Imagine a world where airline delays were predicted in advance.
Agents could rearrange plans for travellers while they are still in the sky, not only saving time but creating a massive value add.
According to Stephane Cheikh, Artificial Intelligence Program Director at SITA, the answer to predicting these delays may well be AI.
Disruption is one of the main issues facing the air transport industry.
We’ve all experienced or heard about disruption causing delays, often with huge impacts on passengers, cargo and operations throughout the rest of the day.
Let me put it into perspective: recent figures show average flight delay times at 51 minutes. Delays could be costing the air transport industry as much as $25 billion a year.
IATA estimates an average on-time performance of 76 per cent for the 26 million flights globally each year. Without action, and given the pressures of growth in air travel, problems will only worsen.
That’s why, in their search for operational excellence, both airlines and airports are seeking more control and predictability.
From my own experience, and in recent collaborations with SITA air transport customers, it’s clear that Artificial intelligence (AI) promises a new way forward.
AI embraces the disciplines of Machine Learning, Machine Vision, Natural Language Processing, and Robotics. It’s attracting great interest from airlines and airports. According to SITA’s 2018 Air Transport Insights research, AI is one of the emerging technologies offering future strategic and operational benefits.
The Insights research shows that 66 per cent of airlines are implementing or planning predictive analytics capabilities by 2021.
In the meantime, 79 per cent of airports are using or planning to use AI for predictive analysis to improve operational efficiency (e.g. task automation) by that same year.
With all eyes are on the potential of AI, I can see several very promising signs, based on SITA’s co-developments and experience with customers in the use of AI for predicting flight arrival times. And we are only at the very beginning.
Of course, in predicting flight times, challenges do exist, not least taking account of the multiple stakeholders in the equation. Having said that, for airlines, many factors are under their control. It might surprise you to learn, for example, that fewer than 5 per cent of flight delays are down to extreme weather (see chart).
The main issue in delays is turnaround time. If fueling, catering or cleaning over-run for whatever reason, then flights will continue to be late from thereon in, a headache for the airline and the airport.
Delay causes by year in the US (2008-2017)
Sita is working closely with a major airport in APAC, which is a great example of pioneering work in AI and predictive disruption. Here we’ve shown that AI can draw valuable insights from masses of historical data accumulated from flights and delays coupled with real-time information.
The airport has a complex network involving over 200 organizations. It can get as low as a 30-minute update time on an arrival. Any insight the airport can get into flight arrival times to plan resources is invaluable. Working closely with the airport, we’ve developed the ability to accurately predict arrival times disruption, through a predictive model using AI and data analytics.
Applying Machine Learning to historical and real-time data, the airport can accurately predict flight arrivals two to four hours out.
This data can be used to plan resources and pre-empt the impact on passengers, some of whom may not make a connecting flight. The next step is to improve the accuracy and predictions, especially for shorter flight durations.
Right now, prediction requires the experience of humans to make decisions, because we must assess a whole list of factors, from ATC, maintenance and crew connections to ground handling and scheduling integrity. That’s why I use the words ‘collaborative intelligence’ as we need a combination of human and AI to improve efficiencies.
AI can help humans look at passenger revenue and value, for example, and quickly re-book high revenue and value individuals first if a flight is cancelled. By accelerating historical data value analysis using AI, we can create a list according to priority – at speeds and with levels of accuracy simply impossible by humans.
Then there’s the potential of using AI (Machine Vision, aka Computer Vision) to visually track and monitor all the stakeholders around the aircraft to proactively manage turnaround, and constantly improve coordination and efficiencies. As another focus area for SITA, I can see this delivering far better visibility of aircraft turnaround, helping to identify issues before they happen and driving significantly better on-time performance.
So the future for airlines and airports is undoubtedly getting smarter. AI is starting to link the players and the processes across the passenger journey, providing a smart way of managing an increasingly complex ecosystem. It’s clear to me that AI is taking us one step closer to operational excellence.