Cooperative Object Tracking for Improved World Modeling in Multi-Agent Systems

This paper presents a cooperative object tracking framework that enables robots to share object observations, improving world model accuracy and consistency in environments with occlusions. An open-source, MHT-based object tracking package was applied in conjunction with an inter-robot communication system developed to fuse object observations across multiple robots. Experimental results demonstrate greater than 1.4-fold improvements in tracking performance in cluttered, occluded environments compared to individual tracking.