Mobile Robot Localization And Mapping Based On Information Fusion

Localization and mapping are the key issues for mobile robot. They are prerequisites for autonomous mobile robot and bases of the following path planning and motion control. Localization is to determine the robot's own pose. Mapping includes environment perceive, target identification, obstacles detection and environment expression.Accurate localization is based on the environment map. To build environment map, robot must know its location at every observation point. If mobile robot moves in a completely unknown environment and can not be located by absolute positioning equipment, such as GPS, it is in a dilemma:to locate itself, robot needs accurate environment map; to build a map, robot needs to know its pose of every moment. In this case, simultaneous localization and mapping is necessary.In order to achieve accurate localization and mapping, robot needs to use sensors various in space, time, confidence, presentation and usage to explore the environment. These sensors information are fused to improve the accuracy, efficiency of environment exploration and so on.For above issues, the following aspects were studied:1,The issue of building a grid map based on multisensor information fusion was studied. The conception of information fusion was defined. The procedure and the basic method of grid map building based on Dempster-Shafer evidence theory were demonstrated.2> Approximate process algorithm was proposed with respect to shortcomings of the Dempster's rule of combination. The algorithm solved the problems that the Dempster's rule of combination of Dempster-Shafer evidence theory can't be applied to information fusion under certain circumstances and there will be counter-intuitive behavior in Dempster's rule of combination in some cases. The D-S grid map was analysed. The characteristics, judgement methods and judgemnet procedure of five types of gird in D-S grid map and improved Dempster-Shafer evidence theory decision rules were proposed. Circumstance valuation and circumstance exploration effect valuation based on D-S grid map were suggested. Simulation result shows that the grid map building method based on improved Dempster-Shater evidence theory is appropriate.3,Ground metal detection mobile robot system was developed and environment exploration and grid map building were achieved based on this system. Experimental result shows that the multi-ultrasonic-sensor mobile robot grid map building method based on improved Dempster-Shafer evidence theory is appropriate. The system has delivered, with favorable results.4,For the issue of simultaneous localization and mapping, an improved RBPFs-SLAM algorithm was suggested. RBPFs-SLAM algorithm was improved in two aspects:proposal distribution and resampling strategy. Accurate proposal distribution computation method, adaptive resampling strategy and the whole algorithm procedure were suggested. Simulation result shows that the improved RBPFs-SLAM algorithm is better than that before improvement.