Lane-Following-Agent Info Vehicle agents trained within simulation for vision-based lane centering. Using reinforcement learning. Using Unity ML Agents platform. Uses a 50x50 px greyscale image of rendered lanes, forward speed, lateral speed, and yaw rate, as input to the neural network. Neural network output is steering and accel / braking actions. Controls Arrow keys when manual torque, and steer override checkboxes are toggled.
