Weighted Covariance Intersection for Range-based Distributed Cooperative Localization of Multi-Agent Systems

Abstract: Precise localization of multi-agent systems (MAS) in harsh environments is a critical challenge for swarm applications, and cooperative localization is considered a key solution to this issue. Among all solutions, distributed cooperative localization (DCL) has garnered widespread attention due to its robustness and scalability. The main challenge of DCL lies in how to fuse relative measurements between agents under unknown correlations. To address this, covariance intersection (CI) was introduced to DCL. However, the classical CI optimization criteria suffer from issues such as scale imbalance and correlation mismatch during the fusion process. These deficiencies are not as pronounced in 2D scenarios, where the state space is relatively simple and the observability of each state component is well. However, in 3D scenarios, where the state space is more complex and there are significant disparities in the scale and observability of state components, performance degradation becomes severe. This necessitates the design of specialized mechanisms to improve the data fusion process. In this paper, we identify three main drawbacks of the classical CI optimization criteria in recursive filtering and introduce a weighting mechanism, namely weighted covariance intersection (WCI), to improve its performance. We then introduce WCI into range-based distributed cooperative localization in 3D scenarios, developing a concurrent fusion strategy for multiple distance measurements and designing a weighting matrix based on the error propagation rule of the inertial navigation system (INS). Simulation results demonstrate that the proposed WCI significantly enhances cooperative localization performance compared to classical CI, while the distributed approach outperforms the centralized approach in terms of robustness, scalability, and is more suitable for large-scale swarms.