机器人群体的完全分布式且具备弹性的源搜索算法
Fully distributed and resilient source seeking for robot swarms
AI总结：
针对现有机器人群体源搜索算法灵活性与故障恢复能力不足的问题，提出一种完全分布式的源搜索方案，通过局部测量与分布式估计实现无需中央协调的运动控制，经大规模仿真验证有效。
摘要：
现有面向机器人群体的源搜索算法通常需要直接的梯度测量或刚性几何编队，这限制了算法的灵活性和对机器人故障的恢复能力。本文提出一种完全分布式解决方案，通过局部场测量与质心相对坐标的分布式估计来计算上升方向，克服了上述限制。该方案的架构包含三种指数收敛算法，运行于快慢闭环系统中，无需中央协调即可同时实现估计与运动控制。本文的框架可适配任意群体几何构型，并分析了机器人的空间分布对梯度可观测性、鲁棒性及故障恢复能力的影响；还确定了能保证与真实梯度对齐的最优群体形状，并展示了形状变形如何操控群体运动。该方法针对$\boldsymbol{\text{R}}^m$中的运动学点开发，并扩展到具有恒定速度的2D单轮机器人。大规模群体的仿真验证了该方法的有效性。
英文摘要：
Existing source-seeking algorithms for robot swarms typically require either direct gradient measurements or rigid geometric formations, limiting their flexibility and resilience to robot failures. We propose a fully distributed solution that overcomes these limitations by computing an ascending direction through local field measurements and distributed estimation of centroid-relative coordinates. The resulting architecture consists of three exponentially convergent algorithms operating in a slow-fast closed-loop system, enabling simultaneous estimation and motion control without central coordination. Our framework accommodates arbitrary swarm geometries and analyzes how the spatial distribution of robots affects gradient observability, robustness, and resilience to failures. We characterize optimal swarm shapes that guarantee alignment with the true gradient and show how shape morphing can maneuver the collective motion. The approach is developed for kinematic points in $\mathbb{R}^m$ and extended to 2D unicycles with constant speed. Simulations with large-scale swarms validate the methodology.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
