Research On Autonomous Positioning Method Of Mobile Robot Based On Information Fusion

As mobile robots become more and more intelligent,visual positioning technology has become a hot research topic as a key technology for mobile robots.However,relying on the visual positioning technology to locate the robot still has the problems of poor robustness and low positioning accuracy,which makes it difficult to meet the positioning requirements of the robot in the actual running process.Therefore,this paper adopts information fusion positioning technology,which uses a variety of information to describe the surrounding environment,and provides more effective positioning information for mobile robots,so that robots can make better decisions.This paper studies and analyzes the commonly used information fusion technology,focuses on the Kalman algorithm,and proposes a v-adaptive Kalman information fusion method for the problems of slow convergence and easy divergence.This method integrates odometer information and Kinect V2 vision sensor.Image information,AR Tag prior information,to achieve real-time precise positioning of the robot.The main work and research results of this paper can be reflected in the following points:(1)Introduced the research status of visual positioning technology and the research status of integrated information technology.(2)Processing information of different sensors,including odometer and AR Tag.And the corresponding motion model is established.The internal structure of the Kinect V2 camera as a visual sensor is also introduced and the calibration experiment is carried out.(3)The system introduces the overall framework of visual SLAM,focusing on the analysis of visual mileage.These include feature extraction,RANSAC,and ICP algorithms.In order to accelerate the convergence of RANSAC algorithm,a RANSAC algorithm based on constraints is proposed to improve the convergence speed of the algorithm.And calculate the important parameters v required in theinformation fusion algorithm.This parameter determines whether the corresponding visual odometer is abnormal by the number of internal points,thereby eliminating some bad visual information.Provides better visual positioning information for information fusion algorithms.(4)Build a variety of visual SLAM framework operating platforms,which are popular today.Through test and analysis,a number of performance comparison charts are obtained.By analyzing the test results,an algorithm framework suitable for the visual odometer studied in this paper is proposed.(5)Introduced the information fusion structure model and focused on the Kalman information fusion algorithm.Aiming at the shortcomings of Kalman information fusion algorithm,a v-adaptive Kalman information fusion method is proposed.Since the Kalman information fusion algorithm has the disadvantages of divergence and slow convergence,the method determines the coefficients by the convergence condition,so that the method can suppress its divergence.At the same time,the observed noise is adjusted online to make the estimated value closer to the actual value.(6)Through different scenarios and different forms of motion,the v-adaptive Kalman information fusion method,EKF,visual odometer and odometer were compared.Experiments show that the positioning effect based on v-adaptive Kalman information fusion method is the best,and the average positioning accuracy can reach0.05 m,which greatly improves the positioning accuracy.The experiment also proves the effectiveness and rationality of the method and better robustness.