Robot motion training using reinforcement learning

In 2020 I took an interest in reinforcement learning and ML/AI and began teaching myself some of the tools in the field. I wanted to build a magnetic swarm robot, and the actuation patterns for movement were so complicated that they could only be learned, not preprogrammed. Ultimately I stopped when it became clear that the compute for the magnetic field simulations was out of my budget, but not before I had experimented with training robots to move using OpenAI baselines, ROS, PyChrono, Simulink, and Simscape Multibody.