Research On Vehicle Dispatch Strategy Based On Time And Safety Joint Optimization

With the widespread use of smart phones,taking a taxi through mobile apps has become a popular mode of transportation.In recent years,as urban traffic congestion continues to intensify,and a number of illegal and criminal cases have occurred in the taxi（online ride-hailing）industry,how to reasonably dispatch vehicles in the city,reduce the occurrence of traffic congestion and ensure the safety of drivers and passengers in the service process has become more and more people’s concern.Taking this as the starting point,this paper proposes a vehicle dispatch strategy based on the joint optimization of time and safety,and makes a reasonable dispatch to passengers’ hailing requests from two aspects of time and safety.Firstly,this thesis starts from the data of the taxi trajectory of shenzhen taxi and the driver’s driving behavior alarm data,and puts forward the removal and correction of the abnormal data.The original trajectory data is converted by the coordinate transformation method,and the geometric matching and the direction of the trajectory processing method are proposed to match the taxi trajectory data,and the speed of each road is calculated.In addition,for the driver’s driving behavior alarm data,after screening and deduplication of different types of alarm data,9 representative driver alarm behaviors are selected,and the membership function method is used to convert the alarm times of each driving behavior into the corresponding score for standardization,so as to provide data support for the subsequent implementation of the corresponding algorithm.Secondly,based on the previously processed road speed data,a road speed prediction algorithm based on improved spatiotemporal graph convolution is proposed to predict the speed of the road network in selected regions,and compare the long-and short-term memory network model and the tree model XGBooost with it,and the improved method has higher prediction accuracy.Thirdly,based on the predicted speed,an improved Dijkstra algorithm is proposed to carry out path planning by building a road network model,so as to obtain the path with the shortest waiting time for the driver to reach the passenger pick-up point.At the same time,compared with the path obtained by the traditional Dijkstra algorithm,the waiting time of the path calculated by the proposed method is shorter.Finally,according to the driver’s driving behavior alarm data,a combination weighting method called CRITIC-AHP（Criteria Importance Though Intercrieria Correlation,CRITIC;The Analytic Hierarchy Process,AHP）is proposed to comprehensively score the drivers.At the same time,a utility function is constructed to jointly optimize the shortest time and safety score calculated above,and a reasonable order distribution result is obtained.