• Overview
• Visitors
• People
• Links

• Groups
• Publications
• Software

• Courses
• SA/MA
• Lab

• News
• Seminars
• Jobs

Optimal operation and real-time pricing of transport services

Large-scale transport operations, whether car rental, train rolling stock, courier services or casual schemes such as Mobility, must spend significant money and effort on ensuring that the correct vehicles are available in the right places at the right time. Small changes in strategy for operating such services can have a large impact on the associated costs. Here we consider how past data can be used to inform real-time pricing decisions for a bicycle hire scheme, for which extremely detailed past data is freely available online.

Many cities, such as Paris, Montreal and London have established city-wide bicycle hire schemes, charging users based on the length of their journey. The goal of such schemes is to encourage cycling within the city at minimum cost to taxpayers and users of the scheme. This brings public health benefits and eases congestion on public transport.

Currently a major operational cost of the scheme is moving bicycles from docking stations where people tend to finish their journeys back to those where they tend to start them. This means that staff must be hired to move them around during the night for use again the next day, which entails significant expense. The problem also prevents schemes from being extended to less densely-populated suburbs, where such costs would be higher due to the wider spacing between stations.

One solution that has been proposed is to give users an incentive to return bicycles to a nearby but less popular docking station. This project investigates how such a scheme should be devised:

• Use historical data, or record data from the online interface to determine demand and supply behind each node. For the London scheme, publicly-available code already exists to extract real-time data.
• Use this data to model the network and simulate for results comparable to historical data.
• Formulate an optimization problem minimizing the cost of operating the scheme over a time horizon, subject to operational constraints and behavioural assumptions on the users, and use it to derive a price controller.
• Verify the results by simulating the system and evaluate potential cost savings.

This project would be rewarding if you have an interest in, and would like to improve your understanding of, the following areas:

• Use of Matlab for system simulation
• Optimization theory and model predictive control
• Interest in online data “mash-ups”/visualization would be a bonus.