Research On Mobile Robot Environment Modeling

The basic requirement for a complete autonomous mobile robot to fulfill its various intelligent tasks in an entirely unknown environment is to enable building its environment's map incrementally and locating itself in this map simultaneously. In this paper, I have done some research on mapping environment, such as how to identify objectives with neural network, mobile robot mapping simulation system with sonar data. The main research results are as following:Firstly, we study the design and realization of sonar-based mobile robot mapping simulation system. A mobile robot mapping simulation system is developed with Visual Basic. With interpreting and fusing simulate data from a sonar mathematical model; a grid map of environment has been built increasingly using occupancy grid method based on probability theory and Dempster-Shafer evidence theory. The simulation system could set and modify environment model and movement status of mobile robot on line. The mathematical model of sonar that its parameters could be adjusted and the map building model are relatively independent in the system. The simulation system could make the research of mobile robot map building more convenient.Secondly, we introduce the design of Back-Propagation neural network based on sonar data for differentiating environmental barrier. Compared to the target differentiation algorithms, the BP NN has higher accuracy and less error.We use MATLAB program to gain sonar data, at the same time, we use the NN toolbox of the MATLAB to process sonar data and get fine error curve.Thirdly, we present a target differentiation method that based on LVQ (Learning Vector Quantization) network for mobile robots. The typical targets are differentiated efficiently in indoor environments with LVQ network fusing the time-of-flight data and amplitude data from sonar systems. The algorithm is simple and real-time and has high accuracy and robustness. It can deal with the uncertainty data of the sonar system effectively. With this method, mobile robots can classify the targets quickly and reliably in indoor environments. The simulation results show that the method is effective.