(PI UMa: Vassilis Kostakos, Co-PI CMU: Anind Dey)

Starting Date: 1 Jan. 2009, Duration: 36 months, Person * month in this task: 50

Description

This work package will focus on motivating sustainable transport behavior. It will:

·       Develop tools and infrastructure to capture individuals’ mobility around a city, and to capture the aggregate behavior of public transport passengers.

·       Construct predictive models of personal and public transport.

·       Deploy interventions that modify transport behavior, and measure their effect on sustainability.

To capture individual mobility, we will use GPS-equipped mobile phones and standalone GPS. These will be issued to a cohort of participants, who will contribute to a longitudinal study of personal transport. To capture public transport behavior, we will rely on our ongoing relationship with the public transport authority of Funchal (industrial partner HF). They will provide ticketing data for the whole of their network, including access to frequent passenger data. In addition, we will expand our existing prototype which uses Bluetooth technology to precisely record passengers’ entry and exit points on the network. The system currently exists on 2 urban buses, but will expand to 20 buses, including interregional services.

We will analyze the data collected using machine learning and data mining techniques to derive models of people’s mobility. In particular, we will develop models to predict when passengers are likely to change mode of transport. This, for example, includes understanding when a passenger is about to drive their car. Relating to the public transport network, we will be aiming to develop individual models of passengers, capable of describing and predicting their unique use of the network. For example, a model could predict the specific services that a passenger is likely to use, or the specific times this is likely to happen.

The final stage of this work package is to develop interventions that result in more sustainable practices and policies for operators and public administration. Our interventions will attempt to to find ways to change people’s transport behavior by 

·       Engaging people when they change mode of transport;

·       Presenting longitudinal information about their behavior;

·       Allowing the transport authority to increase the efficiency of the bus network. 

We will employ motivational techniques (from WP T1) to try and modify people’s behavior when they are about to change mode of transport. In addition, we will work with other work packages to develop interactive visualizations of people’s use of personal and public transport, in an attempt to inform them about their behavior and its impact on sustainability, and to allow people to choose a better transport mode for their own mobility requirements. Finally, we will work closely with the public transport authority to modify and adapt the service network in order to increase seat occupancy across the bus network, adapting better service to the needs of real people. The impact of interventions will be empirically evaluated via ethnographic studies, questionnaires, and data analysis.

Results

The results will include:

·       Development of infrastructure

·       Development of models

·       Evaluation of motivational techniques

Our infrastructure will include software and hardware for capturing personal mobility, and public transport use. The resulting systems will act as a platform for monitoring mobility across a city for any purpose, hence enabling us to design, build, deploy and evaluate a multitude of smart applications. In terms of developing models of transport, we expect to make a number of contributions to the algorithms necessary to capture, filter, and analyse mobility data. We will deliver models of personal transport with a focus on predicting changes in mode of transport, hence our algorithms will need to analyse multiple streams of contextual data (such as location, time, nearby people, public events, and weather). For this work we will: 1) identify people with similar models (while maintaining anonymity) and create lightweight social networks (groups) that share information with each other about best practices; 2) use group membership to improve prediction in underspecified cases (those with little data on either the individual or situation); and 3) aggregate data (to anonymize) and support the bus company in the development of routes.

A unique aspect of our approach is that we are not concentrating on predicting location or destination, but rather identifying and predicting the mode of transport (walking, cycling, car, bus, underground, tram, etc.) Our work on modeling public transport use will rely on our existing prototype that can capture the exact time when passengers board and exit a bus. This information will be used to construct personalised Origin-Destination matrices for passengers of public transport. These will be analysed in conjunction with ticketing data to derive predictive models of use. We expect these to help the public transport authority to redefine and optimise the service network, aiming to increase seat occupancy across their fleet. Finally, this work package will work with WP T1 to deliver an assessment of the effectiveness of a number of motivational techniques used to change people’s attitude towards personal and public transport. We will consider multiple interventions: the first is a simple visualization of their mode of transport and the impact on the environment. A second could be predicting the mode of transport that someone is likely to take for a particular route segment and suggest an alternative for that. A third, leveraging the work going on at CMU, is to predict the route and the mode of transport, and suggest alternatives for the entire route, leveraging knowledge of the bus schedule, bike paths and pedestrian areas. We will assess whether information presented at the time of decision making, as opposed to being accessed on an ad-hoc basis by individuals, has a stronger effect on changing passenger behavior.