On-line target-motion prediction for autonomous multirobot search in realistic terrains with time-expanding boundaries: A novel probabilistic approach

This paper presents a novel methodology that uses probabilistic target information, in the form of space- and time-varying target-location probability density functions, to guide multiple robots to conduct a search task autonomously. The object of the search is a moving target for which information exists to build and maintain a probability distribution for its position at any given time within a continually expanding search area. This position distribution is created, updated, and propagated, based on varying terrain and clues, and also accounts for specific target psychology. The method is demonstrated through a computer simulated example search scenario. This method has applications for search and rescue scenarios in realistic, wilderness terrains.