计算机科学
约束(计算机辅助设计)
钥匙(锁)
鉴定(生物学)
分布式计算
背景(考古学)
理论(学习稳定性)
动作(物理)
工作(物理)
过程(计算)
作者
Zihao Mao,Yunheng Wang,Shenglin Li,Yi Yang,Wenjie Song
标识
DOI:10.1109/lra.2026.3746064
摘要
Multi-robot navigation is challenging in unknown and structurally constrained environments under intermittent communication with limited range and bandwidth. In such settings, robot teams must make long-horizon route decisions from incomplete local observations while resolving dense local conflicts in bottlenecks, dead ends, and other constrained regions. Effective route guidance requires information beyond the sensing horizon, yet sharing dense metric maps is communication-intensive and brittle. The key insight of this work is that long-horizon route selection can be effectively supported by sparse structural knowledge without requiring dense shared metric maps. Based on this insight, a fully decentralized hierarchical navigation framework is developed that combines sparse obstacle-contour sharing for route-level guidance with local metric planning and escape-point-based recovery for real-time conflict resolution. By opportunistically exchanging lightweight structural representations, robots extend navigational awareness beyond local sensing while generating smooth, dynamically feasible local trajectories online. Simulation and real-world experiments demonstrate that the proposed framework enables reliable and efficient multi-robot navigation under low-bandwidth, low-duty-cycle inter-robot communication.
科研通智能强力驱动
Strongly Powered by AbleSci AI