Research On Interval Location Estimation Of On-travel Vehicles On Expressway Based On Multi-source Heterogeneous Data

The expressway in China is an important channel to connect different regions,and it undertakes a large number of vehicle trips with a small proportion of roads.The increasing social demand makes the expressway support increasing pressure,which also leads to frequent traffic congestion,vehicle collisions,and other situations.In the application of intelligent networking,determining the basic location of vehicles on the road is one of the keys for management departments to make intelligent supervision of vehicles,road drainage,and control decision-making,and is an important way to obtain prior information of vehicles in real time.In expressway scenarios,obtaining vehicle prior information can manage vehicles effectively,which is of great significance to reduce or avoid various conditions.This paper proposes an interval location estimation scheme of on-travel vehicles on expressway based on multi-source heterogeneous data.The main work is as follows:Firstly,based on the research,policies,and development of intelligent network at home and abroad,the research progress of multi-source heterogeneous data is reviewed;Subsequently,the research progress of expressway traffic flow prediction at home and abroad is described;Finally,it focuses on the current research status of vehicle location estimation in transit,and expounds the application of vehicle location estimation in expressway scenarios and extended scenarios.Secondly,the multi-source heterogeneous data used in this paper is analyzed and processed.In this paper,multi-source heterogeneous data,including ETC transaction data,vehicle track data of "two passengers,one danger" and other data,are taken as the research object.According to the characteristics of the data,a preprocessing method for expressway vehicle data is proposed.Based on the data field information,the multi-source heterogeneous bidirectional matching data of expressway is completed.Then the matching data is supplemented,converted,and expanded to obtain more data information and characteristics.Then,based on ETC transaction data,a traffic flow prediction scheme for expressway sections based on Kalman Filter and Random Forest algorithm are proposed.Firstly,the original traffic flow data is processed by Kalman Filter model to obtain consistent data;Next,explore the spatio-temporal evolution of traffic flow,and construct the temporal correlation characteristics,spatial correlation characteristic and other characteristics;Finally,The Random Forest model is used for prediction.From the results,the RMSE and MAE are 4.86 and 3.26,respectively.Compared with other baseline models and experiments with different characteristic combinations,this method has higher accuracy,which lays the foundation for subsequent research on interval location estimation of on-travel vehicles.Finally,based on the multi-source heterogeneous bidirectional matching data to conduct the data characteristic modeling for road sections,and hierarchical clustering model is used for training to divide road section into smaller intervals;Next,characteristic modeling is performed again based on the previously predicted section traffic flow and bidirectional matching data.Finally,the XGBoost algorithm is used to estimate the vehicle positions in the interval separately.From the results,compared with other baseline models,the interval location estimation of on-travel vehicles based on XGBoost has smaller error and higher accuracy,which also reversely verifies the rationality of hierarchical clustering model for road modeling.