Decoupled Multi-Robot Localization and Target Tracking via Dynamic Covariance Scaling and Inverse Covariance Fusion

Multi-robot systems excel in complex tasks through collaboration, but reliable collective state estimation remains challenging, especially under limited connectivity and sensory input. This paper presents a unified distributed framework for multi-robot collective localization and distributed target tracking using the Unscented Transform and Covariance Intersection. The proposed approach efficiently handles nonlinear state estimation in decentralized multi-robot systems, effectively balancing computational demands and estimation accuracy. By employing CI for robust fusion of local state estimates without explicit cross-correlation management, our method significantly reduces communication overhead and computational complexity. Inverse covariance weighting optimizes the integration of measurements to achieve more reliable estimates and improve accuracy, while a dynamic covariance scaling strategy is proposed to further enhance robustness under uncertainty. Extensive simulations demonstrate the framework’s high localization accuracy and robustness under varying noise conditions.