We need a new crate that can do all of the RL and Simulation aspects of this project. It needs a method that can take in a Vec of point information and other metadata(Likely Vec<[f32; N]> where N is how many inputs per point you're taking in) along with a Body (either the raw URDF file or a serialized version of our struct).
It can then reconstruct the environment and perform simulations on the agent by moving it's joints and learning where to maximize the goal. The simulation should happen in bevy with rapier-bevy as the physics engine, and should be able to be parallelized.
We need a new crate that can do all of the RL and Simulation aspects of this project. It needs a method that can take in a
Vec
of point information and other metadata(LikelyVec<[f32; N]
> where N is how many inputs per point you're taking in) along with aBody
(either the raw URDF file or a serialized version of our struct).It can then reconstruct the environment and perform simulations on the agent by moving it's joints and learning where to maximize the goal. The simulation should happen in bevy with rapier-bevy as the physics engine, and should be able to be parallelized.
Notes: Bevy Rapier