Estimation of user location for hearing-dog robot using past experience

A hearing dog plays a role in assisting hearing-impaired people. The touch behavior from hearing dogs to the people can alert important life sounds, such as doorbells, alarm clocks and fire alarms, to the people. Hearing dogs are useful to assist the people, but the number of hearing dogs is insufficient all over the world. Instead of hearing dogs, hearing-dog robots have been developed until now. The hearing-dog robot can autonomously move and search a user to be communicated. This paper aims to improve a searching algorithm of the robot. We introduced a stochastic model to the robot behavior. The stochastic model uses past experience about locations where the robot has explored and the user stayed in the past. An experiment was conducted to evaluate the stochastic model. The experimental result indicated that the stochastic model worked efficiently to search the user.