基于大脑启发的边缘智能人工智能：系统性综述

Spiking Neural Networks (SNNs) hold the potential to overcome the severe Size, Weight, and Power (SWaP) constraints of edge intelligence. However, the field currently encounters a 'Deployment Paradox,' where theoretical energy gains are often negated by the inefficiencies of mapping asynchronous, event-driven dynamics onto traditional von Neumann architectures. This survey adopts a rigorous system-level hardware-software co-design perspective to examine the trajectory from 2020 to 2025, specifically targeting 'last mile' technologies—from quantization methodologies to hybrid architectures—that translate biological plausibility into silicon reality. The interplay between training and inference processes is critically dissected, highlighting the key technologies necessary for achieving efficient edge intelligence in practical applications.