During my time studying at USW, I had the opportunity to study AI for Games through a specialized module. I decided to take on a challenging project where I created a robot arm using Unity MLAgents, and the results were astounding! I achieved a perfect score of 100%, which was a fantastic accomplishment.
The robot arm's "brain" was designed to receive its target location and the current locations of its three-axis of movement. This AI system was developed to receive written instructions similar to how a general-purpose robot operates in the real world.
To train the AI model effectively, I implemented a curriculum learning approach that enabled the robot arm to train one axis of movement at a time while also conducting concurrent training. This allowed for optimal hardware utilization and minimized the impact of any fluke behaviors.
This project marked my first significant undertaking in the field of AI, and it served as a catalyst for me to pursue further research and work with openAI models today.