Multisensordata-fusion Based SLAM For Substation Inspection Robot

With the development of smart grid technology,mobile robot Simultaneous localization and mapping(SLAM)worldwide,smart substation inspection robot has become an urgent hot spot for countries to research and develop.At present,the commonly used magnetic tracks and autonomous positioning and navigation technology with RFIDs have disadvantages of large environmental modification,weak anti-interference ability,low positioning accuracy,and poor robot flexibility,thus perform poor in assisting robots execute the daily inspection tasks.SLAM technology,based on single sensor,cannot adapt well to the complicated environment of substations.Based on the traditional RBPF-SLAM idea,this dissertation comes up with a fusion algorithm which incorporates the fusion visual closed-loop detection system,constrains robot position and pose on the back-end with graph optimization theory,and eliminates the cumulative error of traditional laser SLAM.The main research work includes the following contents:This dissertation first analyzes the traditional RBPF-SLAM algorithm idea,models the odometer and the laser radar used in this algorithm,and researches the information extraction and processing methods of the two kinds of sensor.Combining with the algorithm idea,the dissertation proposes the construction methods of importance density function and likelihood function,and analyzes the construction mode of grid map.Secondly,based on the accumulated error of the traditional RBPF-SLAM,a fusion vision closed-loop detection system is proposed to eliminate the cumulative error.This dissertation uses the ORB feature to construct the pouch,extracts the key frame and judges the closed loop.At the same time,an overall relocation mechanism was added to deal with the problem of map rupture caused by the fast movement of the robot.Based on RBPF-SLAM,the robot position and pose and the visual closed-loop signals were obtained,and the back-end optimization system was constructed.By constrained optimization of the robot position and pose,the final global map was built and updated.Finally,a series of experiments were performed in the simulated substation environment both in the laboratory(9.6 m×15.6 m)and outdoors(40 m×80 m)for the traditional algorithm and the fusion algorithm.Experiments show that the fusion algorithm can better improve the closed-loop inaccuracy of maps caused by the accumulated error of the traditional RBPF-SLAM algorithm under.At the same time,it can also avoid the problem of subway fracture caused by robots moving too fast.It also verifies the applicability of the fusion algorithm in the substation environment.