Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Value function approach to managing station-free shared mobility systems



Joe Warrington

Shared mobility systems allow a customer to use a vehicle (typically a bicycle or car) to complete a short journey or task. Many cities now have bike- and car-sharing schemes in which dedicated stations or parking spaces serve as the start and end points of such journeys.

A recent trend of forgoing stations in favour of "free floating" vehicles has emerged, motivated by lower setup costs and a reduced need to coordinate with local authorities. In such schemes, the user typically locates a nearby vehicle using a public app, and at the end of a journey is free to park it in an arbitrary location within the city.

In general, supply and demand will not be evenly distributed, leading to an uneven distribution of shared vehicles around the city, and therefore low service quality for users. Operators of such "dockless" schemes must therefore invest effort in rebalancing the system, principally by attempting to manually relocate them at minimum effort.

This project is concerned with formulating, simulating, and solving this emergent problem, building on our previous work on "conventional" shared mobility systems. We aim to carry out the following steps:

  1. Review emerging literature on the subject, and define the mathematical optimization problem to be solved. This includes deciding on what decisions the operator can make (e.g. manually transferring one bike at a time, or loading several into a truck), and encoding these in the formulation.
  2. Introduce a value function, that is, a function which represents the cost of leaving the system in a particular state at the end of a limited planning horizon;
  3. Use methods from the dynamic programming literature, as well as dual dynamic programming (a related approach used for scheduling of dynamical systems) to define how this value function should be approximated
  4. Devise a suitable algorithm for the operator's decisions (e.g. routing algorithms for trucks) on a short planning horizon, using the value function from step 3.
  5. Simulate operation of the optimization scheme using a simple test system.
  6. If time permits, validate on real-world data, for example using a real bike sharing scheme for proxy data.

Weitere Informationen

John Lygeros

Art der Arbeit: 80% theoretical, 20% simulation
Voraussetzungen: Some optimization knowledge, as well as dynamic programming and/or predictive control, would be highly desirable.
Anzahl StudentInnen: 1
Status: done
Projektstart: Summer/Sept 2017
Semester: Autumn 2017