Neuromorphic Sensing, Perception, and Control for Robotics

Neuromorphic systems bring a paradigm shift in the way computation is performed in robotics. Traditional computing and deep learning supported the recent progress of robotics, but lack robustness and adaptation to bring robots operating in ever-changing scenarios and tasks. Neuromorphic intelligence develops brain-inspired supporting technology featuring brain-like adaptation, robustness and efficiency. These systems are based on event-driven (change detection) sensing, spike-based cooperative and competitive computation for perception and for control. This chapter summarizes state-of-the-art neuromorphic technologies for event-driven sensing and briefly explains how they operate differently to standard sensors. It then describes different frameworks for developing perception modules and neuromorphic computing platforms supporting some of those. The perception modules can cooperate in a closed loop with the actuation system to directly control motors with the output spikes of the networks. The high potential for solving current robotic problems leaves a strong outlook for the field of neuromorphic robots in the coming years.