Probabilistic search of human by autonomous mobile robot

摘要

This paper focuses on the development of an approach for an indoor mobile robot to search the elderly person living alone in an unstructured and dynamic household environment. In order to reach the elderly person, to perform different tasks or in case of an emergency, the mobile robot needs to navigate autonomously in the home environment with the perception of presence of the person at a particular place at a specific time. The estimation of possible locations of a person is carried out using a behavior based Markov Decision Process (MDP). The implemented methodology centers two criteria for searching the person, namely distance to the destination and the probability of finding the person at that location. This estimated destination is reached autonomously by the robot. The autonomous navigation is supported by building a detailed grid map of the environment and planning a path to the required target. Upon reaching the destination, a search for the human face is carried out using a Haar cascade classifier. To validate the methodology over a longer period of time, a 3D simulated environment has also been created. This environment resembles a real house environment and features simulation of a small robot, Artos, which performs the task of searching and a human simulation that walks around in different rooms. The results from simulation shows effectiveness of the developed methodology.