RobotrekkingDRL2019

Here is the code for the paper: Deep Reinforcement Learning Control of Autonomous Terrestrial Wheeled Robots in a Challenge Task, published on Brahur Brasero 2019 Workshop

We used the following python libraries: tensorflow and gym. Simulation was performed with mujoco

Instaling the Environment

If $GYM is your path to the gym folder

The files inside mujoco folder must be in: $GYM/gym/envs/mujoco

The files inside mujoco/assets folder must be in: $GYM/gym/envs/mujoco/assets

And add the following lines of code in the :$GYM/gym/envs/__init__.py file

register(
    id='Rover3W-v0',
    entry_point='gym.envs.mujoco:RoverRobotrek3WEnv',
    reward_threshold=1000,
    )

register(
    id='Rover4W-v0',
    entry_point='gym.envs.mujoco:RoverRobotrek4WEnv',
    reward_threshold=1000,
    )

Demonstration

A Demonstration can be found here: https://youtu.be/y4M_ypfiLig