Research On The Vehicle Localization Algorithm Based On Multi-Sensor Fusion

High-precision positioning can provide high-precision and high-reliability pose information for the environmental perception and planning control of autonomous driving,which is an important basis for the realization of autonomous driving.Due to the complexity of the car driving environment,it is difficult for a single sensor to meet the positioning requirements in the automatic driving of the car,and the multi-sensor fusion positioning algorithm emerges as the times require.The vehicle fusion localization method based on Kalman filter has a relatively mature research foundation,while existing research shows that graph optimization methods can use more historical observations to obtain higher-precision positioning results.Aiming at the positioning problem in automatic driving,this paper proposes a multi-sensor fusion positioning algorithm based on factor graph optimization,which aims to improve the positioning ability in all scenarios of automatic driving,especially the positioning retention ability in special scenarios such as tunnels and large curvature curves.The main work of the paper is as follows:(1)Based on the theoretical framework of factor graph optimization,a factor graph optimization fusion positioning algorithm based on GNSS,IMU,wheel speed sensor was designed,and a series of actual sports car tests were carried out on the algorithm’s positioning ability.And the feasibility and reliability of the algorithm are quantitatively verified through experiments.Compared with the Kalman filter method,there is a 10.9% improvement in the positioning accuracy,and in tunnels there is a 63.7% improvement in the positioning reliability;(2)A visual aided factor graph optimization fusion positioning algorithm is proposed,which uses lane line matching positioning to provide additional absolute position information for the car.Experiments show that the visual aid information can significantly improve the positioning ability of the positioning algorithm in the case of large GNSS positioning errors or loss of lock and the positioning accuracy under tunnel sections is improved by 82.2%,and the positioning reliability is improved by 94.3%.(3)Based on the actual car driving test data,a comparison test with and without wheel speed sensor fusion positioning method and a comparison test with zero speed constraint are designed.Experiments show that the wheel speed sensor can suppress the error divergence of the IMU and significantly improve the positioning accuracy and reliability of the fusion positioning algorithm;the zero-speed constraint can improve the positioning accuracy by 26.9% and the positioning reliability by 12.5% for the fusion positioning algorithm in the section of the toll station.To sum up,the factor graph optimization fusion positioning algorithm based on GNSS/IMU/wheel speed sensor/vision camera proposed in this paper can effectively improve the positioning accuracy and reliability of the whole scene of car driving compared with the traditional Kalman filtering method.,which provides a solution for autonomous vehicles to overcome the problem of large positioning errors in complex scenarios such as tunnels and curves with large curvature.At the same time,it has the characteristics of low cost of sensor hardware and strong scalability.