Development of a virtual test bed for a robotic dead man's switch in high speed driving

Abstract

Driver incapacitation whilst driving can lead to serious hazards for other road users, especially in a high speed driving environment such as a highway or a motorway. This paper examines the possibility of incorporating an autonomous system in road vehicles that can safely steer the vehicle to the highway shoulder in the event of driver incapacitation. As a first step towards this development, we have developed a virtual test bed to rigorously test and simulate a variety of road conditions for such an autonomous system without risking injury or accident during actual tests. A simulated highway driving environment (car racing game) is chosen for implementation. A high performance vision system captures the life-like gaming scenes projected on a high resolution LCD screen. The feedback loop is closed using three modules: the interfacing unit, the image processer and a control unit. The image processor estimates the car state using standard computer vision algorithms. The computed car orientation and speed data is passed on from the image processer to the control unit where steering, acceleration and braking decisions are made using state of the art algorithms such as the Stanley method. Appropriate control signals are transferred to the interfacing unit which is connected to a toy steering wheel and pedal set. We demonstrate basic control maneuvers such as lane changing and steering of the car in controlled experiments with little or no traffic. Reasonably accurate results are achieved for these basic tests and directions for further development have been discussed.