| Now, the research area of the mobile robot has been extended into underwater, outer space and planet environments, and the research has attained the excellent achievements. In all of the mobile robot research, the exact localization of the robot is an essential and important function. The localization of the robot depends on the accurate environment map that the robot is located; however, the accurate environment building depends on the location of the robot. Thus, the environment map building and the robot localization are related with each other. In order to realize the real automatic of the robot, these two problems should be considered as the joint one, which can make that the robot has the ability of the localization and mapping at the same time. The thesis focuses on this problem, especially on the environment mapping, robot localization and the multi-robot map fusion algorithm.In chapter 1, first, the present situation of researches for robotics is illustrated. Then the algorithm of the mobile robot simultaneous localization and mapping algorithms are classified, and final, the research contents and the architecture of the thesis are presented.In chapter 2, for the 2D laser's measurement data, a new line segment method is proposed. First, the prediction model of the adjacent measurement data and the measurement model of the measurement data are presented. Then the unscented kalman filter is used to estimate the output of the prediction model and the measurement model, the Chi-square distribution between the measurement data and the measurement model output is constructed, and according to the attribution of Chi-square distribution, the line breakpoint is detected. Final, a line fitness method is proposed in the polar coordinate. Through constructing the optimal angle and optimal distance in the polar coordinate, the abstract description of the laser measurement can be acquired.In chapter 3, the optimal sparse time, the appropriate method of the sparse operation, and the maximal memory used in information matrix are proposed for the sparse extended information filter of the simultaneous localization and mapping. First the simultaneous localization and mapping is partitioned into three stages, and by comparing the time consumption, information loss, and the optimal time for applying sparse operation is given. Then, by comparing the time consumption of the iterative and batch operation, the optimal operation method is proposed. Final, the maximal memory consumption by the information matrix after all the sparse operation is derived.In chapter 4, the improved re-sample method and the sampling function are proposed for FastSLAM algorithm. First, by Monte Carlo method, the elements that influence the consistency of the robot pose in FastSLAM are analyzed. Then, the distribution of the particle weight, each particle's measurement consistency and the effective particle number are referenced for the re-sampling condition judgment. When producing the new particle, the exponential ranking method is used to select the father particles, and the crossover operator is used to create the new particle. Final, an improved auxiliary particle filter is proposed and is used as the particle sampling function in FastSLAM.In chapter 5, the multi-robot map fusion algorithm is proposed which is based on the relative measurement data between each robot. First the robot pose and landmark position transformation equations are derived based on their relative measurement data, and the fusion map model is constructed. Then, by extended kalman filter, the trace of the fusion map covariance is deduced, and it is used as the select criterion for the map fusion reference coordinate. Final, the Chi-square distribution between each two landmarks is used to detect the duplicated landmarks, and the extended kalman filter is used to fuse the corresponding duplicate information in the fusion map.In the final chapter, the thesis is concluded and the prospect of the future research is presented.
|
PDF Full Text Request