also reallife asteroid fileds are like 1 roid every 100 kms and super boring

i doubt that.

im 100% sure they move relative to eachother

[viral ginkgo]

@silent mural inner orbits will rotate faster and thats it

@silent mural remember the orbit equations from highschool?
your obit speed is only determined by body distance to gravity source

same height == same speed

[silent mural]

Yeah but most orbits arent perfectly round. They can be oval

And that creation motion to eachother bc different shapes

[viral ginkgo]

when you look at it locally everything will still be stationary relative to you
@silent mural

doesnt matter ellipse or not

you wont see different speeds in your view distance anyway

[silent mural]

Zeichnung

The distance from red to blue is a sinus function.

[viral ginkgo]

@silent mural
thats not the case with old galaxies
old asteroid fields
and saturns ring

[silent mural]

Well at least its a repeating function by time. Not sure if sinus tho but likely

[viral ginkgo]

@silent mural thats not a stable system
and wont remain long

in real life, if that was the case,
you'd expect frequent collisions and eventually
one orbit like saturns ring

[silent mural]

Well space is quite empty so collisions arent that problematic i guess

[viral ginkgo]

@silent mural in long long time and with many many bodies yes there will be collisions

[silent mural]

My sketch is extreme but most orbits are elliptical to some degree, and it does cause these sinus movements when looking from a local perspective

[viral ginkgo]

@silent mural ellipse orbit is fine
two ellipse orbits are not natural

[silent mural]

Idc lol i will have my preceding asteroids

Unless i find a better solution

Eff realistic asteroi belts. They are absolutely boring to fly through if you can fast forward

[viral ginkgo]

i'd pool the asteroids,
spawn from far right with velocity to left

destroy on left

[silent mural]

Hm thats an idea

Takes away ability to do longterm stuff with them though

[viral ginkgo]

what long term

@silent mural you want dofferent asteroid fields in different fields?

[silent mural]

Like set up a mining rig there or a treasure chest idk

I like my method

[viral ginkgo]

@silent mural
so you dont want your asteroids to be templorary objects?

[silent mural]

Not if i can avoid it

and the sinus function allows storing them for despawning and respawn with corrwct rotation and position , as if they had never left

Since it just takes in time as a value

[viral ginkgo]

but sinus is weird on a straght line patrolling back and forward like weirdo

[silent mural]

Trust me, you dont notice it right now with 2 min period and you will never notice it if i raise that time

[viral ginkgo]

add like a cosinus term at least on perpendicular axis and make it into a small orbit, place a small planet

@silent mural but there are asteroids in reverse orbit and thats not physically accurate for the "saturn" reasons

@silent mural also in your solution,
if you mounted the asteroid, sitting and chilling
you'd see the asteroid slowing down a go backwards?!

[silent mural]

Well if you have an orientation point to look at yeah

[viral ginkgo]

i'd think the develepor was on lsd if i saw that in a game lmao

sin cos orbit could do it ithink for your case
and would be just as easy

but you have to have a small planet to orbit

@silent mural
if you want to be phyisally accurate and have it easy, you should just make the asteroids static
rotate the planet in the distance slowly

[silent mural]

My 3d space is static with a rotating pla et for fake orbits. I dont caclulate flight maneuvers into the orbit tho

[viral ginkgo]

and have like a particle system towards planet for fake distant asteroids, have them move forward slightly

also another one for other way
make them move backwards slowly

and you'll have the perfect orbit illusion around a massive planet
@silent mural

@silent mural your ship should be in same orbit as nearby asteroids

and if thats the case, nearby asteroids should be relatively stationary to the ship

However asteroids could have rotational velocity on their own axices

@silent mural
https://m.youtube.com/watch?v=pF5ffUhUbHI

I can't* find an animaition how it would look like if you mounted one of these asteroids in realtime (this one is accelerated you know)

International space station should be completing its orbit around the earth every 90 mins from that altitude
(orbit period around same planet only depends on orbit altitude as i said)

You think you'd notice relative movement in nearby asteroids?
If you had an asteroid field on ISS orbit altitude

@silent mural

Hmm, now that i look closer, i see some relative movement in that video
But this is time scaled you know, every second is like 10 days maybe

[drifting sleet]

great @restive moat i'm glad it worked out

[silent mural]

Thats a really cool animation and a ton of roids lol

And you can complain all you want, i will not get rid of my sinus asteroids :D

[night leaf]

How do you make an active ragdoll for a stickman couldn’t find anything good on yt

[silent mural]

apply gravity + current motion to every bone?

but do yourself a favor and restrict how far the bone joints can move in degrees so you dont get rotating heads n stuff

also i found this talk to be very interesting @night leaf https://www.youtube.com/watch?v=LNidsMesxSE

[untold sand]

any nerds that know how to implement a rigid body without any helper functions up to a private chat?

or a resource that shows you how to do that?

[viral ginkgo]

@silent mural thats not ragdoll
overgroth guy turns into a ragdoll only when hit by something
he's looking for an active ragdoll

@night leaf
you are looking to rotate every limb of the stickman via force

Here is a decent way you can archieve it:

have two ragdolls
one is your current one
second is same as current one but no joints, all limbs are kinematic, has an animator

then you will attach joints
connecting every limb in ragdoll to it's match in animated character

then you will
set your animation on animater character,
disable visuals on animated kinematic character,
and there you go you have your active ragdoll

these are called "active ragdolls" btw
use that term and google will help better

@night leaf

[warm birch]

Hello guys, can somebody help me in Marching cube terrain collision detection?

[silent mural]

im sure bargos can. just give more detail on what you need

[warm birch]

okey so my mesh collider doesnt working in marching cube terrain what else can i try?

@viral ginkgo

[viral ginkgo]

@warm birch use set the mesh in the mesh collider?

meshCollider.sharedMesh

(keep tag me, i might miss)

[warm birch]

Thats the proble the generated terrain not registered as a mesh

Its shows int the mesh filter nothing

i dont know how can i fix this

so do u have an idea how to do it? @viral ginkgo

[viral ginkgo]

how is the terrain generated?

@warm birch

you need to access the mesh to set in meshfilter

[warm birch]

cube marching

[viral ginkgo]

can you have the "Mesh" object

or vertices and tris?

[warm birch]

vertices and tris

[viral ginkgo]

you have access to it?

then you create a mesh object from that and set it in mesh filter

[warm birch]

via script?

now the collisions are working but watch this:

the mesh collider is messed up

@viral ginkgo(sorry for taggin u again)

[viral ginkgo]

@warm birch uncheck "convex"

[warm birch]

oh okay

thank u dudeee, u helped me a lot 🙂

[cosmic axle]

basically I want to cancel out the M2 moment, to prevent the car from rolling to the side. I thought by moving F_x and F_y to the center of mass (CoG) but then I have to manually apply the M_1 moment to the CoG as well.

• But cant I just alternatively not do anything with the wheel forces, leave them as they are and just apply a negative M2 torque and that's it?
• Or instead of doing that, I can just apply an orthogonal force F_y pointing down at the wheels position, right?

https://i.imgur.com/xLb7gix.png

[viral ginkgo]

@cosmic axle i can tell you how apply negative torque

Vector3.project(rb.angularVelocity, vector3.forward)
this is rotation velocity on m2 axis

[cosmic axle]

wait that simple, I dont even have to do math? Cant I just cancel rb.angularVelocity setting it to 0? @viral ginkgo

[viral ginkgo]

can do angularvel *= 0.9f

id do that ^
@cosmic axle

i mean

[silent mural]

if you find a way to correctly calculate needed torque for angularVelocity, please let me know.

[viral ginkgo]

thats the compinent of rotation on m2 axis
if you apply a reverse torque looking at also inertia on that axis, you can stop rotation on m2

[silent mural]

sooo angularVel.magnitude = - needed torque?

[viral ginkgo]

@silent mural this is not for targeting a rotation

not good for your case

[cosmic axle]

wait why 0.9 ?

