基于深度学习的智慧共享单车立体车库的研究与实现

The Research and Implementation of Smart Stereo Garage for Shared Bicycle Based on Deep Learning Technology

近年来,很多人将共享单车作为短途代步工具,这些单车会阶段性地聚集在人流密集的交通枢纽,共享单车立体车库可将这些车辆收容,减少拥堵,但现阶段的共享单车立体车库车辆出入库速度较慢,调度不当的话会造成用户等待时间过长。现有的调度算法主要是为小型车立体车库设计,不适合共享单车立体车库。共享单车立体车库的调度算法需要能够识别车辆类型,预测下一时段不同类型车辆的出入库情况,在空闲阶段调整车辆存放位置,保证车辆能够快速入库、快速出库。本文基于深度学习技术实现共享单车识别模型、共享单车出入库预测模型,并基于该预测模型设计结果,结合调度论体系实现一种共享单车立体车库调度策略。通过实验验证,本文提出的调度算法有效地避免了调度不当问题,减少了用户等待时间。

In recent years,many people use shared bicycles as a short-distance transportation tool,and these bicycles will be gathered in traffic hubs with dense traffics periodically.Stereo Garage for Shared Bicycle (SGSB)can accommodate these bicycles and reduce congestion.However,the speed of entering and leaving the SGSB is slow,therefore,it will be time-consuming of improper dispatch.Basically,the existing scheduling algorithms are mainly designed for Stereo Garage for Car,which is not suitable for SGSB. The scheduling algorithm of SGSB needs to identify the type of bicycles,predict the number of in-out bicycles at the next period, adjust the storage location of bicycles at the idle stage,and ensure that bicycles can enter and leave the garage quickly.Based on deep learning,this paper implements a recognition model and a loading and unloading prediction model of shared bicycles.Besides,a scheduling strategy of SGSB based on the prediction model and the dispatching theory system is also proposed.Experimental results show that the proposed scheduling algorithm avoids the problem of improper scheduling effectively and reduces the waiting time of users.