The Lunar Rover Location And Path Planning Based On Stereo Vision

Localization and path planning are two important problems for lunar rover navigation. In order to achieve autonomous navigation in the moon environment, the lunar rover must know where it is all the time, and can planning a reasonable route based on the current environment. This paper is focus on localization and path planning for lunar rover based on stereo vision.The robot must locate itself accurately when it moves around in the environment, unless any autonomous behavior is blindness. In the second chapter, the dissertation analyzes the robot's motion model, calculates the pos and the direction in the ideal condition through reading the encoder. Of course the far the robot moves, the more the locating error grows by this method.In order to modify the error created by original odometer, the iterative closest point algorithm (ICP) is adopted in this paper. There are two improvements to assure the real-time ability and precision. Firstly, reduce the number of the iterative points by introducing the edge detection based on scan line approximation. It is easy to affirm the correspondence relationship in the edge points, thus help to reduce the pseudo point matches. Secondly, as the most computation time of the ICP is spent in the closest point search step, it is necessary to speed up the search process, so a fast algorithm for search the points based on KD-Tree(K=3) is introduced, which allows a point to find its closest point in the time cost of O( log 2N ). The final experiments prove the method possess of real-time ability and precision.Path planning is one of the most important tasks of lunar rover's navigation. It means to search an optimal or approximate optimal free path from start state to target state according to some optimize criterion. A method of distance vector histogram based on expand grids is proposed to resolve the path planning problem in the paper. First, take into account the sensing uncertainties and the robot size, expend the grid which is detected as an obstacle point on the created grid map; secondly, build the local distance vector histogram and segment each obstacle; thirdly, choose a threshold as the minimum distance to determine the Obstacle Avoidance Area (OAA) and Free Walking Area (FWA); Finally, choose a Free Walking Area as the robot's current moving area according to the target direction. The experiments prove that...