A rule-based assistive control algorithm for safe navigation for a powered wheelchair

This paper addresses the problem of safe navigation in an environment with randomly placed static obstacles. We convert a commercial powered wheelchair into a semi-autonomous vehicle with limited sight (environment awareness), by instrumenting it using off-the-shelf ultrasonic sensors and associated electronic boards (Arduino). In the continuity of our previous work where we had used stochastic dynamic programming to formulate optimization problems which led to relatively large size look-up tables that can be used as supervisory control, here we propose to extract rules using those results (that data). The advantage of this approach is a low-computational cost for future online implementation, and the drawback is a suboptimal policy. The feasibility is assessed by running simulations in a fairly realistic environment (Unity3D).