Research On Rough-Set-Theory-Based Mapping For Mobile Robot

With the development of science and technology, the robotics research has been extended from manufacturing industry to many different domains such as aerospace & aviation, military application and so on. Robots should carry out their work autonomously in some environments, where human beings can not enter, for example, the hazardous mine and nucleus relics. Robots should explore the environments and execute different intelligence tasks by themselves. Mapping is the premier task for autonomous navigation and various intelligence tasks. It is also embodies the perceive ability and the intelligent level of mobile robots. Thus, mapping is very important in robotics. In this study, rough set theory and neural network are applied to the mapping-especially the target differentiation in unknown environments for mobile robots.Firstly, the dissertation reviews the research on map building and its relative aspects in the world. And the background and the main contents of this dissertation are described briefly.Secondly, LVQ neural network, which is suitable to classification problem, is employed to process sonar data to differentiate the typical encountered targets in indoor environment. The multi-level hierarchical LVQ neural network is presented to improve differentiation accuracy based on amplitude and TOF data. The robustness of the system to noise, partial removal of amplitude and TOF data and distance anddifferent targets (for example, acute corner with different θ_c, edge with different θ_cfrom training samples) has been studied. The simulation results prove that the LVQ neural network can differentiate the typical targets in a large ranging scope in indoor environment effectively and rapidly with high robustness to noise and partial removal of amplitude and TOF data and different targets in some extent.