RISC-V-Based Neuromorphic Computing: Spiking Neural Networks for Ultra-Efficient Edge AI

Neuromorphic computing, inspired by the human brain's neural architecture, offers a transformative approach to artificial intelligence (AI) by enabling ultra-efficient, event-driven processing through Spiking Neural Networks (SNNs). Integrating this paradigm with the flexible, open-source RISC-V instruction set architecture (ISA) has catalyzed advances in low-power, real-time edge AI systems. This review explores key theoretical frameworks, such as the Leaky Integrate-and-Fire neuron model and Spike-Dependent Synaptic Plasticity, and the development of neuromorphic ISA extensions that accelerate sparse, asynchronous computations critical for energy-constrained environments. It traces the evolution from early analog-inspired neuromorphic hardware to modern RISC-V-based neuromorphic processors with custom accelerators and scalable Network-on-Chip architectures. The discussion highlights hardware-software co-design efforts addressing analog noise, hardware variability, and scalability, essential for robust, adaptive edge intelligence. Real-world applications in autonomous robotics, healthcare monitoring, and industrial anomaly detection demonstrate the practical impact of these systems. The article concludes that RISC-V-based neuromorphic computing represents a paradigm shift toward sustainable, high-performance edge AI by combining biological inspiration with adaptable processor design, and underscores the importance of ongoing research in device calibration, noise robustness, and ISA standardization to fully realize this technology’s potential.

کلیدواژه ها:

Edge AI ، Neuromorphic ISA extensions ، RISC-V architecture ، Spiking Neural Networks ، Ultra-low power computing ، Network-on-Chip.