Toward Efficient Collaboration in Autonomous Mobile Robot Fleets: Addressing Latency and Distributed MPC

Abstract

Abstract Cloud-based multi-robot systems offer significant advantages in computation capacity and coordination capabilities, but face critical challenges related to communication latency that can compromise system performance and safety. This paper presents a latency-tolerant hierarchical control architecture for Autonomous Mobile Robot (AMR) fleet coordination that maintains operational integrity despite network delays. The proposed multi-layer framework integrates (1) a cloud component implementing switchable distributed model predictive control (DMPC) and path sampling for trajectory generation, (2) an AMR component featuring switchable controllers (Pure Pursuit and hybrid LQR) for local trajectory tracking, and (3) a comprehensive simulation environment for systematic evaluation. We introduce a custom ROS2-based communication infrastructure with intelligent finite state machine design to manage complex system behavior and enhance resilience to varying network communication latency conditions. Experimental results across multiple test scenarios demonstrate that our approach maintains over 90% success rates even under high-latency conditions (1000ms), with limited path deviation and execution time. The findings provide quantitative insights into the latency impact of MPC solvers based on different scenarios and offer practical solutions for robust cloudrobotic systems design and implementation.