Unstructured road detection and path tracking for tracked mobile robot

In this paper, we propose an unstructured road navigation system to achieve the visual navigation for tracked mobile robot. The environment information and unstructured road information are obtained through the understanding of sensor data. Then the robot can be automatically controlled to track along the path. We have two contributions. Firstly, we have improved the efficiency of the algorithm for detecting unstructured road and achieve the real-time requirements. We achieve this by down-sampling the pictures and making optimization of a local soft voting method which is used to search for road vanish points. Secondly, we propose the using of the velocities of two wheels to control the robot rather than using velocity and angular velocity, which is generally utilized in most motion control systems but is difficult to control in practice. The road regional center is regarded as the preview points to imitate artificial driving behavior and the velocities of two wheels are exported respectively. Then the robot is controlled to track the available path using fuzzy control technology. Simulation and experiment results validate the efficiency, accuracy and practicality of this proposed system.