[silent mural]

still, ill test that idea

[cosmic axle]

since I dont want the roll at all, why not just = 0

angularVel = 0

[viral ginkgo]

@cosmic axle you should probably *=0.9f
when car is airborne

[cosmic axle]

if the car is airbone we dont have Fx and Fy

since no ground

[viral ginkgo]

when its hit by ball it will gain a small ang vel

[cosmic axle]

in the air I need rolls and pitches 🙂

[viral ginkgo]

if you constantly set it to zero, you wont have it

[cosmic axle]

in rocket league you can rotate the car in the air

then like this:
if(isGround)
angularVel = 0

[viral ginkgo]

@cosmic axle add bigger forces while doing
*0.9f
or set rotation directly

or set angular velocity components one by one

doing the vector3.project

splitting ang vel in 3 axices

modifying each by looking at player input

reconstructing a new angvel

setting it back

[cosmic axle]

wait I want the angular Vel in the air

but I dont want the M2 on the ground, so why just not = 0, I dont get it

[viral ginkgo]

then you will set m2 velocity to zero only on ground

leaving other two components untouched

[cosmic axle]

yes, as I said, in the air I need all moments on all 3 axis

I mena I need all angularVel on al 3 axis

[viral ginkgo]

thats your answer
right above

these 3 axices are:
rotation around transform.right, up, forward

[cosmic axle]

yea gotcha

transform.right I probably dont need since I dont have suspention

and the car cant physically flip forward

[viral ginkgo]

@silent mural
lets say you have a body
you want its forward to point at direction myForDir
you want its up to point at direction myUpDir

what you can do is

AddforceAtPos(myForDir, tr.pos + tr.forward);
AddforceAtPos(myUpDir, tr pos + tr.up);

and then you just apply the same forces in reverse direction to keep the body velocity same while you are doing angular force

and a little angVel *= 0.9f
to stop harmonic motion and make it stable

[silent mural]

im glad for your input but lets not confront my rotational rcs again right now, im trying to figure out my interactive screen system 😛

ill come back to it once i free my brain

[viral ginkgo]

@silent mural just save it
that worked pretty good for me

I dont know how you can have a "delta quaternion"
and use that to come up with a torque

i just have no idea
if i knew that would work, i'd choose that way

[lapis plaza]

@cosmic axle there are two things that you can do to prevent the roll without nasty side effects. proper: make antiroll bar to counter this, easy: apply tire forces at com level so body simply doesn't get roll

both of these are way better alternatives than trying to counter the body roll by some counter torque force

last case basically prevents the roll from happening and first case is just physically correct way to do it

do note that I didn't read all past conversation

if you want to prevent the car rolling in air, there are also better ways to handle it than trying to balance the roll after it's detected

stuff like that make the handling feel fake

also, discord is just messed up right now :p

[silent mural]

what should i use to detect a mouseclick while looking at a collider?
i have a raycast called in update, but im not sure how to proceed

is there the chance that the script would miss the click if i use GetmouseButtonDown?

is there the chance that the script would miss the click if i use GetmouseButtonDown?

[cosmic axle]

@lapis plaza thanks for you input! Which forces would I apply directly on the wheels? The f_y directed to the ground?

[drifting sleet]

@silent mural if you want to detect clicks on objects, the eventsystem is very useful for this.

you will resolve basically a bajillion problems that EventSystem already solved

see https://www.raywenderlich.com/4180726-introduction-to-unity-scripting-part-1

for event system basics

it works for 3D objects using a PhysicsRaycaster

[viral ginkgo]

@silent mural
raycast from camare

you can google to find the code

its easy to find

"how to raycast from camera"

[solemn anvil]

Question: I'm making an 2D autorunner. The player has a rigidbody2d and box collider. When I try to test the game the player falls through the ground. The ground object has a box collider as well and I know the player controller isn't the problem because the problem occurs when the player script is inactive

[stuck bay]

Hi! I've to make a rigidbody controller that has a desired max velocity and a time to get to said velocity from standstill. How should I go about calculating that fixed timestep's force/acceleration?

[silent mural]

addForce with ForceMode acceleration can easily be calculated into the time needed to reach that velocity

it raises the current velocity by the amount * time.DeltaTime -> += amount/second. therefor maxVel is reached after (maxVel/accelerationAmount) seconds

(or should be. its 1am dont quote me)

i gotta be honest, i dont remeber if you need the *time.DeltaTime. chances are the method already does that automatically

@stuck bay

[elder agate]

Hey there! anyone here ever user the animators along with a ragdoll in 2D ?

[viral ginkgo]

@elder agate
You are looking at active ragdolls

Here is a decent way you can archieve it:

have two ragdolls
one is your current one
second is same as current one but no joints, all limbs are kinematic, has an animator

then you will attach joints
connecting every limb in ragdoll to it's match in animated character

then you will
set your animation on animated kinematic character,
disable visuals on animated kinematic character,
and there you go you have your active ragdoll

"active ragdolls"
use that term and google will help better

@night leaf and @elder agate
what you two are doing seems to be very similar btw
might wanna contact eachother

[frigid pier]

@viral ginkgo Keep memes from learning channels, please.

[silent mural]

im fighting again with rotation, whats a method where i can put in an angle in degree and the axis to turn around?

and it has to return a quaternion which i can input into transform.rotation

i want to set an absolut angle to the object around its own, local transform.up

[silent mural]

the quaternions have won.

anyways i fixed my flickering screen and cockpit!

i added a second camera which renders the skybox and background and is overlayed by the first person cam.

and a third camera which renders the cockpit and the canvas with instruments. that one doesnt move so the canvas doesnt flicker

im actually thinking about creating a second complete ship with interior which will be locked at 0,0,0 and rendered as the cockpit. player can move around in it smoothly, experiencing no fast movement wobble or worse

[turbid quest]

How can I stop my player phasing through the corners of my boundary? I don't get how the walls work just fine, but if I force my player into the corner for a couple of seconds I go straight through

I've tried just some basic box colliders, and a custom mesh:

Tried changing the width of the boundary as well, nada

They work in conjunction, but forcing the corner causes the player to spazz out

[proud nova]

@turbid quest How is the player moved?

[turbid quest]

Fixed Update, transform.translate

Only x,z movement

[proud nova]

Yea transform.translate doesn't respect physics at all

You are teleporting the player inside the collider

[turbid quest]

Ahh

What would be a good alternative? I need to redo player movement anyways

[proud nova]

CharacterController.Move or Rigidbody.AddForce/velocity

[turbid quest]

Neat, CC.move looks good, I like the OnControllerColliderHit as well

Thank you!

[silent mural]

If i want to detect the exit point of a raycast hitting a collider, google says i have to activate the backside detection im the physic settings. Is there any reason to not do that?

[fossil storm]

Do you guys use wheels colliders at all?

I've been trying to make them work for a few hours now and from 1 solved problem 10 new spring up

I kinda wanna move on and code it by myself now...

[fossil storm]

I made it work, nvmd 😅

[viral ginkgo]

@silent mural
you wont be able to raycast from within objects as easier

if you ship does raycast hitscan shooting for example
from inside the hull
raycasts will hit in front of the hull

.
if you dont want to enable that setting:
you can do a double raycast to find the exit point

you'll get first raycast hits entry point
add more of raycasts direction on top it
cast a second ray from new point backwards

[silent mural]

okay got that to work.
another question: can i detect a collider with a raycast that is below the first collider hit?

my raycast has to first detect the atmosphere of the planet, and then the planet below

[unreal linden]

How can I make 2 colliders ignore each other ?

[hollow echo]

https://docs.unity3d.com/Manual/LayerBasedCollision.html

[unreal linden]

Thanks !

uhmn...

My 2 colliders are still colliding

[hollow echo]

If you're using 2D make sure you use the right Physics settings

[unreal linden]

I am not using 2D

I am in 3D

And I am trying to get Multiplayer stuff done and it mostly works, except that players don't collide because of the way how I destroy a part of my player prefab wich contains the player controls. But the Capsule collider I am using for the players is a sub object of what I kill of, but it needs to be there for my player controller to work on the player of the local player, so I wanted to add a 2nd collider outside of the player controls, but they always intercollide.

Wich pushes the player away.

And nothing seems to work against it.

Solved it !

[autumn jetty]

Possible to get first contact point when triggers collide for DOTS Phyiscs?

[edgy aurora]

i have this issue with my character controller

the character is standing on this ledge and is just outside the grounddetector sphere cast
i want the character to slide down instead of being stuck
how do i do this without increasing the size of the grounddetector sphere cast

https://media.discordapp.net/attachments/493511468134694912/691266101924855858/unknown.png

[proud nova]

Tried changing physics materials to very slippery?

