Research On Dynamic Dispatching Optimization Of Urban Public Bicycles

With the continuous increase of the number of public bikes and the expanding scale of users,the difficulty in rental and return a bike becomes more and more serious.In order to reduce the occurrence of such situations,reasonable scheduling measures should be taken.Early static scheduling didn't take into account the dynamic changes in demand,resulting in the new requests cannot be processed in a timely manner.Therefore,for the dynamic bicycle scheduling problem,we study the model and optimization algorithm.Based on this research,the scheduling optimization based on short-term demand prediction is studied.The main research work of this paper is as follows:(1)The data set was processed and analyzed.Firstly,the factors affecting public bicycle travel were analyzed and relevant meteorological data were screened and processed.Then,dijkstra algorithm was adopted to calculate the shortest time required for the bicycle to move between any two stations,so as to measure the distance between two stations.Finally,k-means algorithm was adopted to cluster the stations,and the rationality of the number of clustering was demonstrated.(2)In order to improve the processing ability of random requests,we proposed a dynamic scheduling optimization scheme based on improved genetic algorithm in this paper.Firstly,the individuals in the population were selected through the tournament selection and elite selection strategy.Then the optimal savings method and the nearest neighbor priority method were adopted for the hybrid crossover operation.Finally,variable penalty factors were introduced to evaluate the population individuals.In the experimental part,compared with other methods,it was proved that the improved genetic algonthm has better optimization ability.And through experimental cases,it was proved that dynamic scheduling is more suitable for solving public bicycle scheduling problems than static scheduling.(3)In order to improve the timeliness of dynamic scheduling,we proposed a dynamic scheduling optimization based on short-term demand prediction.Firstly,the multi-similarity reasoning model based on random forest was adopted to predict the demand of each site.Then the scheduling path was dynamically optimized based on short-term demand prediction.Finally,in the experimental part,compared with other prediction methods,it was proved that the prediction algorithm has higher accuracy.The comparison between this scheduling optimization approach and other scheduling optimization schemes proved that this scheme has better optimization processing capacity and can effectively improve the response time of transport vehicles to station requests.