| Vehicle speed prediction as an important part of intelligent vehicle, can provide the future traveling data for the decision system of the vehicle. It has great significance for intelligent vehicle, safety assistant driving and powertrain control system research. Since vehicle speed is affected by many factors, high time-varying and nonlinear features make it difficult to predict. In this paper, we study for the speed prediction of the vehicle itself not the traffic flow, analyzing the speed data time series features, using NAR neural network advantages of dealing with the nonlinear and non-stationary time series to establish the prediction model to forecast the speed. The prediction model is used in anti-collision warning system to do simulation analysis to verify the validity of the algorithm.Firstly, the car OBD II equipment is used to collect the vehicle speed data, and the monocular vision camera is used to collect the distance data from the front car. Kalman filter was used to optimize the collected data in order to provide training data of the neural net. Then, a NAR network prediction model which based on the vehicle speed autoregression was established. And the net was trained by back-propagation algorithm in the form of series model. In this paper, we also establish another prediction model using HMM model by the way of pattern recognition. The simulation results show NAR neural network is better than the HMM model to predict vehicle speed. All the results verify NAR network has good prediction accuracy, time-varying and longtime performance. Finally, the NAR neural network prediction algorithm is used in the anti-collision warning systems, the system is established through prediction model which combines car critical safe distance model. The speed and distance predicted by the NAR neural network are used to calculate the critical distance in order to advance the warning time. All the experimental results demonstrate the effectiveness of the prediction algorithm. |
PDF Full Text Request