Research On Multi-sensor Fusion Location And Mapping Of Service Robot

In recent years,mobile service robots have continued to develop towards practicality and intelligence.Simultaneous Localization and Mapping(SLAM)based on sensor information is a prerequisite for mobile service robots to complete a variety of complex tasks.At present,positioning and mapping technology of mobile robot is still immature.Service robots need to face the challenges of complex indoor environment and the interference of dynamic objects on positioning mapping.It is difficult for a single sensor to achieve the established tasks.The fusion of multiple sensors enables the robot to perceive its own motion state and environmental information from multiple dimensions,and can provide rich information to the positioning mapping system,which effectively improves the environmental adaptability of mobile robots.Therefore,it has become a research hotspot in recent years.This paper integrates the advantages of vision sensors and inertial measurement units,studies high-precision SLAM algorithms based on the fusion of multi-camera cameras and inertial measurement units,and uses multi-camera cross-loop relocation algorithms combined with lidar to build high-precision suitable for mobile service robot positioning and navigation.Occupy the grid map.The main work of this article consists of the following three parts:1.In order to improve the adaptability of the positioning mapping system to the environment,this paper proposes a scalable multi-camera and inertial measurement unit tightly coupled,which can effectively expand the range of visual observation per unit time multi-vision SLAM.The system is divided into a multi-vision front-end odometer module and a multi-vision back-end loop optimization module.The multi-vision front-end odometer includes three parts: measurement data preprocessing,initialization and local sliding window optimization.The local optimization strategy of key frames and sliding window ensures the real-time performance of the system.The accuracy of the proposed multi-vision front-end visual odometer is compared with the Vins-Fuison binocular version,and the results show that the proposed algorithm has higher accuracy than the comparison algorithm.2.This paper propose a multi-vision back-end loop optimization method.Based on the monocular loop detection,a multi-camera cross loop detection method is proposed.The main task of the back-end loop detection and optimization part is to eliminate the accumulated error of the front-end odometer through loop detection and global optimization,including loop detection optimization and relocation.Using the multi-camera multi-field information to alternately detect the loop similarity of multiple cameras during the map construction process can improve the loop detection rate.Experiments show that the multi-camera cross loop detection can effectively improve the loop detection success rate.3.This paper propose a method of combining lidar and multi-camera raster map construction.On the basis of high-precision multi-vision relocation and sparse mapping,the wheel odometer and lidar are combined to generate a high-precision occupancy grid map that satisfies the positioning and navigation of mobile service robots in an indoor environment.