Game Against Nature Based Control of an Intelligent Wheelchair with Adaptation to Pedestrians' Behaviour

This paper addresses the problem of synthesis of the control law for intelligent wheelchair navigation assistant, intended to support the navigation in dynamic and populated areas. The method presented in this paper uses a deterministic, model-based prediction strategy to generate the wheelchair motion. The motion has the feature that is acceptable by the patient being transported on the wheelchair. Using the long-term pedestrians motion prediction the minimal collision risk control strategy is applied. While concerning the wheelchair navigation and its environmental interactions, an issue arises which is related to the pedestrians' empathy towards a disabled person being carried by the wheelchair. In this work we propose to use this phenomena for designing adaptive, driving strategy of the intelligent wheelchair. An intelligent motion controller proposed in this paper, generates collision free trajectory based on safe distance policy and evaluation of the environmental response. By comparing the predicted pedestrians behaviour to the real one, the system adapts control strategy coming from game against nature formalism and Hurwicz criterion based solution. The method performance was evaluated in a simulated environment. Relevant simulation scenarios are presented and discussed.