mossy spindle
Hi! I have recently started working on a project where I want to train some AI to play soccer (I am using MLAgents at the moment) and have hit a couple roadblocks:
Problem 1) Currently, I have 2 seperate brains in two seperate scenarios: One to make a ragdoll stand up, and one to make it walk. I want to combine these two so that it can walk to the target AND stand up when it falls to the ground. What would be an appropriate approach? Would it work if I wanted to continue adding actions eg. kicking/passing?
Problem 2) I want to have 2 'levels'. Lower level brain controls the agents immediate action eg. how to walk, how to kick. The 'Higher' level brain will control where the player should move on the field to intercept the ball, how to position, etc. I have seen examples of this done, but have no clue on implementation?
Any other comments are also super appreciated
Thank you!!