SLAM: Teaching Robots to Make Maps

Posted 3 Nov 2005 at 21:23 UTC by steve

A DefenseNews.com article summarizes work at MIT on a variant of the SLAM (Simultaneous Location And Mapping) algorithm called Atlas. John Leonard, Seth Teller, and other MIT researchers have developed a topological approach to SLAM that allows a robot to map large scale environments by combining smaller maps. Applications for the technology ranging from Navy AUVs to flying robots that can map caves in Afganistan are described. Details of the new algorithm can be found in the paper, An Atlas Framework for Scalable Mapping (PDF format). For more on SLAM and robot localization, see John Leonard's numerous publications on the subject. We previously reported on a CMU variant called FastSLAM that was successfully used on an autonomous robot at speeds up to a 90Km/h.