In this paper, the people tracking system of a mobile shopping assistant based on a SCITOS-G5 platform is explained in detail. The robot has two tasks, to find people requiring assistance and to guide a user to a target without losing contact. For that purpose, a probabilistic model and a Bayesian update scheme have been developed, where data of various sensory systems is merged asynchronously. Experimental results of the tracking behavior during several guided tours in a home store demonstrate the reliability of our approach.
展开▼
