Stereovision-based real-time obstacle detection scheme for Unmanned Ground Vehicle with steering wheel drive mechanism

This paper highlights the importance of an autonomous navigation scheme for an Unmanned Ground Vehicle (UGV) operating under a complex operational scenario that requires obstacle detection. The proposed scheme is based on the construction and investigation of disparity images and follows a two-stage perception structure. During detection phase, the relationship between encountered obstacles and the robot's path is inferred. Based on the extracted Region Of Interest (ROI) and statistical information about projections, the confirmation phase characterizes the contours and obstacles positions. Computationally intensive stereovision techniques are optimized for use in real-time applications. The scheme has been tested on a custom-developed wheeled mobile robot. The experimental findings show the benefits of our scheme. The robot detects obstacles of any size and shape within a range of 80–200cm. The results demonstrate that the robot has the ability to precisely navigate in a wide range of illumination conditions. The proposed approach does not need any extraction of lane markers since it fully exploits the information contained in the disparity images. The anticipated applications of the proposed scheme include autonomous navigation of vehicles for environment exploration, perimeter surveillance and automated delivery of goods in a factory.