A Carrot in Probabilistic Grid Approach for Quadrotor Line Following on Vertical Surfaces

Most current research deals with line following at an aerial position with respect to the target object. We address a task scenario of close-up inspection on vertical surfaces. Line following is a basic component for the close-up inspection process on high rise structures, such as building façade, tower skeleton, and wind turbine blade. The inspection process may also require slower and accurate line following movement for anomaly detection in higher resolution. In this paper, we formulate the problem of accurate line following on vertical surfaces. We propose the carrot in probabilistic grid algorithm for accurate line following on vertical surfaces and work through a refinement for performance improvement. We implemented the carrot in probabilistic grid algorithm on a ready-to-fly quadrotor (micro rotary UAV) and evaluated the line following performance with several forms of geometric line segments on a vertical surface. Experimental results based on extensive actual flight tests show satisfactory performance of the carrot in probabilistic grid algorithm over the benchmark line following algorithm.