Speeding Up Person Finding Using Hidden Markov Models

The ability of mobile service robots to efficiently search for a person is needed in a vast domain of applications. The search problem is especially challenging when the user is freely moving across the environment, the robot has only a constrained field of view, and visibility constraints arise from the environment. We propose in this article a novel approach that simulates the user’s presence at different locations in the environment based on a hidden Markov model (HMM). The HMM predicts the user’s motion and computes the observability likelihood at the different locations given the predictions. Our approach then selects ef- fective search locations that maximize the user’s expected observability. The selection criterion hereby considers the visibility constraints along the robot’s path as well as the robot’s travel time to reach the search location. We per- formed both real-world and extensive simulated experiments to evaluate our method. In comparison to a greedy maximum coverage approach as well as to a greedy strategy that uses background information, we show that our framework leads to a significant reduction of the time needed to find the user