A crowdsourced dynamic repositioning strategy for public bike sharing systems

Abstract

<p style='text-indent:20px;'>Public bike sharing systems have become the most popular shared economy application in transportation. The convenience of this system depends on the availability of bikes and empty racks. One of the major challenges in operating a bike sharing system is the repositioning of bikes between rental sites to maintain sufficient bike inventory in each station at all times. Most systems hire trucks to conduct dynamic repositioning of bikes among rental sites. We have analyzed a commonly used repositioning scheme and have demonstrated its ineffectiveness. To realize a higher quality of service, we proposed a crowdsourced dynamic repositioning strategy: first, we analyzed the historical rental data via the random forest algorithm and identified important factors for demand forecasting. Second, considering 30-minute periods, we calculated the optimal bike inventory via integer programming for each rental site in each time period with a sufficient crowd for repositioning bikes. Then, we proposed a minimum cost network flow model in a time-space network for calculating the optimal voluntary rider flows for each period based on the current bike inventory, which is adjusted according to the forecasted demands. The results of computational experiments on real-world data demonstrate that our crowdsourced repositioning strategy may reduce unmet rental demands by more than 30% during rush hours compared to conventional trucks.</p>

A Mode of Intelligent Equipment Monitoring Optimization Based on Dynamic Programming Algorithm

“… Build an intelligent health monitoring data set, C l = { c 1 , c 2 ,…, c i } and eliminate the intelligent monitoring data with digitization degree of <60%. Determine the weight coefficient and adjustment error according to the digital Economic Yearbook of power industry from 2015 to 2020; The monitoring data c l processed by wavelet function are processed structurally [ 14 ], semistructurally and unstructured to realize the “cleaning” of intelligent monitoring data. Judge the fusion relationship of the monitoring data after “cleaning” and obtain the values of Pc and Pm of local and overall fusion relationships.…”

“…The monitoring data c l processed by wavelet function are processed structurally [ 14 ], semistructurally and unstructured to realize the “cleaning” of intelligent monitoring data.…”

Smart Recommendations for Renting Bikes in Bike-Sharing Systems

“…Wang and Hou [27] propose a repositioning strategy that assumes there is a number of voluntary riders at each time period. Trips (including bike repositioning by volunteers) are considered for time periods of 30 minutes.…”