APIs, or application programming interfaces, are at their most basic level a set of standards and protocols enabling two or more computer systems to communicate and share information. But how are they used in the aviation context? We sat down with Dr. Alexander Motzek, Principal Data Scientist at zeroG, a subsidiary of Lufthansa Systems, for an overview of APIs and a deeper discussion around how aviation is using them — as well as what’s coming with the next generation of technology.
Fundamental to the idea of using an API is the exchange of information between at least two systems, whether those systems are within an organisation or cross corporate lines. The inherent benefit here is that the communication between the systems can be automated — and, more recently, can use new technologies like artificial intelligence and machine learning to process that information, and even act upon it.
For example, Motzek tells us, “we are developing artificial intelligence and machine learning models for very specific use cases. But not every use case is so specific that it is only applicable to that single domain. You can reuse these fantastic machine learning models or prediction capabilities in different areas. We can make these predictions available through APIs.”
One example is DeepTurnaround, an app deployed via Microsoft Azure Stack Edge that analyses online cameras covering the airport gate and apron area. From one single camera looking live at the aircraft, DeepTurnaround has been trained to identify every element of the turnaround, from safety cones to staff and different types of vehicles.
“With that,” Motzek says, “we get perfect situational awareness that wasn’t available beforehand. We know at which precise point in time the aircraft got on block. When did fueling arrive? When did the gangway arrive? When did catering arrive? When did baggage unloading start? When did they finish? When did the first passengers get into the plane? We can even identify if the cleaning crew has gone through or when it’s finished.”
With the precise timestamps, DeepTurnaround helps to identify why a delay has happened, and automates the verification of service level agreements by various providers — catering, ground handlers and so on.
The benefit of APIs is that all this information is produced in a standard, specified way that can be immediately and automatically integrated into other systems. On the ramp, staff can be alerted that a key vehicle or service is missing, and the service provider can be automatically chased up via a notification. Airline management information can be updated and the information collated in reports for contractual services.
Services down the line can even be requested automatically: if, say, the lavatories have not been emptied at the usual turnaround (owing to a breakdown of the honeywagon, for example) then systems controlling the next turnaround can be notified to do so.
But not all APIs are built the same, Motzek highlights. “You need to know who provides the best APIs. We develop large revenue management systems based on machine learning and data analysis, and there’s a huge difference if you have such a revenue management optimisation API for a point to point carrier, or for a hub carrier. We care about these details, and it’s very important that you consume the right APIs for your use case, and do not consume an API that does not fit into your individual domain.”
Complementary developments, including the cloud and AI, are inherently linked to modern APIs
Deploying onto cloud computing infrastructure is key to successfully using and integrating APIs, not least because of the extra-large amount of computational resources that some applications need — but, importantly, this resource requirement isn’t constant enough to lend itself to insourcing the requirements. Also in the technology stack are the cloud platforms’ inhouse analytical platforms, which only need to be optimised rather than developed from the ground up. These can then be used across an airline, from head office managers to data scientists.
“Everyone is allowed to view the data, to analyse the data, and to derive their own insights, which is a huge benefit — compared to say, eight years ago, where you precisely needed to know who owns which data, need to approach them, [and] get the data into some kind of analytical environment, which often wasn’t available. Then you could maybe derive some insights,” Motzek notes. “Now, it’s directly available. And within one line of code, you can analyse, for example, the new operational insights that you need, through the help of an analytics platform that we as zeroG can spin up automatically through the use of automated API usage.”
At the same time, the advances in artificial intelligence (AI) in recent years have been nothing less than astounding — and their integration via APIs into airline operations will be revolutionary.
Motzek explains: “in the last two years, we have seen one of the most fundamental advances in artificial intelligence that is now made available through various APIs. For example: hugely large, pre-trained natural language processing models that are able to understand textual information from the scratch, if they are written in normal types of languages — for example, in normal English, or German or French, that you find in a book.”
While this natural language processing data has not yet reached the maturity to understand complex technical texts like an aircraft maintenance manual, say, this level of assimilation is well within the trajectory of AI’s near-to-medium-term horizon.
With today’s level of AI, though, substantial advances are already being made. The Lufthansa Group receives more than 40,000 weekly pieces of free text feedback: from gate agents and cabin crew about delays and defects, but also through written customer feedback and via the proverbial firehose of social media.
“Through Twitter, through Facebook,” Motzek explains, “everywhere you stumble across free text, that gives you the most important information about the complete operations in aviation: from MRO, over catering, up to gate processes, up to the airport experience, and up until inside the cabin. However, these free texts are basically only so far only consumable by a human — it must be read line by line, and then you need to process it with a human. That doesn’t really scale that well.”
Enter AI, which can do it — but which needs substantial experience both within the industry and within the AI field itself. Developing this independently for every airline is unlikely to be feasible, but providing it for use via an API is.
Continuing aviation’s digitalisation revolution, it’s clear that the use of APIs will only expand, with new use cases developing constantly as the benefits of being able to automate and integrate systems, both internally and externally, become compelling.
Author: John Walton
Published 04 May 2023
