Research On Flex-route Transit Scheduling Based On Multi-agent Simulation

Since the State Council issued “The Guidelines on the Priority for the Development of Public Transport in Cities”,the construction of public transport has entered a new period.The bus sharing rate increases,because the construction of subway,light rail,BRT and bus-only lanes has effectively improved the average running speed of public transport.However,the service area is limited and can only meet the needs of passengers in the traffic corridor.Flex-route transit(FRT)refers to the vehicle runs along the benchmark route and stops at a fixed station until receiving the passenger’s request.If the request matches the condition,the vehicle will deviate from the benchmark route to pick up the passenger.FRT has appeared in low-density travel areas in Europe and the United States for a long time,and it has been verified by both theory and practice that FRT can provide better service for passengers and more operating income for bus companies in low density areas.In China,there are many public transport operations similar to the form of FRT,such as customized bus.To complete a more realistic FRT scheduling research to prove the feasibility of FRT running in China,the concept of Agent is introduced to give each FRT bus,each passenger and FRT dispatching center thinking ability,so that it is possible to simulate FRT being scheduled under the actual roads and actual demand.Firstly,multi-agent theory,the critical elements of the FRT and the correlation between the FRT scheduling and the multi-agent simulation is introduced in detail.Based on the correlation,key agents of the FRT and they are communications are designed,the agents including passenger agent,FRT agent and dispatching center agent.To complete the dispatching of FRT,passenger agent and FRT agent send messages to the dispatching center agent,and then the dispatching center agent generates solution.Then,this paper studys the models in passenger agent and dispatching center agent,including the core model of passenger agent – traffic mode choice behavior model,the premise of dispatching center agent – static scheduling planning,the core model of dispatching center agent – dynamic scheduling model.In order to design the traffic mode choice behavior model in passenger agent,a multinomial logit model is constructed for the traffic mode choice behavior,and a questionnaire is designed to caculate the parameter calibration of the model.The traffic mode choice behavior model in the passenger agent can calculate the probability of each transit with the information acquired by passenger agent.In order to generate the initial schedule for dispatching centre agent to find out FRT’s travel baseline,fixed stops and departure time,an improved k-means algorithm is proposed to determine the fixed station of FRT,and the FRT departure schedule is caculated with the method of conventional public transit.In order to complete the construction of dynamic scheduling model and realize the realtime route adjustment of the dispatching centre agent under the feedback of the passenger agent,this paper proposes the pre-scheduling model and real-time scheduling strategy according to whether passengers are booked or not.In the pre-scheduling part,a pre-scheduling model based on mean-variance theory is constructed and a two-stage algorithm is proposed.The first stage is mainly to improve the operational efficiency of the second stage,and the second stage is the classical genetic algorithm.In the real-time scheduling part,a real-time scheduling strategy is proposed to change the vehicle trajectory in real time according to the passenger travel demand.Then,after building agents and setting correspond between agents,this paper integrates the passenger agent,the FRT agent and the dispatching center agent and realizes the communication between the same and different classes of agent with JADE platform and baidu map API platform.The database is established to import or record operating data and used visual window to show scheduling results.Finally,an example of the operation and scheduling of the FRT is carried out in the simulation platform with the potential passenger demand in Chengdu.The scheduling results are analyzed.The analysis showes that FRT has strong competitiveness in the actual roads environment in suburban areas in China with actual demand.