On Simultaneous Localization And Mapping Of Mobile Robots

Simultaneous localisation and mapping (SLAM) is the process that a mobile robot can incrementally build a map by the relative observations of unknown environment and simulta-neously compute its pose within the map. The solution to SLAM problem is hot and difficult in the field of intelligent mobile robot research, and it is considered by many as the key pre-requisite for making a robot fully autonomous.This thesis mainly focuses on the problem of error accumulation of systematic odometry in the SLAM process. The main contributions are summarized as follows:(1) An introduction is given to the related basic theory of nonlinear filtering based SLAM algorithm including the generalized model for the Bayesian Filtering based SLAM and two major SLAM algorithms based on Extended Kalman Filter and particle filter.(2) To study the problem of error accumulation of systematic odometry in the SLAM process, the error motion model is established according to the widely used robot model of SLAM. A new SLAM method is proposed based on the augmented extended Kalman filter (AEKF) algorithm framework. The specific algorithm procedure of this method is deduced in detail, and a simulation feature map is created for the simulation experiment.(3) The applicability of the robot error motion model in FastSLAM is investigated. The procedure of particle filtering with errors in parameters is deduced. Simulation experiments are done in feature map containing many landmarks for the situations when the number of particles is very small and very big.(4) On the basis of studying the mechanism of robot operating system, the robot platform is built. Some related modules in SLAM package are modified. The SLAM experiment is tested in corridor, and the autonomous navigation experiment is carried out in office.The experimental results show that for the wheeled mobile robot, the errors of orientation and odometry are effectively reduced by using the error motion process model, the estimation accuracy of SLAM state is significantly improved, and the navigation task is well accom-plished.