Real-time analytics help improve passenger flows at London Heathrow Airport

By creating passenger flow forecasts at Heathrow, Professor Bert De Reyck used machine learning techniques to improve the passenger experience and support airline, airport, and air space punctuality.

Prior to the COVID-19 global pandemic, Heathrow Airport was the busiest airport in Europe, carrying more than 80 million passengers each year to over 200 destinations worldwide. The airport community employs over 6,500 people and operates collaboratively with over 200 stakeholder organisations. Improving airport collaborative decision-making is at the heart of Heathrow’s Airport Operations Centre (APOC), established in 2014. The APOC brings together all airport stakeholders in a single room, providing real-time access to shared data sources, with the intention to get an accurate overview of the status of all the airport’s operations, enabling real-time communication and collaborative decision making on operational functions such as passenger flow, baggage flow, security, gate management, and crisis management.  

In collaboration with London Heathrow Airport, Professor Bert De Reyck led a team to develop and implement a real-time predictive system that predicts passengers’ journey times through the airport, the expected number of late passengers for each outbound flight, and estimated passenger flows at the airport’s immigration and security areas. The project was based on underpinning research developed by Professor De Reyck on applying machine learning to generate real-time quantile forecasts, in this case for transferring passengers’ connection times, and on copula-based simulation methods to produce aggregate quantile forecasts, in this case for the number of passengers arriving at the immigration and security areas.  

Implementation at London Heathrow Airport 

The system was implemented at Heathrow Airport in 2017, and its real-time predictions are currently being used to improve resourcing levels at security and immigration, to inform airlines of delayed passengers, and to inform the airport of required departure time adjustments. 

The predictions generated by the system are currently used to assist managers at the APOC in making real-time decisions, including: (1) identifying passengers who are likely to be late for their connection flight and contacting ground staff to provide support to these passengers; (2) communicating with airlines to inform them of likely delayed passengers, and to help them adjust their departure times; and (3) adjusting resourcing in real time at the immigration and security areas based on expected passenger flow numbers in 15-minute intervals. A recently completed back-testing study revealed that the system offers a reduction in cost between 12% and 54% at immigration and security areas compared to Heathrow's legacy system. In addition to providing accurate forecasts, the system also helps the APOC mangers understand the key factors that influence passengers’ connection times, including the time of day, type and capacity of aircraft, load factor, travel class, inbound and outbound flight region, and punctuality of the flight). 

The Head of Airport Research at Eurocontrol writes  that this “ground breaking” study has “become a reference”, and “is used by a number of major European airports [including Paris CDG] with a first deployment since 2017 in London Heathrow’s Airport Operations Centre”.  He adds that the study “demonstrates how decision making can be better informed by the flow of data and the use of predictive algorithms, and brings state-of-the-art thinking in machine learning, applied to a problem of crucial importance to airports around the world into the airport operations domain”.  

Links

• Professor Bert De Reyck
• UCL School of Management
• UCL Faculty of Engineering Sciences
• Faculty of Engineering REF 2021

Image 

• Image credit: Shutterstock / EviJanku