Simultaneous Localization And Mapping Of Intelligent Car Based On Multi-Sensor Fusion

With the rapid development of the national economy,the sales volume of automobiles keeps rising.According to the information released by the National Bureau of statistics,in 2019,there are 261.5 million civil automobiles in China.People’s requirements for the safety and intelligence of vehicles are constantly improving.The revolutionary development of the Internet promotes the development of automobile industry towards the direction of intelligence and networking,and the intelligent technology of automobiles is gradually being applied.The intelligent vehicle is equipped with on-board computer,intelligent software and various sensors,including laser radar,visual sensor,millimeter wave radar and ultrasonic radar,so that the vehicle can safely and efficiently reach the destination without human intervention,relying on the vehicle itself for obstacle avoidance,navigation and positioning.Autonomous navigation,as one of the most important technologies of intelligent vehicles,reflects the intelligent degree of intelligent vehicles to a certain extent.The key of autonomous navigation is the simultaneous localization and mapping of intelligent vehicles.First of all,this thesis foucuses on SLAM methods,mainly including SLAM lidar based SLAM And visual based SLAM.Lidar has high accuracy and good real-time performance,but when the environmental characteristics are not obvious and in a large environment,the mapping effect is not ideal due to cumulative errors or erroneous closed-loop detection.Although the visual sensor has rich environmental image information,its accuracy is relatively lower than that of lidar,and it is greatly affected by light.The rich environmental information causes a large calculation load and it is difficult to ensure the real-time performance of the algorithm.In general,lidar and vision sensors have their own advantages and disadvantages.A single sensor is difficult to adapt to the complex and changeable environment.The integration of visual sensor and lidar will be the development trend o f intelligent vehicles.Secondly,this thesis describes the basic principles and framework of SLAM,classifies SLAM from the principles and working methods,and introduces the current mainstream environment map building methods,and then introduces the SLAM algorithm on which this article is based.The system model of the smart car is introduced,including the coordinates of the smart car,the odometer model,and the model of the lidar.The three coordinate systems are introduced into the car body coordinate system,and the basics of the laser-based graph optimization SLAM algorithm are introduced Framework and principles.Finally,based on the existing research,this thesis proposes a new algorithm that combines the two methods of lidar and visual sensor.Its main idea is: in the process of smart car movement,the visual sensor is used for closed-loop detection.Using visual closed-loop detection signals and lidar-based graph optimization SLAM fusion,complete the synchronous positioning and mapping of smart cars.The algorithm can effectively deal with the SLAM problem of environmental conditions with similar environmental structures and weak GPS signals.The robot operating system is used to collect data through smart cars and transplant the data to the laser-based graph optimization SLAM and multi-sensor fusion SLAM algorithm In comparison,the experimental results show that compared with the SLAM algorithm based on Lidar-based graph optimization,the multi-sensor fusion algorithm improves the accuracy of mapping while ensuring real-time performance.