[autumn jetty]

Is this a custom character controller? Or any of unity?

[edgy aurora]

its the unity character controller

using it

https://pastebin.com/JkULGwrW

this is the code

physics materials dont work with character controller @proud nova

[proud nova]

I guess you could slide if raycast down doesn't detect ground within some distance?

Holes in the ground would be risky though

I would just ground in that situation combined with sliding when ground normal is at extreme angle

[edgy aurora]

i want the player to slide down when hes on a ledge like above

in the image

however because the sphere cast doesnt hit the ledge i cant get a ground normal

so how else would i be able to do it

[proud nova]

Resize your sphere cast to match the size of the CC

Make it about the size of the CC and offset it down just a bit

Or use CC's collision events/grounded

[edgy aurora]

im going to try using a rigid body because it handles it already

[hoary rampart]

When working with non-kinematic rigidbodies (especially bouncy ones), there can be situations where collisions with static colliders and other rigidbodies result in large increases in velocity. This can lead to undesirable effects, such as objects suddenly "teleporting" to random parts of the screen or breaking through static collider scenery.

What's the best way to limit the velocity of a non-kinematic rigidbody?

Any idea or tips will.be helpful friends

[neat minnow]

the maxspeed cap doesnt work

nobody has been able to fix it yet

[ornate sapphire]

Hi, I'm having a weird issue with a simple pickup script, when the object gets picked up it doesnt stay in the centre of the screen and for some reason it like, distorts? So sometimes I'll pick up the perfectly symmetrical cube and it will become a slim rectangle like I can't work out for the life of me whats going on its so weird 😂
https://pastebin.com/DbEW2zx4

The script is on my cube btw

[still pecan]

Question: I want to make a collision but it only works for me if I set the rigidbody to be no kinematic. Both gameobjects have rigidbodys. I want it to be kinematic. What I can be doing wrong?

[potent canopy]

I'm using an Overlap Sphere, but it's passing through certain Game Objects that I don't want it to, is there any solution to this problem?

[viral ginkgo]

@neat minnow
the code looks fine
that force will affect velocity after the velocity limiting though

if you are adding like too much force and try to limit velocity right afterwards, that will cause issues

because force is applied to velocity after fixed update has ended

not as soon as you call addforce

[hoary rampart]

@viral ginkgo sorry for mentioning you.....hi I m using impulse force when input is greater than some value I call a function applyimpulse to add impulse force to my ball.
Sometimes the ball goes like "flash" and passes through my static colliders

Like angrybird style when we release touch impulse force is added to my ball based on we aim but sometimes the ball goes like flash any help or tips will.be helpful

[silent mural]

@ornate sapphire do you child the object to the player and is that parent object scaled differently on one axis? Thats what would cause the warp

[viral ginkgo]

@hoary rampart try setting the bird rigidbody continious dynamic collision detection

discrete collisions might skip some collisions

https://answers.unity.com/questions/176953/thin-objects-can-fall-through-the-ground.html

[steel sapphire]

How do I test for a wall collision (so the character can turn around)?
Both the ground under piggy and the ground to the right are in the Terrain layer so both test positive to "Layer == Terrain".
I can think of options:

1. Create a special box collider on all the bits of terrain that might be a wall and tag them "wall" so I test for that tag to decide if it's a side collision or a ground one.
2. Get contact points where the y vector is greater than zero.
3. Create colliders on the right and left of piggy to test for horizontal collisions.

[stuck bay]

There's a whole bunch of information on platforming movement with 6 raycasts here which might help:
http://info.sonicretro.org/SPG:Solid_Tiles

You could also split your horizontal and vertical movement up into two steps, then do a collision test after each one.
Move down, test collision, touching something, this thing must be a floor. Touched a floor!
Move sideways, test collision, touching something, this thing must be a wall. Touched a wall! → Turn around.

You can do that with a basic box collider.

[mighty sluice]

I made this fish body but im a noob regarding how to best use it (blender). I just added articulation joitns to the bone hierarchy (cause articulations rely on hierarchy) and it just works, but I'm wondering if there are any steps I could take that would allow me to use the fish mesh as a collider instead of just a renderer surface, or if there is a sane way to cope with the hierarchy issue if i choose to use configurable joints instead of articulations (which rely on flat hierarchy)

[mighty sluice]

the issue is the masses and the collider

afaik it's one collider per mass

or rather, two masses cannot share the same collider

it seems like to make a truly collider and mass dynamic fish (e.g: getting hit on the tail deforms the tail), I would need deforming meshes at each level of the hierarchy

as it stands, i can just throw primitives on the bones, and that will make the fish collisions work right, but it would be so nice to use the fish mesh itself...

[stark bolt]

so I have 700 static meshes making the scene

10-15 rigidbodies moving

the active constraints keeps spiking

is this normal?

[viral ginkgo]

@mighty sluice just all like small box colliders on each bone

don't try to have a mesh collider, there's no need for that

[mighty sluice]

that's pretty much what i concluded

thanks for the insighttho

[cold light]

Does anyone know how I can stop my players pushing each other around but still be able to collide with each other (like standing on top of each other)?

Also I want object to be pushable still

Help would be awesome! 🙂

[worn lily]

Why is physics in my game causing lag spikes, while the only rigidbody in the scene is the player, and he's not even doing anything

[flat tapir]

@worn lily probably some issues in your code

[worn lily]

Profiler should specify that

It's just saying FixedUpdate.PhysicsFixedUpdate

That's no help

[pallid path]

Well it states BulletManager.OnTriggerEnter() 6.62ms

[stark bolt]

is active constraints spiking inaccurate?

I dont see how 15 rigidbodies can have 85k active constraints

[drifting ore]

What's the correct way to move a gameobject in a script such that a capsulecollider component correctly collides with the world? I was just updating the transform in Update() but that doesn't seem to work...

[drifting sleet]

@worn lily it's still faster than 60FPS so maybe this concern is a little premature

most of your work is not in physics, but in instantiating bullets

[worn lily]

Yeah faster than 60fps

On a 4th day of the project with the scene of the size 25 units by 25 units, made of one capsule, 6 cubes, and 6 low poly models

It's gonna be way less than 60fps when the project is finished then

[drifting sleet]

well it's the bullets

use a pool for the bullets, don't put particles on them

[tender marten]

Hello guys,
If an object has more than one collider and with Queries Start In Colliders unticked, do raycasts it ignore all colliders or only one?

[austere rapids]

Does anybody know how to change the LinierLimit on ConfigurableJoint through sript?

i cant use a float into it

[rare wedge]

can you set the whole thing at once? CJ.linearLimit =

[austere rapids]

i figured it out

you have to input a SoftJointLimit

make a SoftJointLimit var then set a value to it later with SoftJointLimit.limit = 5 or whatever works

just did this
softLimit.limit = Vector3.Distance(transform.position , player.transform.position) - 1;

and it worked

[stuck bay]

So the OnTriggerEnter func doesnt work

[silent mural]

hey, i have a rotation vector and my velocity vector, and i want to figure out how much velocity i have left,right, up down, front back compared to the rotation vector (like when im flying backwards or sideways)

but i cant wrap my head around how to subtract (or whatever) the vectors to get the result i want

help 😄

[silent mural]

i figured it out, measure the angle between heading and velocity vector, use cosinus and hypotenuse (velVector) and you get the magnitude of the heading vector. subtract from velocity, get perpendicular vector.

do the same thing again for transform.right to further divide into left/rgiht + up/down

i wonder if theres a built in function that does that

hm maybe that one does that

https://docs.unity3d.com/ScriptReference/Vector3.ProjectOnPlane.html

[mighty sluice]

Fun fact: the shape of colliders (when too oblong) fucks soo heavily on the inertia tensors of masses that many many things will break

especially joints

solution: manually set the inertiaTensor (and the rotation inertia quaternion if necessary)

inertiaTensor = Vector3.one * 0.1f * mass is usually a good ballpark

[formal trail]

this is the second time this happens to me, the first was when I tried to change players collider when crouching, anybody know how to avoid these uncontrolled spins?

i thought it might be collider of the weapon, because character doesnt do that when you pick up the weapon normally instead of spawning with it, but disabling weapon colliders doesnt help

[autumn jetty]

@formal trail You need to supply with a lot of more info of what you're doing here for anyone to give you some proper answer.

