Skip to content

This is a Unity project that trains reinforcement learning agents that have "vision" used computer vision models instead of raycasts to observe their environements.

License

Notifications You must be signed in to change notification settings

PatP15/cv-walking-rlagents

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CV and RL Walking Agents

Active ragdoll training with Unity ML-Agents (PyTorch).

Ragdoll Agent

Based on walker example Built off of walker github The Robot Kyle model from the Unity assets store is used for the ragdoll.

Features:

  • Heuristic function inlcuded to drive the joints by user input (for development testing only).

  • Added stabilizer to hips and spine. The stabilizer applies torque to help ragdoll balance.

  • Added "earlyTraining" bool for initial balance/walking toward target.

  • Added WallsAgent prefab for navigating around obstacles (using Ray Perception Sensor 3D).

  • Added StairsAgent prefab for navigating small and large steps.

  • Added curiosity to yaml to improve walls and stairs training.

  • Added two environement one for obstacle detection and another for terrain detection

  • Integrate YOLOv8 with Barracuda and pipe the outputs into the observation space of the RL agent

About

This is a Unity project that trains reinforcement learning agents that have "vision" used computer vision models instead of raycasts to observe their environements.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •