The goal of this project is to utilize reinforcement learning (RL) to balance a ball at a specified location on a beam.
The beam was inspired by Sydney Harbour Bridge:
Due to the sparse reward signal common in RL a large amount of data is needed to converge on a policy. Naturally a simulation is made to expedite the process of collecting expereience. This can be done through traditional dynamics modeling:
The env.py file builds such a simulation to use as a virtual enviroment.