[formal trail]

ive got a rigidbody character controller

and a pickup weapon system

how that system works is that weapon has a rigidbody and a collider that gets turned off when player picks it up and then weapon is child of a weaponHolder transform the player has

everything works fine until I try to spawn player already with a weapon in his weaponHolder transform

then the above happens

as I said this happened also when I tried moving upper collider cylinder down (my character has 2 colliders, 1 for upper part and 1 for lower)

[autumn jetty]

Is this your own character controller?

[formal trail]

yes

also, just noticed that this behaviour continues after I drop my weapon

which is strange because in that situation the character is in exactly the same state as if i never gave him the weapon

[autumn jetty]

So I'm not sure why your character would rotate from it's Y axis like this in any case

Do you follow the ground normal?

In any case, just log what you are colliding with, where this behaviour could arise. And see if it's the weapon or something else.

[formal trail]

In any case, just log what you are colliding with, where this behaviour could arise. And see if it's the weapon or something else.
@autumn jetty yeah, logging collisions is a good idea, ill try that

OnCollisionEnter() does not report collision with the weapon

god damn it this drives me mad

this spinning issue forces me to use retarded workarounds and it shouldnt even occur

[autumn jetty]

So you're not even colliding with ground?

[formal trail]

im colliding with ground

im saying im not colliding with weapon

    {
        foreach (ContactPoint contact in collision.contacts)
        {
            Debug.Log(contact.otherCollider);
        }
    }```

[autumn jetty]

try limit rotation on your rigidbody to only Y axis

[formal trail]

this is what i put into the character controller script thats on the object containing collision mesh for the character

try limit rotation on your rigidbody to only Y axis
@autumn jetty done that like 3 weeks ago

[autumn jetty]

Well, if you're sure you've done this, and have checked it right now, then you know rotation is not from physics

[formal trail]

these are my settings

Well, if you're sure you've done this, and have checked it right now, then you know rotation is not from physics
@autumn jetty nowhere in my script do I even touch the tilt of my character and it tilts anyway lol

[autumn jetty]

Is the weapon parented to the character?

[formal trail]

yes

its parented to the weaponHolder transform.

[autumn jetty]

And from scene view when playing you can't get some hints of what is going on?

[formal trail]

looks like I found a fix

froze characters y rotation

I dont see the problem anymore

altho I have a feeling framerate has dropped

also changed rotation code to this:
player.transform.localRotation = Quaternion.Euler(0, currentLook.x , 0);
to make sure x and z stay 0

[autumn jetty]

@autumn jetty nowhere in my script do I even touch the tilt of my character and it tilts anyway lol
@formal trail Well this was a lie then?

[formal trail]

i said tilt

tilt is around Z axis

I didnt change Z axis rotation at any point

neither X axis rotation

only Y

to look around

[serene locust]

I need some help with player movement, currently you can change velocity in mid-air however i would like it so that the player cannot affect his/her velocity while in air, is there any way i can lock velocity while isGrounded is false?

i managed to make it 0 while isGrounded is false but i want the player to maintain momentum and velocity i just want it so they can't change it

[formal trail]

write 2 separate functions

1 for air movement 1 for ground movement

[serene locust]

sounds good thx

[formal trail]

in FixedUpdate() function check if player isGrounded

[serene locust]

i have another problem now unity wiped everything i did from yesterday

[formal trail]

execute ground movement function only if he is grounded

[serene locust]

k thx

can someone explain in detail the difference between update and fixed update i have the general idea

[formal trail]

FixedUpdate happens in regular time intervals

Update happens every frame

[serene locust]

and isnt linked to fps?

[formal trail]

and isnt linked to fps?
@serene locust yes

[serene locust]

ok

so y should fixedupdate only be used for physics?

[formal trail]

you can change FixedUpdate interval

yeah, its mainly for physics, but not only

in general for stuff that needs to done no matter what framerate you have

[serene locust]

ok thx

are transform units for scale and position in metres?

Should I be getting inputs in update or fixedupdate?

[stark bolt]

so is anyone aware of issues with the physics profiler reporting inaccurate values?

[verbal galleon]

Do anyone know of any library that can do boxcast on arbitrary triangles? I can't use Physics.BoxCast for this one

[drifting sleet]

@stark bolt profiling is tricky in the editor. you have to profile built players

@verbal galleon what's your objective?

@serene locust inputs should come in update.

[serene locust]

ok thanks

[stark bolt]

@drifting sleet Im mostly asking because my active constraints keeps spiking to impossibly high numbers with only 15 active rigidbodies

from 160 to 83000 constraints, and then back down to 160

nothing fancy happening just some addforce, there arent even any joints in the scene

[drifting sleet]

Are you using asset store assets

[verbal galleon]

@drifting sleet I have these 4 raycasts, and I need to ensure that one of them is the highest point in the white box. Thus a boxcast that does not result in one of these 4 raycasts will declare the box as invalid.
The objective is to place a thin object on top of that (a slab of stone for instance), and I need to make sure that the ground will not be sticking out of it. I could make some raycasts and hope for the best, but i'd rather be sure of the result.

Some way of ensuring that the rectangle is 'flat enough'

[stark bolt]

@drifting sleet what kind of assets?

[drifting sleet]

@verbal galleon maybe just use 4 parallel raycasts then

@stark bolt any assets

Especially ones related to physics or effects

[stark bolt]

just some meshes/textures?

its vanilla unity + a city scene

[drifting sleet]

Is it your city scene

Or an asset store city scene

Is there something animated in it?

Anyway, it’s hard to know

[stark bolt]

nothing is animated, its just 15 capsule rigidbodies with x,y,z rotation constraint

[drifting sleet]

The constraints don’t necessarily mean what it would appear they mean

Try creating an empty scene

And putting your capsule rigid bodies in it

And a simple plane geometry and see what happens

[stark bolt]

and if that doesnt help?

[drifting sleet]

Then go to boxes, etc

Then your code is wrong

[stark bolt]

lmao there is no code

[drifting sleet]

At least you know where to look

Alright

[verbal galleon]

@drifting sleet 4 parallel raycasts would miss any 'spike' sticking out in the middle. Adding a cross section would still miss some. I guess 3 parallel up/down + 3 parallel left/right + 2 diagonals would catch most

[drifting sleet]

So you’re saying boxcast isn’t working?

It’s just not clear why boxcast does not work

Try to not use concave geometry generally

[stark bolt]

@drifting sleet I think unity is not reporting accurate numbers

does this convince you Im not crazy?

[drifting sleet]

Okay, so it just doesn’t mean what you think it means

Also you have to profile a built player not the editor

[stark bolt]

so you dont know what it means?

the docs say its related to joints

which there are 0 of

also theres a bug report saying some versions might have a regression

65535 is one of those magic numbers

I suspect its a bug

[austere rapids]

anyone has any idea on why this happens?

https://gyazo.com/ce6b01d43199928c61a43e4ea8ef36c6

it stops moving right at the end of the slope

oh nevermind

i had a configurable joint doing shenanigans in the background

[still pecan]

Hi! Can you help me with a spring joint? I'm trying to configure one like the one in Gravity Hoock so that the bomb attracts the player and it goes up.

[cosmic axle]

probably a stupid question but how do I calculate the forces at any given point? Do I simply
(current_Velocity - last_Velocity) * dt * m?
So F= m*a with a = deltaV * Time.fixedDeltaTime

[silent mural]

yeah, sadly unity doesnt keep track of the forces it applies, which is a shame. so you have to track and write down the positions of an object over time if it dooesnt have a rb, or get the velocity from the rigidbody over time to calculate accelearation etc

deltaVelocity * deltaTime = accelearation
acceleration * mass = force

@cosmic axle

[cosmic axle]

deltaVelocity * deltaTime = accelearation
acceleration * mass = force
@silent mural

Didnt I write the same in my question though? 🙂

[silent mural]

yeah i just wrote it down again from my mind to make sure we have the same thing 🙂

[cosmic axle]

but thanks for reassuring me I am doing it correctly 🙂

ok gotcha

[silent mural]

if you find a better way to track forces, let me know

[cosmic axle]

nah, google said the same 😦

[silent mural]

its a pain to track them. i have a missile script where the missile predicts wher you will be with current vel and aims ahead and i want it to be able to calculate how far you could deviate from that with your max acceleration to make the tracking even more deadly. but logging the acceleartion is painful and not very precise

[cosmic axle]

hmm

[silent mural]

i mean, i dont get why there isnt a forces value for a rb. just like velocity.

