Design Of Vehicle Odometer System Based On Multi-sensor Fusion

With the vigorous development of artificial intelligence technology,autonomous driving technology has gradually entered people’s lives.Vehicle localization is the basis of the whole automatic driving ecology,and provides real-time location and posture information of vehicles.High precision localization has become one of the key issues in the autonomous driving system.In this paper,a vehicle odometer system based on multi-sensor fusion was proposed to meet the high precision and real-time requirements of positioning system,which solved the problem that single sensor positioning scheme cannot meet the high precision and high stability.In this paper,error state Kalman filter(ESKF)was used as the filtering framework to estimate the position,speed and attitude information of autonomous vehicles,and a multisensor fusion vehicle odometer system was proposed.The inertial measurement unit(IMU)was used as the data of state prediction.Status update includes sensors such as wheel speed encoders and GPS.According to the sensor model and measurement principle in the positioning system,the sensor error model was obtained,which become the prediction equation and update equation of computing system to provide mathematical basis.According to the principle of filter,an eskf-based multi-sensor fusion vehicle odometer system was designed.And the state variables,prediction process and update process of the system are clearly defined.A spatiotemporal synchronization scheme and outlier preprocessing scheme are used to ensure the effectiveness of sensor data.The initialization scheme of the system was improved by using the complementarity of multi-sensor.Based on the vehicle kinematics model,the static and dynamic zero velocity update are proposed to improve the accuracy of odometer.This paper tested the odometer system on multiple data sequences of KAIST urban dataset.The improvement schemes and different observation schemes are analyzed and verified.The proposed algorithm was compared with the wheel odometer and the classical msckf-based odometer system.The results show that the positioning accuracy of the odometer system proposed in this paper was better than the comparison algorithm.