Research On Particle Swarm Optimized Fast Simultaneous Localization And Mapping

With the fast development of new technology in intelligent control, computer science, networking, bionics and artificial intelligence, mobile robot has become the focus in the field of robotics and automation. Simultaneous localization and mapping for mobile robot is a key to realizing the navigation of intelligent mobile robots, and an essential foundation when mobile robots engage path planning, environment exploration and global map building.This thesis, combining the demand of research project, emphatically studies the problem of simultaneous localization and mapping for mobile robot.Main contribution and work are described as follows:1. According to the demand of subject, Mobile robot motion model is researched carefully; combining the actual requirement of the motion control, a set of algorithm of robot odometry pose sample estimation is designed. The sampling points generated by the algorithm can cover the mobile robot real position better. Such sampling provides prediction data effectively for robot in next time step's position prediction.2. In the processing of the map building, using 2D laser radar as the environmental perception sensor, according to the indoor's structured environment, an extended line model is designed, and a fast, reliable line segments extraction algorithm is used by such model. This algorithm taking advantage of characteristic of laser degree of accuracy and rapidity fully, using sliding window method extracts the structured environment features. Such algorithm achieves the aim of environmental perception rapidly and simply, and fulfills the requirement of real time navigation.3. A particle swarm optimized simultaneous localization and mapping(SLAM) approach is presented, which integrates particle swarm optimization(PSO) into the SLAM. The particle swarm optimization fused the newest observation is exploited to adjust the particle's proposal distribution, which enhances the effectiveness of the prediction sampling, meanwhile, makes the particle obey the robot's pose posteriori probability distribution and concentrate into the robot's true pose. Because of the optimization of the particle set, the impoverishment of particle is overcomed effectively. The particle number and the computational complexity are reduced. At last, this method is proven correctly and feasibly in simulation experiments and online experimental data.