A Multimodal A* Algorithm to Solve the Two-Dimensional Optimization Problem of Accompanying a Person for an Intelligent Wheelchair

Abstract

Impaired mobility have far-reaching consequences for handicapped persons and their relatives. Mobile robotic technologies enable intelligent wheelchairs to regain mobility for those affected. Multiple research projects address human aware navigation and the task of following a person for assistive robots. But just a few projects focus on accompanying a person to enable social interaction. Therefore, we present a navigation system for indoor navigation in dynamic and cluttered environments as well as a novel algorithm for accompanying a person.First, we developed an autonomous driving wheelchair for indoor navigation based on the robot operating system (ROS). Thereby, a multi sensor setup using cameras and laser scanner enables localization within a map. People are detected by the same sensors and tracked by a Kalman filter. Afterwards we propose a novel algorithm to achieve a dynamic accompanying behaviour. An attractiveness distribution is introduced to evaluate the possible accompanying positions next to the manually selected target person regarding social interaction. The resulting two-dimensional optimization problem is solved by a novel multimodal extension of the A* algorithm.The proposed intelligent wheelchair is able to navigate in indoor environments and to accompany any person. In addition it allows social interaction while walking to relieve relatives or nursing staff, which otherwise need to push the wheelchair. The aim is to increase participation in everyday life for those affected.