Research On Localization Algorithm Of Patrol Robot Based On Fusion Of Vision And Wheel Encoder

Robot positioning technology is a problem that has been studied extensively in robotics,but still needs to be further improved.Patrol robot is an increasingly common and important application scenario of robot positioning technology.Classic used to patrol the positioning method of robot mainly global sensor,need fixed position in the environment to deploy auxiliary positioning equipment,the method need to elaborate on the auxiliary positioning equipment calibration in advance,and work can only be limited to a certain range,the patrol robot cannot be widely applied.In the national key research project "key technology research" patrol robot(2017 yfc0806504)subject,focus on research and development in anhui province plan "has the industry's capacity for independent judgment and disposal the key areas of outdoor all-weather patrol robot is the key technology research and demonstration prototype"(201904 d07020007),the support of the project,in order to achieve the patrol robot in the rapid deployment of indoor and outdoor unknown environment,this article uses visual and wheel speed meter two local sensors to locate,through a combination of visual and wheel speed sensor characteristics,This paper puts forward a kind of based on visual And wheel speed meter patrol robot navigation And positioning system,sensor information fusion And the positioning technology is studied,using a framework based on called Teach-And-Replay,namely is divided to two phase diagram And autonomous navigation,under construction phase diagram,control robot on patrol route during movement,the process is proposed in this paper a kind of based on visual And wheel speed meter tightly coupled SLAM algorithm for reconstruction of environment And patrol track estimates,in the stage of autonomous navigation,this paper proposes a map based on transcendental vision And real-time positioning algorithm of wheel speed meter.The algorithm proposed in this paper does not need to modify the patrol environment and can quickly deploy tasks.The research work has certain academic research value and potential application prospect in promoting the application of patrol robot.The main research and innovation work of this paper is as follows:(1)In view of the real-time and precision problems in the mapping stage,this paper adopts the nonlinear sliding window state estimation optimization combined with loop detection and pose map optimization methods,and proposes a SLAM system based on the tight coupling of vision and wheel tachometer sensor information.In this system,aiming at the problem of the fusion of low frequency visual information and high frequency wheel tachometer information,the pre-integral measurement and error propagation process of wheel tachometer are deduced.In order to solve the scale-free initialization problem of monocular vision,a fast and accurate system initialization process combining vision and wheel tachometer information is proposed.Through simulation experiments and practical verification in representative indoor and outdoor environments,the algorithm proposed in this paper has been greatly improved in precision and efficiency.(2)Aiming at the real-time positioning problem based on the prior visual map in the autonomous navigation stage,this paper adopts the fusion method of extended Kalman filter and proposes an EKF fusion system based on the repositioning information of the prior visual map and the information of the wheel tachometer.Through the design of three independent threads,the high frequency location information output based on local reckoning of wheel tachometer and the low frequency pose correction based on EKF fusion are realized.Finally,the experimental results show that the proposed algorithm can obtain the localization information output with smoother local and lower global errors.