Research On Positioning Technology Of Electric Inspection Robot Based On Multi-sensor Fusion

With the development of smart grid technology and synchronous positioning technology of mobile robots,intelligent robots for electric field inspection has become a hotspot that is urgently needed for the research focus in various electric industries.At present,preset magnetic tracks and RFID technology are commonly used for positioning of inspection robots.This method has the disadvantages of low flexibility,weak anti-jamming ability and low positioning accuracy,which leads to the low efficiency of the robot and even unable to complete the daily tasks.The positioning method based on single sensor can not adapt to complex the working environment such as power plant and substation.In order to ensure the stability and accuracy of robot positioning,it has become a trend to integrate multiple sensors for positioning.To meet the requirements of low cost,high precision and high stability of the robot positioning system,the multisensor fusion positioning algorithm is studied in this thesis.The main works are as follows:1.In this thesis,ORB(Oriented FAST and Rotated BRIEF)feature point method and optical flow tracking method are compared.Aiming at the problem of feature point aggregation and mismatch in optical flow method,the existing LK(Lucas-Kanade)optical flow method is improved,and its effectiveness is proved by experiments.In order to ensure the real-time performance of the system,a sliding window and key frame selection strategy is designed for monocular visual odometer.A closed-loop detection module is added for monocular vision positioning to eliminate the cumulative error caused by the system’s long-term operation.Through experiments,it is verified that the effect of the improved algorithm and the role of the closed-loop detection module.2.The characteristics of odometer and IMU are analyzed,and they are calibrated and corrected by the error correction model established in this thesis.A multi-sensor fusion external parameter online calibration method is proposed,which synchronizes different frequency sensors by the pre-integration method.Then,taking the external parameter calibration of camera and IMU as an example,the specific implementation process of the algorithm is given.Through experiments,it is verified that the effectiveness of the proposed algorithm which lays the foundation for the realization of multisensor fusion.3.In order to solve the problems of low initialization success rate and large error caused by insufficient IMU excitation on mobile platform,The process of odometry pre-integration and camera alignment is considered,and a method based on dynamic weighting to solve the scaling parameters is also proposed in this thesis.This method improves the success rate of the initialization process and the accuracy of the estimated parameters.A tight coupling back-end optimization algorithm is used,and the system model and observation model are built in order to solve the problem that the monocular vision inertial navigation positioning algorithm is not highly accurate on a mobile platform.Then the pose of the robot is estimated by nonlinear optimization.Finally,the required robot software and hardware platforms and indoor detection platforms are built to test the algorithms in indoor and outdoor environments.Through experiments,it is verified that the fusion positioning algorithm has been improved greatly in accuracy and stability compared with monocular vision-IMU positioning.