Research On Localization And Mapping Technology For Mobile Robot In Unknown Environment

The capability of autonomous navigation for mobile robot is one of the mostimportant aspect to determining the robot intelligence level. In order to achieveautonomous navigation, we must solve the problem of localization first. Locatingmobile robot in unknown environment, first we need to build an environmental map.On the other hand building environmental maps need position information of mobilerobot. So Locating a robot in unknown environment is a processe of simultaneouslocalization and mapping (SLAM). SLAM for mobile robots is a fundamentalproblem in the area of intelligent mobile robot, and also thought to be the key to therealization of truly autonomous mobile robot. It is also an important embodiment ofrobot intelligence level.Due to wide range of mobile robot’s movement, its operating environment is ofstrong randomicity. In this article, the locating problem of mobile robot is totallybased on mathematical model of probability theory. And specific works are asfollows:(1) We design the mobile robot system. In this paper, we divide mobile robotsystem into three parts: mechanical system, hardware system and software system.Hardware system is divided into the ground information processing and controllayer, the on-board equipment management and data collection layer, On-boardbottom control layer. The software system is divided into application layer, functionlayer, system layer and the underlying control layer. By using hierarchical andmodular technology, we greatly improve the system’s openness and scalability.(2) We successfully perform the calibration on external sensor KINECT. At thesame time we extract the compressed image information of the KINECT, becausethe data of KINECT RGB image and depth image are too big, and can not meet therequirement of the real-time. For internal sensor CMPS10in this paper, we designthe movement measurement subsystem, using wireless transmission way, increasingthe flexibility of the robot.(3) Base on CMPS10, in this paper, we established the probability of the mobilerobot motion model by reading CMPS10module information of acceleration anddirection angle. Base on the KINECT, in this paper we analyzes the main factorscausing observation error and the observation probability of mobile robot model isestablished. After that we use expectation maximization algorithm to determine therelated parameters of the model.(4) Under the MRPT platform used in this paper, we implemented FastSLAM algorithm based on the robot probability model and probability model. And performpractical application on the robot platform. At the same time, in order to improve theparticle diversity loss caused in the resampling phase of particle filter algorithm inFastLSAM algorithm, we put forward fuzzy resampling algorithm based on fuzzytheory. By comparing with common resampling algorithms, it can be seen that thefuzzy resampling algorithm reduce the loss of the diversity of particles.