[cosmic axle]

yep

it could be PhysX doesnt track it either

[silent mural]

and you do rb.Forces and get the vector which is all forces combined

[cosmic axle]

they just apply forces to change the vel

[silent mural]

no, they apply impuls once at physics update and repeat that to simulate a constant force

[cosmic axle]

I dont see rb.Forces

[silent mural]

yeah, it doesnt exist 😄

but i wish it would

[cosmic axle]

an impulse is p = m*v ?

oh ok they change the vel that way

[silent mural]

yeah, basically its a force added where you dont have a time variable.

[cosmic axle]

i am pretty sure impulse is not a force 🙂

[silent mural]

impuls over time is force

[cosmic axle]

Force is the change of the impulse, so P will be F and V will be a

yea exactly

[silent mural]

the way things are you add impuls to an object by using rb.AddForce.
it applies instantly, with no time value and only once. therefor its an impuls 😄

[cosmic axle]

hmm never thought of that 🙂

[silent mural]

you can simulate a force by adding mulitple impulses over time, but they are still impulses and not forces

and that is a dumb and very confusing practice, especially if you name the method addForce

[cosmic axle]

well if you contantly apply impulses over time you do apply a force

an approximation of the force, p integrated

[silent mural]

if you average the whole thing, you can say "it experienced a force for x secodns" but if you look in detail, its still multiple spikes of impuls

eh anyways enough rant lol

lets just settle on "it would be nice to have a rb.Forces method to get the current forces applying"

[cosmic axle]

yea 🙂

but if you look in detail, its still multiple spikes of impuls
@silent mural

yep, you approximate a continuous F function with an integral over a quantized impulse samples

[silent mural]

the thing is, in reality you cant have instant impuls spikes, because deltaTime can never be zero

[cosmic axle]

I think smth like this

https://i.imgur.com/E6DPVk1.png

[silent mural]

yup exactly

[cosmic axle]

dt, not dp 🙂

or wait

p * dt * dp I think

[silent mural]

y is p and f, x is t

[cosmic axle]

omg I am dumb, yea, y is p LOL

[silent mural]

eh, i havent touched integrals since highschool so i wouldn tknow lol

[cosmic axle]

I always "hated" physics but now I love it since I NEED it

[silent mural]

yeah its a lot cooler if you can actually do sth with it

[cosmic axle]

hated because why the hell do I need to calc. the friction forces for a cube on a slope in real life lol

yea exactly

[silent mural]

i almost failed grade 10 because of vectors and now im playing with them in my freetime

[cosmic axle]

if only I had this back then

https://i.imgur.com/AeunN8b.jpg

[silent mural]

dude thats cool!

[cosmic axle]

its not mine 🙂 Its free from Unity

for educators to teach physics

[silent mural]

thats really cool

[cosmic axle]

wow the forces work, almost 🙂

[silent mural]

can you send me your forces tracker if you re done?

i would like to steal it for my missle

[cosmic axle]

I only did this

_wheelVelocity = _rb.GetPointVelocity(_wheelContactPoint);
        _wheelAcceleration = (_wheelVelocity - _lastWheelVelocity) * Time.fixedTime;
        _lastWheelVelocity = _wheelVelocity;

https://gfycat.com/NastySafeAlaskanmalamute

at the end the forces bug out lol

white lines

[silent mural]

maybe you need a "close to zero -> dont draw lines" condition

[cosmic axle]

y

[silent mural]

i imagine if the forces it registers is super small, it might jump between negative and positive and thats where the flickering comes from

like, -0 and 0

[cosmic axle]

yea it does

registering a 1 m/s^2 acceleration lol

[silent mural]

1 ms2 is not small tho

[cosmic axle]

it is much, the car is stationary, probably the code where I apply brakes

[silent mural]

what are the white lines showing?

[cosmic axle]

the F/m

forward acceleration

[silent mural]

i gotta say, i really love how clean your debugging is. with te forces etc drawn out.

thats something i need to do myself and i just tend to not do it way to often lol

[cosmic axle]

thx, I did it out of necessity 😄

[silent mural]

Can someone help me with collision detection?
my projectile tends to sometimes not register the collisions with the target im shooting.
the proj. collider and rb are set to continous and interpolate, the target is at discrete.
i cant really recreate the times it glitches through tho, not sure whats causing it

[cosmic axle]

try cont. dynamic for the rocket

and cont. for the target

[silent mural]

what does that do?

[cosmic axle]

https://docs.unity3d.com/Manual/class-Rigidbody.html 🙂

it will better detect collisions for very fast objects

[silent mural]

oh nice, it works now

yeah my projectile moves at almost 2 km a second and i believe if the target collider was too small, it glitched through

it does seem to work reliable now.

Does that mean every object in my game has to have continous collision detection?

[cosmic axle]

the docu says then yes

you can try discrete for the target and let me know 🙂

but the docu says you should use cont. target with cont. dyn. rocket

[silent mural]

alright will experiement 🙂

first, i need to get my autocannon working tho

[silent mural]

now look at that, that already looks cool

https://imgur.com/bkBTshR

[silent mural]

now anwards to greater destiny. time to implement point defense cannons to shoot down missiles

[cosmic axle]

looks cool! I love space

if I disable gravity on my car I have basically a space ship with 5 DOF

[silent mural]

Yeah im a huge spacenerd
Thats why im building this project, implement everything as realistic as possiblr and then see and figure out how combat evolves

Probably a lot more chess game than star wars dogfights

[autumn jetty]

@silent mural are you using floating origin?

[viral ginkgo]

@silent mural I think if there would be space fight in real life now, people would use these kinds of weapons
https://www.youtube.com/watch?v=tyUh_xSjvXQ
Seems like the most efficient way to destroy something in space

Not as cool as pew pew pulse lasers though hehe

[silent mural]

Yeah maybe, but i think lasers are easy to counter

Just slap a mirror on the front of your ship, right?

Instead, i would use a bunch of gbu equivalents. You fly towards the enemy, drop a bunch of self guiding bombs and the silently glide towards the enemy. Hopefully long enough undetected to hit

[naive remnant]

silently in space😅

[silent mural]

Eh, i think its not that hard to have spme kind of material that reflects or disperses enough energy to counter lasers

Yeah i mean "undetected" lol
And they only start thrusting when close enough

Lol use the enemys laser to power up your own and shoot back

Thats a fun idea

[naive remnant]

i think easiest way to counter laser attack in spaces to warp space😏

[silent mural]

Thats a bit too futuristic for my case :P

Im using the values of todays tank cannons for my weapons atm

[viral ginkgo]

@silent mural missiles would be shot down with lasers though
unless they are covered with mirrors as well

[naive remnant]

you can power your warp engine through enemy's laser attack😂

[silent mural]

The missiles will ride eternal, shiny and chrome

So to conclude, space warships will be covered on chrome looking like ghetto pimp rides

[viral ginkgo]

and the chrome missiles are the supreme weaponry

[silent mural]

Eventually

[viral ginkgo]

maybe flak for missle defence

[silent mural]

https://imgur.com/roIrllU

[viral ginkgo]

pew pew variance of lazors are weak for big distances

they are slow and not smart

[naive remnant]

slow?🤔

[viral ginkgo]

the laser in gif is too slow for reallife orbit situations

[silent mural]

Thats tracer slugs tho. I added tracer effects for coolness

[viral ginkgo]

if the target alters its orbit randomly at very small pace, offset will be too big

[silent mural]

Its kinetic projectiles

[viral ginkgo]

i'd use that for very close range

smaller than 5 km

[silent mural]

Yup

[viral ginkgo]

but orbit fights shouldn't be close range if it was realistic imo

[silent mural]

Im playing with the tought of implementing a kinetic railgun that shoots slugs at higher speed, like 3km/s but thats still slow for big distances

[viral ginkgo]

cant beat lights speed
or missile smarts

[silent mural]

Yeah

It really depends on how well you can counter lasers

[naive remnant]

little mirror ships which will redirect your laser attacks😏

[viral ginkgo]

a drone with deployable origami mirror umbrella

that could always remain deployed

[naive remnant]

or long steel cables😅

[silent mural]

Like a net?

