Research On Dynamic Scheduling Of Public Traffic Vehicles Based On GPS Data

Transit vehicle scheduling is a major impact on the operating efficiency of public transportation system and service level,and is the core content of intelligent transportation system.The use of big data on the dynamic scheduling of public transport vehicles not only saves human resources,but also to meet the needs of passengers,and improving the attractiveness of public transport while ensuring the vehicle loading rate.In order to realize the Hangzhou public transportation intelligent vehicle scheduling,the paper establishes the bus arrival time prediction model,the passenger flow forecast model and the optimization of bus frequency model with the data of bus GPS positioning data and video monitoring statistical data.Bus arrival time prediction,which using Euclidean distance formula and the bus speed to calculation of bus arrival information,and then predicting the bus arrival time according to the historical data with the BP neural network algorithm.Prediction of bus station passenger flow,calculating the boarding rate with the change of standing time and distance and getting off rate with the change of standing time and cross section according to the number of passengers on the bus to get off the bus.And then predict the number of passengers for different lines and different sites and different time with the exponential smoothing method.Arranging the bus departure frequency reasonable to reduce the passengers' waiting time and improve bus utilization rate and increase the profitability of the company by establishing objective function,which is restrained by the conditions.Solve the problem with the improved genetic algorithm after forming a weighted objective function.The bus frequency of the goal is to reach the minimum of average waiting time of passengers,and the average passenger volume reached the maximum,but also standard deviation of the maximum cross-section to meet the minimum.The maximization goal should be converted to minimize goal,and then plus the goals with the weight to calculate the optimization target.This paper presents a modified genetic algorithm from several aspects,including elite co evolutionary,adaptive probability,simulated annealing algorithm and genetic encoding method,these improvements gradually applied to the bus schedule through MATLAB simulation,the results become better in varying degrees,the convergence gradually decreased while the optimal fitness value is also reduced.