Research On Cooperative Scheduling Framework Of Multi-Period Dynamic Heterogeneous Shared Transportation Vehicles

Bike-sharing,electric bike-sharing,and electric car-sharing respectively satisfy people’s short,medium and short-medium distance travels.While bringing convenience to people’s travel,they also waste resources due to untimely and unreasonable scheduling.In this context,this article considers multi-period dynamic attributes,with the goal of minimizing driving cost and demand deviation,establishes a multi-period dynamic scheduling model for bike-sharing,electric bike-sharing,and electric car-sharing,and designs a coordinated scheduling of the system framework to realize the dynamic coordinated scheduling of them.The specific research content is as follows:(1)Designed the framework of the collaborative scheduling system for bike-sharing,electric bike-sharing,and electric car-sharing,and generated a collaborative dispatching plan through the area recognition module,the prediction module,and the dynamic dispatch module in conjunction with the real-time monitoring module;(2)Divide the demand area based on the Mean-Shift clustering algorithm.Based on the area division,a feature processing method based on the combination of sub-model and category feature probabilization is proposed,and the Good-Turing idea is applied to the unappeared normalization of categorical features,the prediction performance of the prediction model is improved;(3)A new multi-period dynamic scheduling model is proposed,and a spatio-temporal network diagram construction algorithm for multi-period scheduling is designed to achieve the balance of the system through the balance between the excess demand area and the insufficient demand;(4)A greedy heuristic algorithm was designed to solve the initial solution for the multiperiod scheduling model of bike-sharing,electric bike-sharing,and electric car-sharing,and multi-period Adaptive Large Neighborhood Search(ALNS)algorithm was designed to optimize the initial solution,so as to solve the scheduling model.The experimental results show that the feature engineering processing method based on submodel correction can effectively alleviate the large peak prediction error caused by data imbalance.The average absolute error in Lasso,Ridge,Elastic Net,SVR,BP neural network models(Mean Absolute Error,MAE)is reduced by 37.6%on average;the designed multi-period greedy heuristic combined with multi-period ALNS to solve the algorithm can effectively solve the problem of urban shared traffic vehicle dynamic scheduling;the designed collaborative dispatching system framework can generate effective collaborative dispatching paths when shared transportation vehicles are replaced,promoting energy conservation and emission reduction in the city,while improving the travel experience of urban residents.