Navigation Of Driverless Vehicle Based On GNSS/Lidar/INS

The self-driving car is one of the main research fields.The navigation system is an important part of self-driving car.The global navigation satellite system(GNSS)has the advantages of no accumulated error in positioning,all-weather positioning,etc.However,GNSS has problems such as low operating frequency,multipath effect and signal loss of lock.Inertial Navigation System(INS)has the advantages of fast data update,high short-term accuracy and good stability.But it also has the shortcomings of large cumulative error and long-term accuracy poorly.Lidar has the advantages of high resolution,strong anti-interference ability,but is greatly affected by the weather and the environment.Using their respective advantages and disadvantages,the study of integrated navigation and positioning becomes one of the main directions.In this paper,we mainly study the positioning methods of GNSS?INS and Lidar combination,and use Kalman filtering and information fusion to process the information given to achieve continuous and accurate vehicle location.Finally,the map matching technology is used to meet the requirements of high precision positioning of autonomous vehicles.This paper mainly completed the following work:1?locator-related algorithms.(1)The self-driving frame and the principle of vehicle positioning are introduced.The transformation matrices between inertial coordinate system,carrier coordinate system,geocentric solid coordinate system and station coordinate system are analyzed.(2)The attitude,velocity and position algorithm of inertial navigation system are studied.(3)GNSS satellite navigation algorithm and its position estimation algorithm.(4)The calculation of 3D point cloud coordinates generated by Lidar localization,the principle of ranging and the construction of Lidar/INS geometric model are studied.Finally,the locator uses the fusion algorithm to solve the attitude of the vehicle at moment K,and transmits the data to the map end and displays it.2?Kalman filter algorithm improved,reducing the space complexity and time complexity of the algorithm.(1)In terms of time complexity,this paper presents a method to reduce the computational complexity of the covariance matrix prediction equation by using sparse state transition matrix.With matrix blocking and symmetry,the number of multiplications is reduced to less than 5% of the usual algorithms.Experiments show that the CPU of the improved method consumes less than 10% of the usual algorithm.(2)In terms of spatial complexity,this paper proposes a matrix transposition algorithm of block location pointers,which reduces the storage space of transposed matrix to 51%.3?Design a map-based environment model for experimental research.Aiming at the problem of spatial mismatch in map,an intelligent algorithm is proposed to accurately identify the vehicle's trajectory by taking into account the historical trajectory of the vehicle and the topology information of the road network.It determines the location of the vehicle on the road by all information sources related to the positioning system and the digital map database.In this paper,a simulation experiment platform is set up and the feasibility of Kalman filter and integrated navigation algorithm are simulated.At the same time,experiments of straight line and curve are carried out on the automatic driving agricultural machine,which verifies the accuracy and validity of the dead reckoning method.