Extraction Of Road Feasible Region For Mobile Robot With 2D Laser Range Finder

As an important technology for autonomous navigation of mobile robot, road detection attractsmuch attention from overseas and domestic researchers. The laser rangefinder is widely used inmobile robot navigation for its advantages, such as high measurement speed and high precision.Therefore, the studies on drivable region extraction based on laser rangefinder are crucial for robotautonomous navigation in theory and applications.Firstly, the thesis analyses the existing 3D laser distance measuring systems and their ownsituations of use, characteristics, etc. According to the specific situation of the road environment,the thesis analyses the proposed 3D laser distance measuring system used in this study—thestrategy of laser range finder combined with mobile robot platform. Three coordinate systems of themeasuring system are established: the local coordinate system of laser sensor, the local coordinatesystem of robot and the global coordinate system of robot. And then, the transformation formula ofthe measurement data the laser range finder acquired in the three coordinate systems is analyzed.Experimental results show that the measuring system can acquire the real-time and onlineenvironment information. The distance measuring system obtains the environment information, andthen, supplies the data for subsequent extraction algorithms.Secondly, a kind of road feasible region extraction method based on a 2D laser rangefinder isproposed in this study. 8-Neighbor searching method has obvious local characteristics. However,the global searching method can make full use of all data points. After discussing the advantagesand disadvantages of the above two methods and combining with the interlacing searching approach,a data point searching strategy based on eight-neighborhood is proposed. Considering the greatcorrelation between the point and neighboring points, a kind of road boundary and obstaclesdetection method is proposed based on eight-neighborhood, in which, judgment criterion of heightdifference threshold and angle threshold are used. Experimental results show that the proposedalgorithm can extract the feasible region of the environment effectively. It can satisfy therequirements of mobile robot navigation.Finally, considering that the feasible region is relatively flat and contains a lot of straight linefeature, a kind of feasible region extraction algorithm based on clustering and region growing isproposed. Straight line feature is more advanced than point feature in describing roads, thus, linefeature is applied in this study to extract the feasible region of roads. In order to reduce thecomplexity of processing data in 3D space, the laser scanning plane is selected as the 2D projective plane. In the laser scanning plane, the single frame laser radar data is classified by the clusteringmethod based on data point space distribution proposed in this study. According to the size of robot,a rough solution is proposed to deal with the width problem of feasible road region. Straight linefeature of the road is extracted by using the least square straight line fitting method, and the twoendpoints in place of the whole line segment are acquired by calculation. Moreover, the twoendpoints of line segment are mapped to the global space to replace space conversion of straightline segment. Meanwhile, sideslip problem of robot is taken into account in this study. On this basis,region growing is executed for the local area composed of the two line segments with the shortestdistance between them in the former and latter moments. Experimental results show that theproposed algorithm can effectively extract road feasible region.