Vehicle Driving State Prediction Based On Wireless Data Fusion

With the rapid development of urbanization and motorization,China is facing serious traffic problems,such as traffic jams,traffic environment deteriorates,and increased traffic accident rate.In order to effectively alleviate traffic pressure and reduce the accident rate,it is important to obtain the driving state of surrounding vehicles in real time and accurately.To this end,the short-term vehicle state prediction based on wireless sensing information is proposed as one of the promising solutions to traffic safety problems.For this,this thesis first investigates the short-term vehicle state prediction technologies,and then proposes effective predicton algorithms based on Kalman filtering theories.The main contributions are as follows:(1)For the speed prediction problem of high-speed moving vehicles,this thesis proposes a radio frequency identification(RFID)system to obtain the front vehicle information by exchanging the vehicle data between RFID tags and RFID readers.In this RFID environment,this thesis proposes a vehicle speed prediction algorithm based on adaptive Kalman filter(AKF).In order to enhance the influence of current data and reduce the influence of old data,the prediction algorithm introduces an adaptive forgetting factor based on the traditional Kalman filter(KF).Numerical results show that the proposed algorithm improves the speed prediction accuracy by 87.5% and 50% respectively compared with the least squares(LS)and the traditional Kalman filter algorithm.Therefore,the real-time and accuracy of the proposed algorithm are fully verified.(2)For the state prediction problem of moving vehicles at intersections,this thesis proposes a vehicular networking system to obtain vehicle information by exchanging traffic data between on-board unit(OBU)and roadside unit(RSU).In this vehicular networking system,this thesis proposes a two-level quantized adaptive Kalman filter(QAKF)algorithm based on auto-regressive moving average(ARMA)model to predict the vehicle state(i.e.,the moving direction,driving lane,vehicle speed,and acceleration).Numerical results show that compared with the ARMA model,the traditional Kalman filter(KF)algorithm,the regression tree(RT)algorithm,the unscented Kalman filter(UKF)algorithm and the long short term memory(LSTM)algorithm,the proposed algorithm improves the speed prediction by 90.62%,89.81%,88.91%,82.76% and 70.77%,respectively.