Event-Based Distributed Linear Quadratic Gaussian for Multi-Robot Coordination with Localization Uncertainty
Tohid Kargar Tasooji, Sakineh Khodadadi
TL;DR
The paper tackles multi-robot rendezvous under localization uncertainty by proposing an event-triggered distributed LQG framework that couples Kalman-filtered estimates with a distributed consensus control law while decoupling the LQG controller from the event scheduler. The approach derives an optimal, Riccati-based consensus gain $L_k$ and a mean-square stability guarantee, accounting for stochastic disturbances and communication constraints. Key contributions include the decoupled design, an explicit Riccati recursion for $ Sigma_k$, and empirical validation on a Robotarium/Robotrium platform showing a tunable trade-off between rendezvous accuracy and transmission rate. This work advances energy-efficient, robust coordination in noisy, bandwidth-limited multi-robot systems with practical relevance to real-world deployments.
Abstract
This paper addresses the problem of event-based distributed Linear Quadratic Gaussian (LQG) control for multirobot coordination under localization uncertainty. An event-triggered LQG rendezvous control strategy is proposed to ensure coordinated motion while reducing communication overhead. The design framework decouples the LQG controller from the event-triggering mechanism, although the scheduler parameters critically influence rendezvous performance. We establish stochastic stability for the closed-loop multi-robot system and demonstrate that a carefully tuned event-triggering scheduler can effectively balance rendezvous accuracy with communication efficiency by limiting the upper bound of the rendezvous error while minimizing the average transmission rate. Experimental results using a group of Robotarium mobile robots validate the proposed approach, confirming its efficacy in achieving robust coordination under uncertainty.