[naive remnant]

just long 1 cable at huge velocity

[viral ginkgo]

laser weapon would be like silencing the enemy
enemy would need to deploy the reflectives immidently
therefore coving most of the weaponry

[silent mural]

But it also eats tons of energy and gives off you position

[viral ginkgo]

maybe you have to repair some stuff real quick

like in FTL

patch some holes n stuff

actually ftl would be exactly how i imagine realistic orbital fight is
i mean like, no dogfighting and everything, position,velocity doesn't matter
better enginepower/mass, better projectide dodge chances
sort of like a turn based thing
trying to counter eachothers actions
@silent mural

[arctic totem]

also working on orbital dogfighting https://mogacreative.itch.io/orbital-dogfight

[alpine jolt]

Hello everyone, I have a bit of problem. For some reason, checking if a character is grounded is yielding crazy results.

My code:

[arctic totem]

how are you checking?

[alpine jolt]

grounded = Physics.SphereCast(cController.bounds.center, .01f, Vector3.down, out hit, myLayerMask);

I'm using the character controller.

[arctic totem]

.01f is the offset?

[alpine jolt]

It's the radius of the sphere.

[arctic totem]

and what are the funny results?

[alpine jolt]

It doesn't know when it's in the air. It knows it's on the ground, but not when it's in the air.

So it only checks when very close to the ground.

If I lift up the character manually to just a little above the ground, it'll be false and do the fall animation.

But if I take the character high into the air and drop them, it doesn't know it's in the air.

[arctic totem]

but if you know when its on the ground this mean that you know when its not with the same bool

a

[alpine jolt]

Yeah. It should work.

[arctic totem]

okay

[alpine jolt]

But it doesn't.

[arctic totem]

i know

Vector.down is a normalized vector meaning it has a length of 1

try Vector.down * 10f; ...

or take your height off the ground and multiply by 2f to make sure the raycast is always twice the distance off tghe ground?

might work

[alpine jolt]

Let me give that a shot.

[arctic totem]

or just say Vector.down *1000

give it a number you know it will not be higher than

[alpine jolt]

Doesn't work.

[arctic totem]

does it say grounded the whole time?

[alpine jolt]

Now it does.

[silent mural]

@arctic totem thats really cool

[alpine jolt]

It's always grounded.

[silent mural]

and the exact thing im trying to do, only that im not in VR

[arctic totem]

okay that is as expected

@silent mural Thank you. Big workin in progress

love to see your stuff

@alpine jolt okay ready

https://docs.unity3d.com/ScriptReference/Physics.SphereCast.html

has a maxDistance overload you can add to your spherecast

[silent mural]

you can see what i have so far here, its not much but big progress for myself :)
https://imgur.com/a/noHRT9T

more of a logbook than a showcase 😛

[alpine jolt]

@arctic totem What would be the point of changing that?

[arctic totem]

you want to have the spherecast executing always... when the hit returns true; (wich it should always unless upside down or something) you can get the distance of that hit by using Hit.distance

[alpine jolt]

I guess I don't understand something fundamental about this.

[arctic totem]

if Hit.distance is greater then your character height then you are off the ground

its a simple fix

make this : grounded = Physics.SphereCast(cController.bounds.center, .01f, Vector3.down, out hit, myLayerMask);

grounded = Physics.SphereCast(cController.bounds.center, .01f, Vector3.down, out hit, 1000f, myLayerMask);

if there is a HIT your 'out hit' will be filled with hit information https://docs.unity3d.com/ScriptReference/RaycastHit.html

you simply say float distanceToGround = hit.distance; if(distanceToGround > characterHeight) { grounded = false } else {grounded = true};

[alpine jolt]

@silent mural Looks cool.

[arctic totem]

@silent mural okay taking a proper look now

[alpine jolt]

Ah. I see what you mean. Okay. Let me give that a shot. @arctic totem

[arctic totem]

@silent mural the lasers look good at that speed but I will point a few things out as I am now deeply immersed in the same topic(these only generally relate to your game.. i like your laser): OKay lasers are light, which not coincidentally travel at light speed, therefore any representation of lasers as 'bolts of light' moving at speed is utter bullshit BUT since gamers understand 'hot tracer round' style mechanics, making lasers 'look' this way works very for gaming despite it being totally unrealistic. A beam laser that was being pulsed might look closer but still.... in orbital dogfight I have tested shooting lasers (the same kind you are showing) while also moving at orbital speeds in actual orbit.... in this case the lasers look super weird...here is why: the ship is traveling at 2000+ meters per second. When I shoot a laser blast and the bolt is traveling to its target I am moving away from it at orbital speeds and it looks as though the shot came from behind me ,,

[silent mural]

ill stop you right there, they are not lasers but knietic projectils with tracers

[alpine jolt]

Ran some tests. Looks like it basically works, but for some reason the animator is really not doing it's job.

[arctic totem]

O they look sick then!

[silent mural]

i just added the tracers for cool looks 😛 but they are realistically behaving, scaled projectiles 🙂

[arctic totem]

@alpine jolt it may be because in this case the animator work restart with every true state..

[silent mural]

but the problem you are describing still applies, i didnt look into that yet 🙂

[arctic totem]

have you started on floating origin?

[silent mural]

idk what that means lol

[arctic totem]

o boy 🙂

[silent mural]

im just learning unity along the way

[arctic totem]

mee too 🙂

its an esoteric necesity of space games... some of them anyhow.. Kerbal Space PRogram

[silent mural]

ah, its the "move to far and shit start breaking" thing

[arctic totem]

THE KRAKEN

[silent mural]

i have it somewhat solved by locking my cockpit (and interior) in place and having the ship move around. that way you dont notice that stuff is vibrating when you get too far

[arctic totem]

shit starts breaking over 1000 meters if your shit is small.. like buttons in a cockpit

not a bad move

[silent mural]

well it doesnt solve the problem that the worldspace is too small. ill probably have to write a chunk/tile system at some point

i just havent gotten there, ill look into the floating origin thing

[arctic totem]

how about orbital mechanics?

[silent mural]

not a bad move
@arctic totem yeah my UI canvas was wobbling so bad already at start that it was unbearable

[arctic totem]

that is floating point

also UI canvas is terrible for FPS

[silent mural]

the "gameroom" is just a 3d system and you cant escape it, ego fly towards the planet. but i do simulate the orbital mechanic of the comeplete gameroom, ergo the background and planet and sun move according to the "gamerooms" orbit

[arctic totem]

should try to only use those in LObby if possible.. a tip that took me a while to learn

@silent mural no just the two body system of the ship and planet?

shit have toi run be back soon

[silent mural]

nope. the ship does not experience orbital mechanics. the thing is, i want it to be multiplayer focused and form KSp i learned that anything with orbit takes forever to get somewhere

imagine you want to hunt someone on the other side of the planet, youll never reach him bc he can just do evasive thrusting a bit

and imagine having to wait one and a half hour for an orbital maneuver to get somewhere

i do have asteroids that simulate orbital mechanics and i have it planned to be able to adjust the asteroids orbit, but not for player ships.

[autumn jetty]

@silent mural Do you use floating origin?

[silent mural]

not yet

im at a state where you can move around in the ship a bit and shoot at the 5 asteroids that are close. thats all you can do at the moment 🙂 im still creating the very basic systems

i mean, i only learned how to use inheritance yesterday

[alpine jolt]

@arctic totem Thanks a bunch. Looks like I got it working.

[arctic totem]

@silent mural totally right... I have had to scale down the planet to make rendev times less insane.. my planet is currently 200km diameter, the atmosphere is 10k thick and orbital peroid for a small ship just in space is about 1minute 30 seconds.

[naive remnant]

you have to use futurama style engine. It is not your space ship moving through space, it is your ship moving entire universe😂

[arctic totem]

thats the one... what coming from GDC Kerbal Talks?@naive remnant

[naive remnant]

this is basically how you render 3d stuff 😅

[arctic totem]

for everyone who has not watched this and is interested in space game ... its a no brainer https://www.youtube.com/watch?v=mXTxQko-JH0

[halcyon oyster]

wondering a squirrely plane raycast math issue would go here, or general code

[halcyon oyster]

imma put it here since there's a lot of discussion already. I am trying to do a raycast from an arbirtrary point and direction. I am checking against multiple ui elements, each with their own complicated transform hierarchy. Some are scaled, others are not. Some are angled, others are not.

