基于鱼群涌现行为启发的集群机器人有限交互深度模型  

A Finite Interaction Depth Model for Swarm Robotics Inspired by Fish Schooling Behavior

作　　者：蔡佳浩 刘磊[1,2] Jiahao Cai;Lei Liu(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai;School of Management,University of Shanghai for Science and Technology,Shanghai)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海 [2]上海理工大学管理学院,上海

出　　处：《建模与仿真》2025年第2期460-474,共15页Modeling and Simulation

摘　　要：在人工智能和机器学习技术的推动下,集群机器人系统作为一种先进的智能系统在多个应用领域表现出了显著的潜力。特别是在环境探测、搜索救援及灾害响应等领域,集群机器人通过其高度的协同性和灵活性,能够有效提高操作效率和安全性。然而,实现有效的集群机器人控制在动态和不确定环境中仍面临诸多挑战,如计算资源的大量需求和系统的鲁棒性问题。本研究提出了一种新型的基于Transformer的有限交互集群机器人控制模型,灵感来源于自然界鱼群的涌现行为。我们采用深度强化学习方法,结合生物群体动态数据,对模型进行训练和优化,进而在复杂环境中实施精确的路径规划和动态避障。通过大量实验验证,结果表明该模型能够显著提升机器人群体的协同操作性能和环境适应性。With the advancement of artificial intelligence and machine learning technologies,swarm robotic systems have demonstrated significant potential as advanced intelligent systems in various application domains.Particularly in areas such as environmental exploration,search and rescue,and disaster response,swarm robots,with their high degree of coordination and flexibility,can effectively enhance operational efficiency and safety.However,achieving effective swarm robot control in dynamic and uncertain environments still faces many challenges,such as high computational demands and system robustness issues.This study proposes a novel Transformer-based hard attention swarm robot control model,inspired by the emergent behavior of natural fish schools.We employ deep reinforcement learning methods,combined with dynamic biological swarm data,to train and optimize the model,enabling precise path planning and dynamic obstacle avoidance in complex environments.Extensive experimental validation demonstrates that the model significantly improves the swarm’s cooperative performance and adaptability to environmental changes.

关 键 词：集群机器人 智能系统 深度强化学习 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]