Consistent, Convergent, and Constant-Time SLAM

This paper presents a new efficient algorithm for simultaneous localization and mapping (SLAM), using multiple overlapping submaps, each built with respect to a local frame of reference defined by one of the features in the submap. The global position of each submap is estimated using information from other submaps in an efficient, provably consistent manner. For situations where the mobile robot is able to make repeated visits to all regions of the environment, the method achieves convergence to a near-optimal result with O(1) time complexity while maintaining consistent error bounds. Simulation results demonstrate the ability of the technique to converge to errors that are only slightly greater than the full solution, while maintaining consistency.

• Type: Conference
• Year: 2003
• Where: IJCAI
• Authors: John J. Leonard, Paul M. Newman
• Keywords: Consistent Error Bounds, IJCAI 2003, IJCAI 2007, Method Achieves Convergence, Multiple Overlapping Submaps