Going beyond PyBullet into autonomous driving
PyBullet is a great environment to test the capabilities of reinforcement learning algorithms in a high-fidelity physics simulation. Some of the other libraries you will come across at the intersection of robotics and reinforcement learning are:
- Gazebo: http://gazebosim.org/,
- MuJoCo (requires license): http://www.mujoco.org/,
- Adroit: https://github.com/vikashplus/Adroit.
In addition, you will also see Unity and Unreal Engine-based environments used in training reinforcement learning agents.
The next and more popular level of autonomy is of course autonomous vehicles. RL is increasingly experimented in realistic autonomous vehicle simulations as well. The most popular libraries in this area are:
- CARLA: https://github.com/carla-simulator/carla,
- AirSim: https://github.com/microsoft/AirSim. (Disclaimer: The author is a Microsoft employee at the time of authoring this book and part of the organization developing...