Research On Mobile Robot Decision And Planning Based On Multi-sensor Fusion

In recent years, multi-sensor fusion technology is more and more attention, it contains control theory, probability and statistics, intelligent algorithm, signal processing and other fields. Multiple sensor fusion technology is different from the previous sensor independent testing .It’s rather through simple specific algorithm, different sensors to collect information on comprehensive treatment, than the accumulation of the data collected from the sensors, and get a more accurate estimate of the information. As is known to all, in the robot technology, robot surrounding environment is often uncertain, complex, or unknown, this needs the robot can get accurate perception system the information of environment to make the robot more accurate decision making and planning. Therefore, multi-sensor fusion technology in decision making and planning of the robot, is now a hot study issue, and one of the important research direction in the future.This paper mainly analyses the application of multi-sensor fusion in the mobile robot decision and planning, and compare the precision of robot localization and the planning effect before and after adding multi-sensor fusion link. The article elaborates the development trend of mobile robots at home and abroad, the main methods of multiple sensor fusion,and the advantages of multi-sensor fusion technology in complex and unknown environment around the robot. Then many sensor fusion algorithms is compared, and square root no trace kalman filter (SR-UKF) is selected as a multi-sensor fusion algorithm. This paper elaborates the steps of sensor fusion algorithm based on the square root no trace kalman filtering, and the improvement on this basis. Then the improved method is used to fuse the cross-correlated sensor information to get accurate information. The real-time robot localization and building robot’s surrounding environment map which is simultaneous localization and mapping problem (SLAM), can proceed by using environment information fusion. In local maps the nearest feature points is selected from the target points, as a local target, and then plans out a optimal route to get to local target in the local map by ant colony algorithm. And when it gets to the local target, the robot builds a new environment map, and real-time update local target. This step is iterated until robot getting to global target.Finally, the algorithm simulation of MATLAB is based on sensor systems model of the ultrasonic sensor and laser range-finders. The contrast of positioning error before and after adding sensor fusion link can be got, and the positioning error based on SR-UKF multi-sensor fusion and extended kalman filtering (EKF) now commonly used carries on the comparison, to demonstrate multi-sensor fusion algorithm’s superiority. Then, the whole system planning simulation is made to test the feasibility.