Study On Intelligent Scheduling Optimization Algorithm Of Urban Public Bicycle System

With the increasingly serious urban road congestion and continued climate deterioration, bicycles, the healthy and zero pollution transportation means, are gradually returning to the sight of citizens due to the help and support of the government. However, as the result of the unbalanced traffic flow and commuter travelling peak, an awkward phenomenon occurs frequently that people cannot rent bikes at empty stations and return bikes at full stations, which blows down the enthusiasm of citizens to choose bikes as their travelling means, dramatically hindering the development of Public Bicycle System(PBS).Based on the optimal computation theory and Vehicle Routing Problems(VRP), the paper researches into the intelligent scheduling alogrithms of PBS, including multiple vehicle task coordination, static route planning and dynamic scheduling optimization strategy. Dynamic regional scheduling model based on clustering partition, route planning strategy based on genetic algorithm(GA) and dynamic scheduling scheme based on iterative feedback model are proposed, which optimize the scheduling costs and station service capability, providing schedulers and managers with decision supports and intelligent scheduling solutions.Firstly, on the basis of dynamic regional scheduling model, the paper proposes a K-medoids based multistage reoptimization clustering algorithm with composite considerations of station distances, vehicle tasks and station demands to accomplish the dynamic region partition and solve the station allocation problem for multiple vehicles.Secondly, the mathematical model with minimal costs is established and a GA based route planning strategy is proposed for static scheduling problem. Inverse number based chromosome distance and its minimum error correction analysis method are proposed to transform infeasible individuals of GA. Bidirectional correlation based crossover operator, adaptive mutation operator, and 2-opt local search strategy are employed to promote the solution quality and reduce computation time.Thirdly, the paper establishes the mathematical model with maximal station service capability and minimal scheduling costs for dynamic scheduling problem and then solves it with an iterative feedback scheme. Three components are incorporated into the scheme: a GM(1,1) based demand prediction model, a scheduling planning model that is implemented in terms of “demand-distance ratio” and Pareto preference ordering method, and a tabu strategy for starving and pathological station inhibition.Finally, algorithms of the paper are implemented with modular design method. Simulation tests performed on the dataset of Yi Xing PBS and TSP/VRP standard instance library demonstrate the effectiveness of the scheduling strategies proposed in the paper. Moreover, a B/S based simulation system is developed, which provides users with a convenient interface to invoke the algorithms and check the results.