Intelligent Navigation Research Of Autonomous Mobile Robot

This dissertation research the intelligent navigation problem of a four-wheeled autonomous mobile robot, it mainly contains three parts.In the first part, an indoor global map-building method based on machine vision is proposed. After transformation of color space to the original RGB image, the obstacles can be picked up exactly and fully through threshold segmentation. And a quick algorithm of polygon approximation based on geometrical form and rules is presented to fit the obstacle edges, the processing result — an indoor map described by polygons, is in good agreement with the edges, and the critical points is few, so the computing burden of the succeeding path-planning is reduced.In the second part, considering the requirement for safety of the path, a novel algorithm for mobile robot path-planning based on vector field model is proposed, using this model, a satisfying path can be planned with the two criteria, which are length and safety respectively. And a difficult problem suffered often by potential field method — local minimum is researched, a preventive method based on virtual local target is presented, the key of the method is when and how to appoint the virtual local target, simulation results in several complex environments prove the validity and adaptability of the algorithm. Also, in order to extend the generality of the algorithm, the geometrical catalog of the obstacle in the planning workspace is expanded from rectangle to polygon, and in order to optimize the path in the sense of several factors including the length, the smoothness and the safety, a hybrid optimization algorithm, HGSA, which incorporates the simulated annealing algorithm (SA) into the genetic algorithm (GA), is proposed, the effectiveness of the planning method is verified by computer simulation in three environments with different obstacle distribution, and the comparison of the optimized paths show that the hybrid algorithm outperforms either GA or SA.In the third part, after deducing the kinematical model of the four-wheeled mobile robot in discretized form, a fuzzy controller, which can benefit from the driving experience of human, is designed to control the mobile robot for the pathtracking. And an algorithm named slow-down FID control is applied in the last section of the tracking to drive the mobile robot to stop at the anticipant goal correctly. The simulation and experiment results show that this dual-model control method realizes the expectant objective, and it ensures the smoothness of the robot moving.