A Radiation-Hardened Neuromorphic Imager with Self-Healing Spiking Pixels and Unified Spiking Neural Network for Space Robotics

摘要

A radiation-hardened neuromorphic imager prototype is developed for space exploration, featuring a fully spike-based neuromorphic vision system architecture, in-pixel self-healing against radiation-induced damage, and integrated unified spiking neural network (USNN) with adaptive neurons and synapses and contrast enhancement at low-contrast conditions. Self-healing reduces dark current by 6.25× at 14kGy cumulative dose, recovering recognition accuracy by 27.8%. USNN consumes 0.0529 pj/SOP at 5,000 events/s.