Empty Vehicle Relocation Method For Autonomous Taxi Services

With the rapid development of the Internet of Things and autonomous driving technology,autonomous taxis have gradually appeared all around the world.Compared with the traditional manual driving taxi,the autonomous taxi system is not constrained by the personal preference of the driver.It automatically collects and processes traffic information,intelligently matches passengers and taxi,and optimizes the driving route globally,avoiding the problem of high empty driving rates,choosing passengers,and cheating passengers.After completing the order-taking task,the autonomous taxi will become the empty state.How to relocate the vehicles in the empty state is the key problem to complete the global optimization of the driverpassengers matching.Aiming at the problem of the autonomous driving system’s relocation algorithm for taxis in empty state,this paper designs an autonomous driving taxi relocation algorithm based on demand forecast.it is mainly divided into two stages.Firstly,this paper converts a large amount of vehicle trajectory data into grid OD matrix data,and uses the Bi Lstm network to establish a demand forecasting model.This forecasting model can use historical demand data to forecast demand data at the next moment.Secondly,in terms of the stage of empty relocation.In this paper,according to the OD information predicted by the demand prediction model,a flow network model is established for different regions of the grid for relocation between regions,and the minimum cost maximum flow algorithm is used to ensure the minimum cost of completing regional relocation.After that,this paper discusses the traditional empty relocation algorithm,and proposes an improved empty relocation algorithm based on the demand prediction model and the regional relocation model.Finally,we considered the actual application scenarios,simulated the experiment based on real map data,and conducted experiments on the traditional relocation algorithm and relocation algorithm based on demand prediction in the system.By comparing with traditional relocation algorithms,the empty relocation algorithm proposed in this paper can reduce the waiting time of 45% of passengers and increase the total length of taxi passengers by 54%,which has better performance than traditional empty relocation algorithms.