Each element has their own plane, aligned with their arbritrary forward (technically - forward, because they're all backwards). I create a ray from the source point and forward. I call plane.raycast against each plane. This gives me a distance, which I can use with the ray to calculate the hit point for the plane.

I am trying to figure out if the generated point is within the bounds of the objects mathematically. Right now I am using Transform.inversetransform on the hit point, and then checking against bounds I generate using collider.size * transform.localscale.

the issue is that it's just not behaving properly. Up/down works, side to side is not working. (collision movement)

it's like left/right movement is mathematically dampened by something

[halcyon oyster]

and it turns out i was just overthinking it. Once you inversetransform the point, you can check if it's in bounds of 1,1,1. No scaling, no fuss.

[silent mural]

@arctic totem the different spaces mentioned at 12:07 is actually exactly what i came up with too

tiny background planets, a "world space" and a interior space

[arctic totem]

@silent mural I think that is the way to go

[silent mural]

i also like their idea of multiple cameras for far view distances on terrain

[arctic totem]

its was getting tricky in VR for a while... too many cameras with both eyes to build up..

the new integration is making it simpler though

have to keep track of each others work.. you up on Itch or somewahere yet?

[silent mural]

no, the imgur logbook is the only thing i have right now

also my trello which i never update lol

Unbelievable how much time it cost me to get my point defense tracking system working

I wasted at least 2 hours trying to figure out why the hell my firecontrol computer didnt caclulate that targets velocity correctly. And then i noticed that the mock target i had moving had a rigidbidy set to lock position.... and glitched back and forth and messed up tracking

The damn target was broken not my firecontrol

[viral ginkgo]

@silent mural
use PID please

PID is exactly made for these stuff

[arctic totem]

porportional integral derivatives?

[viral ginkgo]

yes

[arctic totem]

I use those to control the orientation of my space craft

[viral ginkgo]

had told him about pids before

[cosmic axle]

Yeah im a huge spacenerd
Thats why im building this project, implement everything as realistic as possiblr and then see and figure out how combat evolves
@silent mural

Me too! If you want I can help with training AI with reinforcement learning do to attacking etc. I have plans for my next project to do combat in space, but I see you already doing what I was planning to do

[viral ginkgo]

@arctic totem
i had used pid for my active ragdoll limbs back then

but it was very bad and wrong

@arctic totem i couldnt pid control rotation correctly

[cosmic axle]

PID control is awesome, everyone should use it 😛

[viral ginkgo]

i was trying to pid control up and forward vectors

how did you do your rotation control?
@arctic totem

[arctic totem]

straight off my spaceship " public void AlignDirToVector(Vector3 dir, Vector3 ToVector)
{
float errorP = Vector3.Angle(dir, ToVector);
Vector3 TorqueAxis = Vector3.Cross(dir, ToVector).normalized;
errorP = AnyAxisMainPID.Update(errorP, Time.deltaTime);
Vector3 correctionTorqueP = errorP * TorqueAxis;
m_Rigidbody.AddTorque(correctionTorqueP);

    Vector3 am = m_Rigidbody.angularVelocity;
    float error = am.magnitude;
    error = AnyAxisErrorPID.Update(error, Time.deltaTime);
    Vector3 correctionTorque = -error * am.normalized;
    m_Rigidbody.AddTorque(correctionTorque);
}"

and

using UnityEngine;

[System.Serializable]
public class NavPID {

public float pFactor, iFactor, dFactor;

float _integral;
float _lastProp; // aka last error or amount we want to move

public NavPID(float pFactor, float iFactor, float dFactor)
{
    this.pFactor = pFactor;
    this.iFactor = iFactor;
    this.dFactor = dFactor;
}

public float Update(float proportional, float timeFrame)
{
    float derive = (proportional - _lastProp) / timeFrame;
    _integral += proportional * timeFrame;
    _lastProp = proportional;

    return proportional * pFactor + _integral * iFactor + derive * dFactor;
}

}

[viral ginkgo]

@arctic totem
thanks
only if i had this back in ragdoll days haha

[arctic totem]

np. I got most of it off Github so.. thanks to those before us

[viral ginkgo]

you the one who makes orbital dogfight?

[arctic totem]

yeah working on it

[viral ginkgo]

pasted in some other channel i think

[arctic totem]

yeah in VR sometimes

[viral ginkgo]

cool

a realistic orbital dogfight huh

[arctic totem]

well its all using newtonian mechanics

Multiplayer will be hard across floating origin space and orbital velocities but im going for it

[viral ginkgo]

cool

should do snapshot interpolation probably

but you can also send velocities for extrapolation

so clients aim at right pos

[arctic totem]

yeah! I have a solver for the six orbital elements... I'll use that to extrap future position

[viral ginkgo]

if orbits are big enough
considering gravity may not be a big deal

can just do linear translation

for future pos prediction

[arctic totem]

you can go from ground to any height orbit that is not hyperbolic... but

kinda of aiming for a Kerbal /Warthunder/ elite dangerous baby

War Thunder has some great multiplayer mechanics

[viral ginkgo]

you wanna do contious math for this stuff?

[arctic totem]

probably be some kind of subsampling

[viral ginkgo]

doing the trajectory equations n stuff

it wont be super accurate

because unity physics intervals move objects linearly

[arctic totem]

Yeah I am cosidering double precision vectors

but for now my orbital prediction is close enough..

[viral ginkgo]

unless you handle it yourself, solving trajectory equations every tick

[arctic totem]

thats what I do

it still swings a bit with the error

as higher speeds the calculation begin to screw

[viral ginkgo]

@arctic totemits not about vector precision
50 linear physics steps per second is not going to be very synced with your trajectory equations

oh you say subsampling

[arctic totem]

ah but the trajectory is calculated in each step

[viral ginkgo]

oh i was about to say that

[arctic totem]

it will have to be a combination of off and on rails

in the end

[viral ginkgo]

since you have the kerbal equations
you will also use those for predict
very nice imdeed

[arctic totem]

yeah I am working on using the Orbital Elements to configure orbits instead of the other way around..

[viral ginkgo]

wait, are these kerbal like equations or

you got the orbit math?

[arctic totem]

Newtonian orbital mechanics with a two body solver to be exact

[viral ginkgo]

that doesnt work for high speeds?

you said?

[arctic totem]

It does but the Kraken lurks

its all a matter of resolution

[viral ginkgo]

Newtonian orbital mechanics with a two body solver to be exact
@arctic totem
does this work iteratively?

[arctic totem]

hardest problem I have taken on... except these baby girls I have to put to bed

[viral ginkgo]

hahah

[arctic totem]

Nope is straight up does everything in one go

got all my math from http://www.braeunig.us/space/orbmech.htm

[viral ginkgo]

whats the matter with resolution tho?

i dont understand that part

[arctic totem]

gotta come back bath trime

time*

[viral ginkgo]

allright

[arctic totem]

🙂

[mighty sluice]

something is going on with my physics updates... I'm using ML-agents, which has a python API that may or may not fuck with some unity settings... Long story short, when i am using the python API for training the physics updates happen faster than deploying the learned brains without python being active

it's quite strange because i cant locate any changes that reflect in project settings

https://cdn.discordapp.com/attachments/502171346503991307/693323699247513621/unknown.png

https://cdn.discordapp.com/attachments/502171346503991307/693323849139355679/unknown.png

this is what the difference looks like...

is there any straightforward way to track down the cause of my sparse physics updates?

when inferencing instead of training, the player loop is wrapped up in additional editor loops

[drifting sleet]

@mighty sluice you can't profile the editor.

you can, but it will give you wrong answers

[mighty sluice]

I realize that some of the results will be skewed, but it still identifies an issue

i am profiling the editor twice, once with the python background in use

so im at least comparing like with like

the real issue is why the physics updates become more sparse despite everything else speeding up

i suspected that the python api might be altering a setting, but i cannot identify what

checked usual suspects like time scale, frame rate, fixed update, etc...

it could be that they are being altered and the editor values aren't updating

been looking into it, and i think this may be related to maximum allowed time step

[silent mural]

@viral ginkgo the floating point issue. The bigger your velocity or position relative to 0.0.0 gets, the more digits behind the comma you loose bc both are 32 bit numbers
Thats the resolution thing. Close to spawn, lots of resolution bc lots of digits behind the comma; very precise velocity and pos.
Get far out, go fast you loose precision and get wobble

Ksp guy mentions they also have "floating velocity" to zero out objects moving fast relative to eachother but i gotta be honest i cant imagine how that would work
If you move 1km/s and your target is still, you can zero out by moving you and the target at 500m/s. But thats literally it. You can half the speed, so double your maximum speed before engine breaks.
Not really a solution

[viral ginkgo]

oh i see

[timber python]

What must I do with rigidbody physics to optimize it....

