Unified Framework for Multi-Robot Localization and Target Tracking via Unscented Transform
Abstract: Accurate localization is crucial for autonomous systems operating in diverse environments, from underwater exploration to space missions. These systems rely on precise pose information to perform critical functions and operations like surveillance, search and rescue, distributed coordination etc. Multi-Robot Systems (MRS) excel in such complex applications due to their collaborative capabilities. However, achieving consistent collective state estimation as a team remains a significant challenge for MRS, especially in environments characterized by low connectivity and limited sensory input.
In this study, we propose a unified distributed framework that leverages the Unscented Transform for multi-robot collective self-localization and distributed target tracking. Our approach integrates conservative covariance intersection methods and simple yet effective diffusion strategies. This integration effectively manages inter-robot cross-covariances and achieves reliable global consensus on uncertain state information. This facilitates robust state estimation even in partially observable environments. The framework is designed to be scalable and efficient, without imposing extra memory or computational demands. It enables robots to self-localize accurately using relative measurements and reach consensus on the states of a moving target with minimal communication overhead, making it ideal for mobile robots with constrained computational resources.
Rigorous simulations demonstrate the framework's effectiveness, showcasing consistent and reliable state estimation for all agents and targets involved. This highlights its potential to significantly enhance operational capabilities of MRS in challenging conditions, setting a standard for low-resource autonomous systems.
Advisor: Professor Siavash Farzan
Committee: Professor Berk Calli, Professor Constantinos Chamzas, and Professor Siavash Farzan
Zoom Link: https://us05web.zoom.us/j/87441696089?pwd=XU7WRshWqYbV275iuFSFjaZigPmvhU.1