Traffic State Prediction For Predictive Cruise Control

Commercial vehicles have characteristics of large cargo capacity,long mileage and high fuel consumption.When vehicle is driving on a composite road with slopes,unreasonable throttle,emergency braking or frequent shift behavior due to traffic flow or slopes will lead to a significant increase in fuel consumption.Predictive Cruise Control（PCC）is aimed at vehicles driving on composite roads with slopes.The slope information in front is obtained from the ADAS map,and the gear and engine speed and torque state are adjusted.However,traditional PCC lacks the processing of traffic information,and is often disturbed by traffic flow during the cruise process,which affects actual fuel saving effect of PCC.It is of great significance to predict road traffic conditions and apply it to PCC,which can improve the adaptability of traffic conditions and the fuel economy of vehicles.The research includes the following aspects.Firstly,PCC system is analyzed and modeled.The vehicle longitudinal dynamics analysis is carried out,and the dynamic programming solution is used for PCC algorithm based on distance domain.System state equation and vehicle kinetic energy equation are given,and optimization focus on engine energy optimization rather than fuel consumption optimization.In view of the fact that the traditional PCC focuses on the change of slope and lacks the global or local consideration of traffic condition,PCC algorithm integrating traffic conditions is studied.Taking road expected speed obtained by global traffic conditions prediction as PCC cruise reference speed,combined with local obstacle vehicle trajectory prediction to make PCC overtake or yield decision,can enhance the road traffic adaptability of the PCC system and improve the actual fuel saving effect in the complex traffic environment.Secondly,in order to obtain the road expected speed as PCC cruise reference speed,a traffic condition prediction framework based on the spatio-temporal positioning data of the Io V（Internet of Vehicles）is proposed to predict the road expected speed after rasterization.Road traffic prediction framework is proposed for the spatio-temporal positioning data of a single brand of commercial vehicles.The framework consists of Conv LSTM coding and Res Net residual structure,and introduces Attention mechanism and Mask module.Spatio-temporal positioning data of specific brand vehicles are used to predict the global traffic condition,and the expected speed of different sections of the road network in the future is obtained.Verified by actual data,the road expected speed predicted by the framework can be used as PCC cruise reference speed.Thirdly,local obstacle vehicle trajectory prediction algorithm considering vehicle lane change behavior is designed for PCC,which provides specific prediction parameter values of surrounding obstacle vehicles for PCC decision-making.In order to improve the computational efficiency,local vehicle trajectory prediction algorithm considering vehicle lane change behavior is divided into vehicle lane change intention recognition based on XGBoost and vehicle lane change trajectory prediction based on vectorization processing.Vehicle lane change intention recognition uses XGBoost machine learning model to determine‘cut in vehicle’,and predicts the trajectory of the‘cut in vehicle’to obtain accurate specific prediction parameter values.The specific prediction parameter values include obstacle vehicles cut-in time t_cut,cut-in distance s_cut and the cut-in speed v_cut.Different parameters are selected to evaluate the prediction accuracy and recall rate.The results show that the local obstacle vehicle trajectory prediction algorithm can match PCC requirements.Last,Truck Sim/Simulink co-simulation is used to study the application of global traffic condition prediction and local vehicle trajectory prediction in PCC system.In Truck Sim,the longitudinal vehicle dynamic model of FAW Jiefang JH6 is established,and typical road conditions such as uphill,downhill,down-uphill and up-downhill are simulated for PCC applied to global traffic condition prediction,and overtake or yield conditions are simulated for PCC applied to local obstacle vehicle trajectory prediction.Compared with traditional PCC,the simulation results show that the traffic condtion prediction can provide the traffic conditions information for PCC,which makes the PCC more suitable for the road traffic conditions,improves the fuel saving effect,and plays an obvious role in maintaining the cruise state,reducing the rapid acceleration and deceleration.In summary,this paper proposes global and local traffic condition prediction methods for PCC,and discusses the application of traffic state prediction and road scenarios in PCC.The actual data verification shows that the global and local traffic condition prediction methods in this paper can match PCC demand.PCC system combined with traffic condition prediction has 2%-5%fuel economy improvement without reducing the cruising speed,and significantly increases following distance in cruising state,reduces braking times,and improves traffic adaptability,fuel economy and driving safety of PCC system.