Stigmergic search for a lost target in wilderness

The problem of searching for a missing person in a wilderness search and rescue application is often modelled as a straightforward application of Bayes' Rule to a conventional occupancy grid. However, this model fails to exploit many potentially valuable secondary cues - such as material dropped by the missing person or unmarked tracks - which could aid in the search process. In this paper, we develop a Bayesian approach to exploit this secondary evidence. Our approach is inspired by the stigmergic approach to indirect coordination: evidence left by the missing person on the ground is used to coordinate the actions with the searching UAV. To achieve this coordination, we compute the joint probability over multiple cells using a path-base representation of the missing person trajectory. The trajectory is modelled using an agent-based simulation. As new evidence becomes available, a resampling scheme is used to update the ensemble of paths. We demonstrate the performance of the algorithm in a simple search scenario, and show a significant improvement over current search methods. (5 pages)