Abstract: Planetary rovers' wheels will encounter sinkage and slip when driving on soft terrain. Real-time sinkage and slip detection plays an important role in rover simulation, wheel coordinated control and soil parameter identification. A real-time wheel sinkage and slip detection method based on monocular vision is presented here. Edge detection is performed on the region of interest of the wheel, and the boundary of wheel-terrain is fitted to a straight line to obtain sinkage. Hough circle detection and corner detection are performed on the wheel image to obtain the pixel length of wheel rotation distance. Using optical flow method to derive the pixel length of the wheel displacement distance from ground image, then calculating slip from wheel rotation distance and displacement distance. Only one monocular camera is used to detect the two important parameters, wheel slip and sinkage, in real-time. There is no need to convert the pixel length into actual distance during the data processing, the calculation is small and the accuracy is high. The validity and effectiveness of proposed method is verified by using a single-wheel testbed to carry out the typical wheel and typical working condition tests, and compare results of proposed method and single-test bed. The results show that the relative error of wheel slip is no more than 3%, the measurement error of wheel sinkage is no more than 3 mm.

Key words: rover wheel, slip, sinkage, monocular vision