Self-Parking Car
About
For my final year project, I wanted to train a virtual car using reinforcement learning (PPO) with Unity's ML-Agents toolkit. The aim was to have the agent park in one of the given spaces with a 85-95% success rate. It performs better in sparse environments, and in some cases does meet the aim.
The video above demonstrates the performance of the agent in the same environment it was trained in. For a better look at how it performs in other environments, please press the 'Play Game' button.