Improved TSDF-based map merging with Kalman filter and covariance intersection

Abstract

This study proposes an improved TSDF-based map merging framework that incorporates a Kalman filter (KF) and covariance intersection (CI) to enhance mapping accuracy and robustness in collaborative robotic systems. In the single-agent phase, the iterative refinement of TSDF grid maps is achieved using the KF to reduce noise and improve accuracy. For the collaborative phase, a rectified occupancy grid map and sequential spectral map merging approach are introduced to reduce the computation complexity by aligning individual maps. The CI method improved the precision of updates to overlapping grid cells, effectively addressing uncertainties in map merging. Experimental validation across two publicly available datasets demonstrated the proposed method’s superiority over conventional approaches, with significant improvements in SLAM accuracy, computation time, and covariance convergence. These results highlight the robustness and adaptability of the method for diverse mapping scenarios. Future research will extend our enhanced framework to encompass multi-agent exploration systems, expanding its applicability to diverse and complex scenarios.