TravelInt

Context

A complete understanding of the behaviour of passengers, as final users of the air transport system, is a key aspect for airport planning and management. Airports must be able to adapt to the constantly changing market demand in order to acquire a better understanding of passenger behaviour and adapt their strategies accordingly. The traditional approach for collecting this information is based on passenger surveys, which have important limitations in terms of cost, sample size and frequency of update. In recent years, the widespread use of personal mobile devices and IoT sensors has opened new opportunities to collect a huge amount of passenger behaviour and mobility data in a continuous and cost-effective way, with a level of detail not achievable with traditional methods.

The project TravelInt

TravelInt is a research project funded by Red.es that proposes to take advantage of these opportunities by exploring how big data and artificial intelligence technologies can be used to enable the detailed analysis of the passengers’ behaviour and support the airport decision-making processes. 

TravelInt tackles three specific use cases in which the improvement of the understanding of the users’ behaviour is particularly valuable:

1. Route development: this use case will analyse and model long-distance travel behaviour, with the aim of identifying new attractive destinations and offering new routes. 
2. Accessibility and intermodality: the behaviour of passengers, employees, and other airport visitors in the airport access leg will be analysed, with the goal of optimising the connectivity between the airport and its catchment area. This is expected to improve passenger experience, facilitate the use of more sustainable access modes, and attract potential demand that would otherwise opt for a more accessible airport or an alternative long-distance mode.
3. Non-aeronautical revenue: the detailed characterisation of the behaviour of passengers within the airport terminal will help optimise the passenger experience and adapt the shopping area and other airport services to the passengers’ preferences and needs.

Goals

The specific objectives of TravelInt are the following:

1. To identify the opportunities offered by personal mobile devices to improve the understanding of the behaviour of passengers and other airport users regarding three main aspects:
   1. Long-distance travel demand behaviour.
   2. Airport access and egress.
   3. Behaviour within the airport terminal.
2. To develop new data analysis and fusion methodologies that allow the detailed characterisation of the profile and behaviour of passengers and other airport users.
3. To develop a set of machine learning predictive models able to forecast passenger behaviour and support what-if analyses of alternative scenarios and management actions. 
4. To develop a prototype decision support system that integrates the new algorithms and models developed in the project with an interactive dashboard.
5. To validate the proposed system through a set of experiments carried out in close collaboration with the final users in order to demonstrate the maturity of the technology and assess its benefits.

Methodology

TravelInt aims to develop a big data platform that supports the airport operator decision-making process. The platform will consist on three main components: 

1. An ETL (Extract, Transform, Load) process that will integrate data from heterogeneous data sources and that will perform different data pre-processing tasks (cleaning, formatting, etc.) for their subsequent analysis.
2. A processing core which consist of:
   1. A data analysis layer that will integrate different algorithms to analyse and characterise the airport users’ profile and behaviour using the different data sources available. 
   2. A layer for the generation and prediction of indicators that will be fed by the information generated in the previous layer, and that will consist of three different modules, each of them covering the use case of the platform.
3. A visualisation tool that will allow the analysis of the results and the benchmarking of different scenarios through a visual interface.

The architecture of the proposed platform is presented in the following figure: