Qiu Guangping, Deng Jizhong, Li Jincan, Wang Weixing
School of Artificial Intelligence, Zhujiang College of South China Agricultural University, Guangzhou 510900, China.
Biomimetics (Basel). 2025 May 26;10(6):347. doi: 10.3390/biomimetics10060347.
To address the core challenges in multi-robot path planning (MRPP) within large-scale, complex environments-namely path conflicts, suboptimal task allocation, and computational inefficiency-this paper introduces a Hybrid Clustering-Enhanced Brain Storm Optimization (HC-BSO) algorithm designed to improve both path quality and computational efficiency significantly. For optimizing initial task assignment, the conventional K-Means clustering method is supplanted by a hybrid clustering methodology that integrates Mini-Batch K-Means with Density-Based Spatial Clustering of Applications with Noise (DBSCAN), facilitating an efficient and robust partitioning of task points. Concurrently, we incorporate a two-stage exploration-perturbation evolutionary strategy. This strategy effectively balances global exploration with local exploitation, thereby enhancing solution diversity and search depth. Comparative analyses against the standard Brain Storm Optimization (BSO) and other prominent swarm intelligence algorithms reveal that HC-BSO exhibits significant advantages in terms of total path length, computational time, and path conflict avoidance. Notably, in large-scale, multi-task scenarios, HC-BSO consistently generates high-quality, conflict-free paths, demonstrating superior stability, convergence, and scalability.
为解决大规模复杂环境中多机器人路径规划（MRPP）的核心挑战，即路径冲突、任务分配次优和计算效率低下问题，本文介绍了一种混合聚类增强的头脑风暴优化（HC-BSO）算法，旨在显著提高路径质量和计算效率。为优化初始任务分配，传统的K均值聚类方法被一种混合聚类方法所取代，该方法将Mini-Batch K均值与基于密度的带噪声应用空间聚类（DBSCAN）相结合，有助于对任务点进行高效且稳健的划分。同时，我们引入了一种两阶段探索-扰动进化策略。该策略有效地平衡了全局探索与局部开发，从而增强了解的多样性和搜索深度。与标准头脑风暴优化（BSO）和其他著名的群体智能算法的对比分析表明，HC-BSO在总路径长度、计算时间和路径冲突避免方面具有显著优势。值得注意的是，在大规模多任务场景中，HC-BSO始终能生成高质量、无冲突的路径，展现出卓越的稳定性、收敛性和可扩展性。
