Multi-robot Multiple Hypothesis Tracking for pedestrian tracking with detection uncertainty

Abstract

The problem of tracking walking people with a team of moving robots is tackled in this paper. We extend the Multiple Hypothesis Tracking method so as to handle measurements coming from multiple sensors and to allow for one-to-many associations between targets and measurements. Derivation of hypotheses probabilities accounts for the overlapping fields of view of the robots sensors and for uncertainty in detection. In the context of two experiments involving people walking among moving robots, the successful integration of our tracking algorithm to a real-world scenario is assessed.