International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
Neuromorphic Edge Intelligence: Brain-Inspired Computing for UltraLow Latency IoT Systems S. Cynthia Juliet1, B. Devendran2 1Assistant Professor & Head, Department of Computer Applications, Jaya College of Arts and Science, 2PG Student, Department of Computer Applications, Jaya College of Arts and Science, Chennai
---------------------------------------------------------------------***--------------------------------------------------------------------2. LITERATURE REVIEW Abstract – Neuromorphi Edge Intelligence (NEI) represents a convergence of brain-inspired computing and edge architectures, enabling event-driven, ultra-low-power processing for Internet of Things (IoT) applications. This paper introduces a novel framework that integrates Spiking Neural Networks (SNNs) with edge computing infrastructure to achieve sub-millisecond inference latency while consuming 95% less energy than traditional deep learning approaches. We propose a hierarchical neuromorphic architecture with adaptive spike-timing learning mechanisms for real-time pattern recognition in resource-constrained environments. The framework implements bio-inspired plasticity rules combined with hardware-aware optimization for deployment on neuromorphic chips like Intel Loihi and IBM True North. Experimental validation demonstrates superior performance in autonomous robotics, predictive maintenance, and smart sensor networks, achieving 98.7% accuracy with 40μW average power consumption. This research establishes NEI as a transformative paradigm for next-generation intelligent edge systems.
Neuromorphic computing traces its origins to Carver Mead's pioneering work in the 1980s on analog VLSI implementations of neural computation. Recent advances in neuromorphic hardware, particularly Intel's Loihi chip (2017) and IBM's True North processor (2014), have demonstrated the feasibility of largescale spiking neural network deployment. Mass (1997) established theoretical foundations for Spiking Neural Networks, proving their computational superiority over traditional rate-coded networks. The intersection of neuromorphic computing and edge intelligence remains largely unexplored. Davies et al. (2018) demonstrated Loihi's capabilities for real-time learning but focused on centralized deployments. Our work bridges this gap by developing architectures and algorithms specifically designed for neuromorphic edge deployment, incorporating bio-inspired plasticity mechanisms with distributed learning protocols.
3. NEUROMORPHIC ARCHITECTURE
Key Words: Neuromorphic Computing, Edge Intelligence, Spiking Neural Networks, Event Driven Processing, UltraLow-Power AI, Brain-Inspired Computing
The exponential growth of IoT devices has created unprecedented demand for intelligent edge processing capable of real-time decision-making with minimal energy consumption. Traditional Artificial Neural Networks (ANNs), despite their effectiveness, suffer from high computational overhead and power requirements that limit deployment in battery-operated edge devices. Neuromorphic computing, inspired by biological neural systems, offers a radical alternative through event-driven, asynchronous processing that mimics the energy efficiency of the human brain.
Figure 1 illustrates the complete NEI architecture, showing the flow of spike events from neuromorphic sensors through hierarchical processing layers with asynchronous, eventdriven communication that minimizes energy consumption by processing information only when meaningful changes occur.
This paper presents a comprehensive framework for Neuromorphic Edge Intelligence (NEI) that integrates Spiking Neural Networks with hierarchical edge architectures, implements adaptive learning algorithms compatible with neuromorphic hardware, and demonstrates practical deployment strategies for resource-constrained environments. Our contributions establish theoretical foundations and practical methodologies for building the next generation of intelligent, energy-efficient edge systems.
|
Impact Factor value: 8.315
INTELLIGENCE
The NEI architecture comprises three hierarchical layers: sensor nodes with neuromorphic preprocessing, edge gateways with SNN inference engines, and optional cloud connectivity for model evolution. This design exploits the event-driven nature of both sensory data and neuromorphic computation, eliminating unnecessary processing cycles and achieving orders-of-magnitude improvement in energy efficiency. At the sensor layer, neuromorphic vision sensors (DVS cameras) and spike-encoding circuits convert continuous signals into discrete temporal events, while edge gateways implement leaky integrate and fire (LIF) neuron models with STDP learning capabilities.
1.INTRODUCTION
© 2025, IRJET
EDGE
|
ISO 9001:2008 Certified Journal
|
Page 457