Multilevel intention estimation for wheelchair control

A novel approach for supporting severely disabled in steering electrically powered wheelchairs is presented. The proposed assistance system estimates the user's intentions on different levels of abstraction to enable navigational support in known and unknown environments. Fusing the estimation levels' outputs according to the quality of estimation ensures that optimal support can be provided in the present situation. The person's gaze behavior is used as additional source of information, to compensate for the reduced input, severely disabled can provide via speciality controls. Thus, the described assistance system will considerably increase the usability of powered wheelchairs for this group of people.