Stereo vision-based autonomous navigation for lunar rovers

Purpose - To provide an autonomous navigation system to endow lunar rovers with increased autonomy both for exploration achievement of scientific goals and for safe navigation. Design/methodology/approach - First, algorithm and technique of initial position determination of lunar rovers are introduced. Then, matched-features set is build by multi steps of image processing such as feature detection, feature tracking and feature matching. Based on the analysis of the image processing error, a two-stage estimation algorithm is used to estimate the motion, robust linear motion estimation is executed to estimate the motion initially and to reject the outliers, and Levenberg-Marquardt non-linear estimation is used to estimate the motion precisely. Next, a weighted ZSSD algorithm is presented to estimate the image disparities by analyzing the traditional ZSSD. Finally, a virtual simulation system is constructed using the development tool of open inventor, this simulation system can provide stereo images for simulations of stereo vision and motion estimation techniques, simulation results are provided and future research work is addressed in the end. Findings - An autonomous navigation system is build based on stereo vision, the motion estimation algorithm and disparity estimation algorithm are developed. Research limitations/implications - The field test will be done in the near future to valid the autonomous navigation algorithm presented in this paper. Practical implications - A very useful source of information for graduate students and technical reference for researchers who work on lunar rovers. Originality/value - In this paper, stereo vision-based autonomous navigation techniques for lunar rovers are discussed, and an autonomous navigation scheme which based on stereo vision is presented, and the validity of all the algorithms involved is confirmed by simulations.