T-Maze Mouse Here is an reinforcement learning model that habituates a mouse-like agent to a T-Maze using only vision input. The agent learned reliably which part of the t-maze to explore to get the reward with no investigation of the alternated t-maze section. Show me the game!
Visual Tile Search Here is an reinforcement learning model that habituates a cartoon cat agent to a multi-colored tile space using only vision input and rotation information. The agent learns to navigate to the closest tile that has the same color as the tile it landed on. Show me the game!
Tool-Use Here is an reinforcement learning model that requires the agent to use a tool placed in the environment to reach a target object. The agent learns with some fidelity to jump on the tool in order to jump to get to the target. Show me the game!
BrAItenberg BrAIn Vehicles I recreated the famous Braitenberg fear and love vehicles where a small reiforcement learning model has to learn to control the wheel motors to create the behavior objective. The model learning with only one layer and 32 units to control the vehicle. We also observed some complex behavior emerge in the love vehicle in the form of cyclical driving patterns. Show me the game!