Research Of Real-Time Mobile Robot Localization And Map Building In Unknown Environment

The robot has interdisciplinary combined with the microelectronic technology, automatic control, sensor technology, artificial intelligence, computer technology and so on, and it becomes a new high-tech frontier disciplines, attracting a large number of domestic and foreign scholars to expand and research. The mobile robot technology is an important part of the field, and its main purpose is to enable the mobile robot to achieve some voluntary movement and complete some of the work specified through its own perception of the sensor in unknown environment. Among them, the intelligent navigation research goal is to make the robots to move to a specific location and complete a specific task or a specific operation according to pre-set purpose under the condition of fully autonomous. The robot’s real-time localization and map building is the foundation of the intelligent navigation. Robots are not identified in its location conditions, to autonomously locate and draw the map of the environment, and on the basis of navigation in a completely unknown environment, This is the real-time mobile robot localization and mapping (SLAM) problem.This paper is mainly to improve UKF-SLAM algorithm, enhance the estimation accuracy of the UKF-SLAM algorithm, at the same time, reduce the computational complexity. Then create indoor robot experiment platform for application on the algorithm.The main contents of this paper include:First of all, do mathematical modeling for a mobile robot and environment SLAM research. Respectively establish motion model, observation model of robot, feature map and its augmented model environment. Also, using the noise model to deal with uncertainty about the noise.Secondly, it introduces the theoretical basis of the Kalman filter, and stands in the angle of statistics to analyze the SLAM problem. Respectively expound two kinds of commonly algorithm, based on the extended Kalman filter and unscented SLAM Kalman filter SLAM, and in the MATLAB environment, comparison of the two algorithms for simulation experiments and analysis of experimental results.Then, according to the filter divergence caused by the computer rounding errors of UKF-SLAM algorithm, using covariance square root covariance update covariance replacement; And for sampling points of symmetric sampling more, poor real-time and easy to produce non-local effects and other issues, transform sampling strategy for single sampling scaled minimal skew, thus put forward the square root UKF-SLAM algorithm based on minimum skew sampling proportion. Moreover, simulation experiments are used to verify the estimation accuracy and computational complexity of it.Finally, odometer、gyroscope、laser radar and other sensors are used to build SLAM experimental platform for the robot. The pose of robot and the environmental feature maps are extended to the state vector while the laser radar collects the initial environmental data, furthermore, the information of initial pose gathered by speedometer and gyroscope works as observed data for the improved UKF-SLAM algorithm in the experiment on the experimental platform, as a result,it gains real time information of the pose of robot and the environmental feature maps, which proves the effectiveness of the algorithm in practical application.