main-content

2020 | Buch

# Active Balancing of Bike Sharing Systems

verfasst von: Prof. Jan Brinkmann

Buchreihe : Lecture Notes in Mobility

Enthalten in:

insite
SUCHEN

### Über dieses Buch

This book reports on an operational management approach to improving bike-sharing systems by compensating for fluctuating demand patterns. The aim is to redistribute bikes within the system, allowing it to be “actively” balanced. The book describes a mathematical model, as well as data-driven and simulation-based approaches. Further, it shows how these elements can be combined in a decision-making support system for service providers. In closing, the book uses real-world data to evaluate the method developed and demonstrates that it can successfully anticipate changes in demand, thus supporting efficient scheduling of transport vehicles to manually relocate bikes between stations.

### Inhaltsverzeichnis

##### Chapter 1. Introduction
Abstract
Increasing urbanization and mobility demand lead to a large volume of traffic in urban areas. As the traffic infrastructure is limited, too much traffic results in traffic jams, noise, and air pollution. City administrations focus on a reduction of the individual traffic to tackle these discomforts. Therefore, collective traffic modes, i.e., public transport systems (PTSs), are launched or expanded. Conventional modes are buses and trams. On the one hand side, PTSs may be able to reduce urban traffic. But on the other hand side, the comfort is reduced. First, the users’ actual origins and destinations are not necessarily in walking distance to a bus or tram station. Second, buses and trams are often delayed or crowded.
Jan Brinkmann

#### Preliminaries

##### Chapter 2. Bike Sharing Systems
Abstract
In this chapter, we introduce the concept of BSSs and point out the resulting challenges.
Jan Brinkmann
##### Chapter 3. Optimization Problems
Abstract
In this chapter, we highlight the literature related to the balancing of BSSs by means of manual relocationing. In Sect. 3.1, we introduce different types of vehicle routing problems addressing various applications. Section 3.2 focuses on inventory routing problems for BSSs and provides a comprehensive literature classification.
Jan Brinkmann
##### Chapter 4. Dynamic Decision Making
Abstract
In this chapter, we define the basic concepts on dynamic decision making. We first introduce the concept of a dynamic decision process (DDP). In Sect. 4.1, we define the Markov decision process as a special case of DDPs and as the modeling technique. In Sect. 4.2, we introduce approximate dynamic programming as a collection of policies to solve MDPs.
Jan Brinkmann

#### Application

##### Chapter 5. The Stochastic-Dynamic Multi-Vehicle Inventory Routing Problem for Bike Sharing Systems
Abstract
In this chapter, we introduce the multi-vehicle stochastic-dynamic inventory routing problem for bike sharing systems ($$\text {IRP}_\text {BSS}$$) as first presented by Brinkmann et al. (2019b).
Jan Brinkmann
Abstract
In this chapter, we introduce the lookahead policies originally presented by Brinkmann et al. (2019a, b) to solve the $$\text {IRP}_\text {BSS}$$. The LAs base on the blue print introduced in Sect. 4.​2.​2 but are strongly adapted to the requirements of the $$\text {IRP}_\text {BSS}$$.
Jan Brinkmann
##### Chapter 7. Dynamic Lookahead Horizons
Abstract
As observed by Ghiani et al., Transp Res Part E: Logist Transp Rev 45(1):96–106, 2009, Voccia et al., Transp Sci, 2017, the length of a lookahead horizon has a significant impact on the solution quality when solving a stochastic-dynamic optimization problem. Therefore, the lookahead horizon must neither be too short nor too long. More precisely, the lookahead horizon needs to match the request pattern. In BSSs, we are facing spatio-temporal request patterns (Sect. 2.​4). Therefore, in this chapter, we approach lookahead horizons for LAs changing in the course of the day. We define a dynamic LA (DLA) to be an LA with lookahead horizons individually selected for certain periods of the time horizon. To this end, we use a procedure originally introduced by Brinkmann et al., Comput Oper Res 106:260–279, 2019a.
Jan Brinkmann
##### Chapter 8. Case Studies
Abstract
In this chapter, we present the case studies to evaluate the policies introduced in the previous chapter. If not stated otherwise, the computations are carried out on an Intel Core i5-3470 with 3.2 GHz and 32 GB RAM. Our approaches are implemented in Java 8.0u121. In Sect. 8.1, we describe the real-world data sets we use. Section 8.2 addresses the actual generation of instances. We describe the implementation of transitions, i.e., the handling of relocations and requests in Sect. 8.3. Benchmark policies are defined in Sect. 8.4. In Sect. 8.5, we present the parametrization of our policies as well as the benchmark policies. The results are presented in Sect. 8.6. In Sect. 8.7, the solutions’ structures are analyzed.
Jan Brinkmann

#### Conclusion

##### Chapter 9. Managerial Implications
Abstract
In this chapter, we summarize the findings of the previous chapters to accentuate managerial implications.
Jan Brinkmann
##### Chapter 10. Future Research
Abstract
In this chapter, we point out future research. In Sect. 10.1, we suggest possible extensions of the model to account for challenges in practice. Possible extensions of the method to obtain a higher solution quality are suggested in Sect. 10.2.
Jan Brinkmann