Research On Dynamic Ride-Sharing Matching And Vehicle Routing Optimization Of Online Taxi

With the growth of China’s economy and the increasing desire for a higher standard of living,there has been a surge in demand for travel,resulting in a continuous rise in motor vehicle ownership.This has led to worsening traffic congestion,increased energy consumption,and environmental pollution.The emergence of taxi ride-sharing offers a potential solution to optimize taxi dispatching systems,improve passenger service levels,and reduce waiting times and travel costs while enhancing vehicle utilization.This study presents the Dynamic Ride-sharing Matching Problem of Online Taxi(DRMP-OT),which is developed based on an extensive analysis of the strengths and weaknesses of current ride-sharing models on online taxi platforms and recent research on ride-sharing travel problems in both domestic and international contexts.The DRMP-OT model aims to minimize the sum of the waiting time cost and travel cost for all passengers in the ride-sharing system.The model incorporates a soft time window constraint,considering route,ride-sharing,time,and load constraints,and introduces a cost constraint to safeguard the fundamental interests of passengers and drivers.The model is then tested using real-life cases and the Gurobi optimization solver.To solve the large-scale dynamic ride-sharing problem,an optimization algorithm combining the multi-agent reinforcement learning QMIX algorithm and the adaptive large-neighborhood search algorithm,and the corresponding dynamic simulation framework are designed in this study.Lastly,two kinds of case studies,small and large-scale,are performed to evaluate the proposed algorithm based on the real taxi order data of a certain district in a certain city.The test results indicate that the proposed algorithm outperforms the actual situation in terms of average travel cost savings of 29.84%in the large-scale case test.Moreover,the algorithm achieves a 22.40% reduction in vehicle use,although the minimized number of vehicles is not considered.The study also demonstrates significant benefits of ride-sharing over non-ride-sharing systems,with an average reduction of 20.05% in waiting time costs,15.87% in travel costs,and 18.96% in the number of vehicles used.The case test also analyzes the impact of rolling time domain cycle length,cluster size,and the discount rate of the ride-sharing fare on the algorithm’s solution effectiveness and passenger cost.The proposed mathematical model and algorithm can provide theoretical support for optimizing the passenger-vehicle matching and vehicle path optimization of the dynamic ridesharing system for online taxis and provide more travel options for the general public.