Reinforcement Learning Swimmer

I used Stable-baselines-3 with a custom environment to train a 2D human how to swim. This project was more challenging than I expected, since getting the RL algorithm to converge required quite a bit of trial and error, playing with different observation and rewards until I got stable converging solutions. A fun project that gave me quite a bit of insight to the physics of swimming and the difficulties using RL on a continuous observation and input space.