Study On Particle Filter Localization For Internet Based Mobile Robots

Internet based mobile robot (also called network mobile robot in this thesis) is the combination of WWW technology and mobile robotics. The problem of remote measurement & control of mobile robots via Internet has not been successfully sovled due to the practical limitations such as Internet bandwidth. One of feasible solutions of Internet based tele-operation is supervisory control. However, it needs that mobile robots have relatively strong autonomous capability.Self-localization, especially the global localization in which the initial pose of the robot is unknown, is the key for a mobile robot to realize autonomy. The main research work of this thesis can be concluded as follows.1. A Bayes grid map is constructed using single rotating sonar. The map is improved based on data fusion and successfully used in the mobile robot global localization.2. The detailed implementation procedure of particle filter algorithm for small low-cost mobile robots is studied and successfully applied to localize the small mobile robots with low performance. A histogram matching is utilized in the sensor-data-update phase of particle filter localization. Experiments show that it is quite suitable for large amounts of low-precision sonar data and that it makes the algorithm more efficient and robust.3. Java is used to build the remote measurement & control system of mobile robots based on B/S mode. The idea of"online measurement, offline calculation"is introduced to the mobile robot localization, which has effectively covered the shortage of slow-data-processing speed of low-cost robots, and successfully solved the global localization of low-cost network mobile robots working in supervisory control mode.