D-CLIPSE: Distributed localization achieves near-centralized accuracy

Multi-robot teams rely on accurate, consistent localization for planning and control, but centralized filtering—which optimally fuses all sensor data—is often impractical due to hardware, communication, and computational constraints. Distributed approaches, where each robot runs its own filter and communicates with neighbors, offer scalability but can suffer from inconsistency and degraded accuracy. The new D-CLIPSE framework addresses these gaps by introducing a consensus-based distributed filtering method that shares both preintegrated odometry and relevant shared states among communicating robots. This passive listening on shared state exchange ensures each robot maintains a consistent estimate without requiring a central fusion center.

D-CLIPSE was validated in both simulated and experimental scenarios, showing near-centralized performance in accuracy and especially consistency compared to current state-of-the-art decentralized methods. The paper, submitted to IEEE Robotics and Automation Letters, includes 8 pages of results with 7 figures and 1 table. The method’s communication efficiency and consensus mechanism make it practical for real-world deployments in robot swarms, autonomous warehouses, and search-and-rescue operations where bandwidth is limited and reliable localization is critical.