Scholarlink

<p>The increasing need for decentralized, cognitive behavior in cooperative mobile robots has spurred interest in low-power, swarm-inspired reinforcement designs optimized for edge platforms. This work introduces the Neuromorphic-Enhanced Swarm Agent Reinforcement Architecture (NESARA), a biologically inspired architecture that combines Glowworm Swarm Optimization (GSO) with Multi-Agent Deep Q-Learning (MADQL) and neuromorphic computation. NESARA allows distributed mobile robots to cooperatively explore, learn, and decide in real-time with negligible computational overhead. Neuromorphic units emulate spiking neural responses for power-efficient edge processing as light-weight cognitive nuclei. GSO enables clustering and spatial self-organization of dynamic agents, and MADQL enables reinforcement-based policy learning to address uncertainty, task coordination, and support collaborative decision-making. By integrating cognition into every agent via neuromorphic processing and facilitating swarm-level adaptation through GSO and MADQL, NESARA provides a context-aware and self-regulating decision layer. Performance analysis over benchmarked maze worlds and dynamic fields of obstacles proves NESARA to outperform individual GSO or MADQL systems. The architecture realizes 22.7% quicker convergence, 31.4% enhanced task accomplishment rates, and 26.8% reduced energy consumption relative to baselines. The findings confirm the validity of NESARA’s utility in creating robust, real-time, and smart multi-robotic ecosystems under edge computing limitations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Neuromorphic Enhanced Swarm Agent Reinforcement Architecture (NESARA) for collaborative mobile robotics in edge environment

  • 期刊论文

  • 发表日期：

    2025年12月4日

查看全文
  • K. Punitha,
  • R. Jayanthi,
  • C. A. Rishikeshan

摘要

The increasing need for decentralized, cognitive behavior in cooperative mobile robots has spurred interest in low-power, swarm-inspired reinforcement designs optimized for edge platforms. This work introduces the Neuromorphic-Enhanced Swarm Agent Reinforcement Architecture (NESARA), a biologically inspired architecture that combines Glowworm Swarm Optimization (GSO) with Multi-Agent Deep Q-Learning (MADQL) and neuromorphic computation. NESARA allows distributed mobile robots to cooperatively explore, learn, and decide in real-time with negligible computational overhead. Neuromorphic units emulate spiking neural responses for power-efficient edge processing as light-weight cognitive nuclei. GSO enables clustering and spatial self-organization of dynamic agents, and MADQL enables reinforcement-based policy learning to address uncertainty, task coordination, and support collaborative decision-making. By integrating cognition into every agent via neuromorphic processing and facilitating swarm-level adaptation through GSO and MADQL, NESARA provides a context-aware and self-regulating decision layer. Performance analysis over benchmarked maze worlds and dynamic fields of obstacles proves NESARA to outperform individual GSO or MADQL systems. The architecture realizes 22.7% quicker convergence, 31.4% enhanced task accomplishment rates, and 26.8% reduced energy consumption relative to baselines. The findings confirm the validity of NESARA’s utility in creating robust, real-time, and smart multi-robotic ecosystems under edge computing limitations.