Mobile Robot 3D Map Building And Path Planning In A Complex Outdoor Environment

Map building and path planning are two essential problems for mobile robot, and there are many research results towards indoor environment. But with the fast development of new technology in intelligent control, computer science, networking, bionics and artificial intelligence, the research field of mobile robot has transferred from indoor environment to outdoor environment, which is a challenge for mobile robot map building and path planning. This paper mainly studies the two essential problems in 3D (three-dimensional) outdoor environment.In order to acquire 3D laser point cloud, a 3D laser range finder system is built on the basis of a 2D (two-dimensional) scanner. And a CCD (charge coupled device) camera, which can provide color and texture information, is used as auxiliary equipment. A novel method is proposed for extrinsic calibration of the 3D laser range finder system and the CCD camera. Calibration results are demonstrated by coloring 3D points.Towards outdoor environment, an elevation map is utilized to represent outdoor terrain feature. After projecting laser points into different units, according to the height distribution the units are classified into vertical ones and horizontal ones for purpose of describing overhanging structures. A layer projection method is presented in order to obtain more environment information. By connecting projection clusters of vertical units in different layers, topological structure models are formed, which are utilized to identify typical objects by means of BOW classifier. Otherwise, a relationship network composed of topological structure models can be applied for initial map registration and loop closing. A hierarchical strategy, in which vertical units and horizontal units are used for scene matching separately, is presented for achieving a globally consistent registration. Finally, experiment results show the validity and practicability of the proposed methods.By projecting the units of the elevation map into the image, the fusion of terrain feature and ground surface character is achieved using statistical method. Giving consideration to terrain information and ground surface information, the environment constraint is generated. The probabilistic roadmap method is improved for path planning in elevation map. Unlike total random distribution, the nodes in different grid regions are chosen randomly to ensure a uniform distribution in the free space, so that the un-uniform problem of PRM can be solved. The path composed of horizontal units is smoothed to satisfy the nonholonomic constraint of a mobile robot. Experiment results prove that the proposed method can plan safe paths by giving consideration to both environment constraint and nonholonomic constraint.