- Multi-agent reinforcement learning approaches for collaborative locomotion
- Implementations of centralized value-based approaches, distributed actor critic methods and hierarchical decomposition algorithms
- Geared towards increasing scalability for multi-agent learning and applications for locomotion through difficult terrain
- Path planning with revised DStar Lite algorithm for collaborative robot navigation
- Perception using RGB values and greedy algorithms
- ROS interface to design of ROS infrastructure to facilitate multirobot updates
- Box_ws: Collaborative Box-Pushing for Bridging
- Bridge_ws: Tether-based Approach for Bridging
- Double DQN with dual networks and TD Error Priority Sampling
- Q-SARSA
- TD3 (Twin Delayed DDPG)
- Soft Actor Critic ~ coming soon
- MADDPG ~ coming soon
- Hierarchical Feudal-Inspired Algorith ~ coming soon
- Unfinished
- Catkin workspace using ROS interface for collaborative robots
- Uses multiple robots in V-Rep environment to navigate unpredicatable terrain using revised D-StarLite algorithm
- Utilizes proxy ROS node to effectively communicate new obstacles between robots
- Naive, greedy implementation for robot following red light