In order to build a RL agent for climbing task on the robot, we need to define a state space, a action space and reward function, such that RL agent is able to know to control the robot to do which action under a certain state via trainning. Besides, states defined in the space space need to be observable via sensors, in our case, we only have a camera on a smartphone looking at the scenario from above. Thus, states need to be some visual information that can be captured by camera. Interface, thourgh with the RL agent is able to send the control message to robot's motors, is implemented.
WenzelHu/rl-robot-climbing
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
About
Reinforcement learning program for metabot robot to learn to climb over an obstacle.
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published