Learning-based adaptation for personalized mobility assistance

Abstract

Mobility assistance is of key importance for people with disabilities to remain autonomous in their preferred environments. In severe cases, assistance can be provided by robotized wheelchairs that can perform complex maneuvers and/or correct the user's commands. User's acceptance is of key importance, as some users do not like their commands to be modified. This work presents a solution to improve acceptance. It consists of making the robot learn how the user drives so corrections will not be so noticeable to the user. Case Based Reasoning (CBR) is used to acquire a user's driving model reactive level. Experiments with volunteers at Fondazione Santa Lucia (FSL) have proven that, indeed, this customized approach at assistance increases acceptance by the user. © 2013 Springer-Verlag.