Ride-Sourcing Multi-Agent Based Modeling And Simulation System Development And Applications

Traffic travel is an important part of people’s life.With the continuous improvement of urbanization,urban traffic problems are becoming more and more serious,and urban traffic management has attracted more and more attention.The Internet era has spawned many travel modes,such as online ride-hailing and car-sharing,which have brought convenience to personalized travel,but also brought new problems to traffic management.How to regulate the traffic supply and demand relationship of large-scale urban road network has become a key scientific problem.As an important traffic analysis tool,traffic simulation has the advantages of fine granularity,high efficiency and low cost,which can help decision makers to simulate and test traffic strategies and provide technical support for decision optimization.In this paper,a multi-agent traffic simulation model is proposed for the ride-sourcing big data environment,and a complete simulation system is independently developed,which is applied to simulation and strategy optimization of various complex ride-sourcing scenarios.The main content and research results of this article are as follows:(1)Multi-agent based ride-sourcing simulation framework.An agent-based simulation framework is proposed,and the driver and passenger behavior models are constructed.Based on massive traffic data and deep learning,we build a data-driven agent behavior decision model to accurately simulate heterogeneous driver behaviors.Based on the open source map data,we build the simulation road network and realize multi simulation process,such as vehicle navigation,cruise,and human-vehicle matching.(2)An agent simulation system is developed independently based on open source Golang,which realizes functions such as back-end simulation and front-end visual display,and performs simulation results statistics and visualization demonstrations from multiple perspectives such as passengers,drivers,platforms,and society.According to experiment,it can simulate 130,000 trips in Hangzhou in a single day on a single machine(CPU i7-4790@3.6 GHz)in only 14 minutes,where simulation speedup ratio reaches 103 times.(3)Research on pricing of ride-sourcing platform based on multi-agent simulation system.Based on the multi-agent simulation system,a model is built to describe the relationship between platform pricing and market supply and demand,so as to study platform pricing.The results show that there is a specific travel price that maximizes the platform’s revenue,while the platform tends to increase the order price and reduce the driver’s income.The social utility maximization pricing scheme is maintained within a strip zone.(4)Human-driving Vehicles(HV)and Automated Vehicles(AV)ride-sourcing hybrid simulation.We expands ride-sourcing to HV and AV hybrid scenario.The simulation results show that AV and HV can complement each other.A small amount of AV(about 10%)can greatly reduce the waiting time of passengers.As the proportion of AV increases,the total mileage and exhaust emissions of online car-hailing are also continuously reduced.The exhaust emissions in the pure AV scenario are reduced by 12.3%,compared to the HV scenario.In summary,this paper proposes an agent-based ride-sourcing simulation model framework and develops a complete simulation system,which is applied to multiple simulation scenarios such as pricing optimization and AV and HV hybrid scenario.The system has high scalability and adaptability.It can be used as an important tool for ride-sourcing research.