Research On Network Performance And Scheduling Optimization Of Shared Bicycle Based On NB-IoT

In the development of shared bicycles,the demand forecast and dispatching of bicycles have always been the pain points of bicycle development and are also the focus of research in academia.Besides,the high power consumption of communication in shared bicycle has become a prominent problem.In order to study and analyze the above problems and propose the solutions,the following works are done in this paper:Firstly,this paper preprocesses the obtained bicycle trip data,including eliminating error data and the data irrelevant to this experiment.In order to simplify subsequent operations,the convertion of geohashed geo-location and the normalization of trip data are operated in paper.This paper also makes a brief analysis of the trip distance,and the results show that 90% users prefer a trip distance less than 2000 meters,which further indicates that shared bicycle is more of a short-distance transportation in urban life.Secondly,this paper handles all the pre-processed bicycle trip data and divides the whole region into multiple regional clusters,and then selects the regional cluster where the row points are measured for further study.Considering that the prediction of bicycle demand is not only related to a single isolated point,but also involves a bicycle travel point near that point,this paper selects the trip records of some areas near the treatment measurement point to represent the demand at the point to be predicted and names them as bicycle travel points.In this paper,ARMA model,a fast regression algorithm,is selected to predict the demand of cycling points.Thirdly,this paper proposes a hybrid scheduling strategy based on the predicted data.First of all,this paper simplifies the bicycle trip scene in reality,establishes an undirected graph of the trip region by using mathematical methods,and then selects the trip record resampling data of multiple travel points in the previous step as the experimental trip data.In addition,the Monte Carlo simulation method is used to generate several initial distribution maps of the number of bicycles with the characteristics of bicycle distribution for the verification of the model.In this paper,a manual dispatching algorithm and a user incentive algorithm based on the predicted value of a bicycle trip point are proposed.Finally,we use a number of initial maps with different distributions and travel records with different flows to test the dispatching algorithm.