Unity doesn't allow to do it I guess

I have 100 cubes with rigidbody and the FPS drop to 20

😂 I have a potato rig, but that's a different issue

[viral ginkgo]

@timber python the not moving ones should go sleep any stuff

if you left them stacked somewhere and they are not moving most of the time, it should be fine

[arctic totem]

@silent mural its an OLD known issues. Check this out: https://wiki.unity3d.com/index.php/Floating_Origin

I think their floating Velocity system is just going on rails///

[silent mural]

Is there a reason why unity cant just make float store stuff in 256bit ir sth?

[fringe spruce]

that's not at all how IEEE754 works

[silent mural]

It wouldnt solve the problem forever but make it large enougj to not matter i guess

I dont know how data stuff works, thatd just my first thought :)

[fringe spruce]

it's basically glorified scientific notation

and the hardware works with pretty specific sizes-- mostly 32 and 64 bit

[silent mural]

How about splitting it in 2 parts, and add together. 1457399392 becomes 14573 and 99392 and voila twice the digits available

Is the floating origin script multipöayer compatible?

I know jack shit about how multiplayrr works, just wondering

[timber python]

I am working on a game something like demolition racer, so it will be slightly physics oriented. Just rigidbody will be active on 20 cars! May not be a huge problem @viral ginkgo ..... Still game needs to be optimized 😁

It will be a problem for potato rigs though

[viral ginkgo]

20 cars should be fine on even mobile i think
@timber python

if you dont have some super funky code on cars

[autumn jetty]

You could go 64 bit floating point computation to erase the problem, but it creates longer computation for all math, so its a tradeoff, and somone has to rewrite all that code as well. There's currently a thread on the forums discussing this heavily.

I was thinking of going for floating point origin for my space game as well, but what do you know, traveling in 1000m/s + projecitiles going X that speed doesn't really bring interesting visuals and has gameplay problems. Either you're so far off the enemy is only a dot on your screen, or in a dogfight and one do a turn is almost impossible to keep up with.

[silent mural]

In the ksp video they say they fix the problen by having floating origin and translating the positon etc into doubles for storing

Couldnt physx use doubles from the start? I read that they are more performant too

[autumn jetty]

Where did you read this? https://stackoverflow.com/questions/4584637/double-or-float-which-is-faster

[silent mural]

German website

https://www.softwareok.de/?seite=faq-Windows-CPP-und-C&faq=10

[proud nova]

Seems like it comes down to 1. hardware implementation 2. how fast can that data be loaded to caches

[mighty sluice]

anyone know why unity is skipping a physics update on some ticks?

https://media.discordapp.net/attachments/502171346503991307/693432326784548864/unknown.png

[charred marten]

Physics update only runs 50 times a second

(or whatever number you've set that to)

[mighty sluice]

according to fixed time step

[charred marten]

So if you run 100 frames per second, itll skip every other frame

if you run 25 frames per second, itll run fixedupdate twice per frame

[mighty sluice]

so that makes sense...

but...

im trying to figure out what is causing a difference between 'inferencing' and training with MLAgents

somehow, training results in faster moving physics, but time step appears untouched

during inferencing, there are no (or few) physicsless editor loops

https://cdn.discordapp.com/attachments/502171346503991307/693432806507806750/unknown.png

[charred marten]

that probably just means that inferencing costs more performance

so you're closer to that 50 fps

[mighty sluice]

https://cdn.discordapp.com/attachments/502171346503991307/693301335508123728/unknown.png

280 fps vs 780 fps

training is slow despite having faster physics

slower computationally

[charred marten]

The statistics screen is inaccurate

https://forum.unity.com/threads/is-fps-in-the-editor-accurate.195549/

use a script like http://wiki.unity3d.com/index.php?title=FramesPerSecond&_ga=2.124542402.939345135.1585417642-246594539.1542030173 to accurately display FPS

[silent mural]

Couldnt you just count the time.deltatime of update and calculate fps from that?

[mighty sluice]

According to that script, the FPS is 400 and 200 for inferencing and training respectively

[silent mural]

Or is there a time where update doesbt run every frame?

[mighty sluice]

that's what i just did basically

inferencing should always be faster; during training unity needs to pause and reach out to a python external api that handles learning

[charred marten]

Fair. Just making sure

I'm unsure then, using basic math FixedUpdate should run every 4 frames during training, and every 8 frames during inferencing

[mighty sluice]

indeed, but somehow inferencing decides to skip physics updates on some ticks

[charred marten]

well both of them will

[mighty sluice]

inferencing skips more

[charred marten]

inferencing should be skipping more

because the FPS is higher

[mighty sluice]

but fewer physics ticks result

its not timing them correctly

[charred marten]

yes, fewer physics ticks because there are more ticks in general

with inferencing running 400 frames per second, every 8th frame should roughly be a physics update

[mighty sluice]

but unity is settings its own framerate. so why would it sacrifice fixed updates to increase fps

[charred marten]

(50 physics updates per second, 400/50 = 8)

huh?

its not sacrificing fixed updates

its making it so that theres only 50 physics calls per second

no matter what, it runs the same amount of physics calls per second

with more fps, that just means that you have more frames without physics calls

[mighty sluice]

that all makes sense, but i still cannot explain why my agents are slower

[charred marten]

as in they move slower?

or they compute slower?

physics is framerate agnostic, framerate never impacts physics

what kinda code are you using to interact with the physics system?

[mighty sluice]

just monobehaviors

[charred marten]

if you're doing something like

void Update()
{
  rigidbody.AddForce(movement);
}

then ofcourse itll move slower if you have less frames because update isnt called as often

[mighty sluice]

im not using Update anywhere

everything is done in fixed update

[charred marten]

ok, what are you using?

are you able to show me?

[mighty sluice]

not sure what the question is

im using articulation bodies, skinned mesh renderers, custom physics scripts, etc

[charred marten]

well im mostly just curious how you use the physics engine

because Unity's physics dont change based on FPS, only the scripts you write might

[mighty sluice]

what about maximum allowed time step?

[charred marten]

maximum allowed timestep might enable physics updates to skip if the framerate is too low

but you're running at 200+ fps so thats not the issue

[mighty sluice]

yea that's what i figured

my main suspicion is that the python API is flicking some particular unity setting that i cannot find

[charred marten]

does the python api cause frame dips?

like severe ones?

[mighty sluice]

only when it needs to do a backpropagation

[charred marten]

how often does this happen

and does this happen in FixedUpdate

[mighty sluice]

rarely, and no

[charred marten]

alright

then that shouldnt really be the issue

if your fixedupdate calls are really slow (about 10ms slow) it might skip physics updates

but thats impossible coz ur running 200fps

[mighty sluice]

my thoughts exactly

[charred marten]

Given that all our current conclusions are correct then physics should behave normally for you

And the framerate difference between 400 and 200 shouldnt make a difference

[mighty sluice]

I think i can confirm that fixed updates are happening more slowly during inferencing

fixedDeltaTime still reports the same

[charred marten]

Well yes it should

fixedDeltaTime never changes

[mighty sluice]

so its as if the program is slowed.... but we know that is not true

everything leads back to these editloop updates that should have physics in them

it's the only thing making physics more sparse

is it possible that the very high fps causes it to start drawing a new non-physics frame, but that this takes longer than expected?

therefore the next fixed update gets delayed?

[charred marten]

no then fixedupdate would just catch up

[mighty sluice]

the difference is just so extreme...

this is inferencing and training side by side (above is training, below is inferencing)