Playing the Pinball game with Spiking Neural Networks

Inspired by biological principles, we propose a neuromorphic model to control the timing of flipper activation in a pinball game. This task is particularly challenging due to the need for high temporal precision and accurate tracking of the ball's motion in the drain region between the two flippers. At the same time, it offers a valuable toy problem for exploring solutions to more complex real-time robotic challenges. Event-based cameras, combined with spiking neural networks (SNNs), provide distinct advantages over traditional artificial neural networks (ANNs) in such contexts. Unlike frame-based sensors, event-based cameras detect asynchronous changes in brightness at microsecond resolution, enabling low-latency, high-speed motion detection. SNNs process this data stream, preserving temporal dynamics and offering high energy efficiency. To detect the direction, speed, and position of the moving ball with millisecond latency, we draw inspiration from the MT/V5 area of the brain, which plays a central role in motion processing.