Examples

The following examples show how to create an environment, how to evaluate a baseline policy (written by hand), how to train and evaluate a RL policy, and how to generate a video. We repeat these four steps for three scenario, where we

• control a single robot

• control two robots of the same type with the same policy

• control two different robots.

Each scenario is implemented as a little package enki_env.examples.<scenario> in examples, subdividing the common steps in four modules, which can be run separately like

python -m enki_env.examples.<scenario>.<part>

where <part> is one of environment, baseline, rl or video and <scenario> is one of single_robot, same_robots, or different_robots.

• Single Robot
  • Environment
  • Baseline
  • Reinforcement learning
  • Video
• Two robots of the same type
  • Environment
  • Baseline
  • Reinforcement Learning
  • Video
• Two different robots
  • Environment
  • Baseline
  • Reinforcement Learning
  • Learning a centralized policy
  • Video