Lufthansa Group is one of the world’s largest airlines, operating globally with close to 110,000 employees and bringing in over EUR 32 billion in revenues annually. Its passenger airlines include Lufthansa itself, but also Eurowings, SWISS, Austrian Airlines and Brussels Airlines.
This week, Lufthansa shared details about how it is working with Google Cloud to optimize its complex operations, making it easier for operations controllers to make decisions. This in turn is also reducing its carbon footprint, saving on costs, and improving passenger satisfaction rates.
Daniel Bogado Duffner, Product Owner "Irregularity Solver & Cost Functions" at Lufthansa Group, compared the operations of the airline to a complex body - or the body of an athlete - and the operations controllers to medical doctors that take care of these bodies. He said that they work 24/7 across the company’s hubs in Zurich, Vienna, Munich and Frankfurt to take care of the organization’s flights, network operations, crew members, and passengers.
These operations controllers have to make complex decisions on a daily basis, using data from a variety of sources, to help ensure that things run smoothly. The aim of the project with Google Cloud is to build a holistic optimizer panel that pulls the data from these different sources and uses machine learning to provide recommendations and predictions all in one place.
Providing an example of just one of the complex decisions a controller may have to make at any given time, Duffner said:
If we have a flight coming from Paris back to Zurich, to our hub, and it was supposed to arrive at 4.20pm - but it is now delayed by 60 minutes. This will have a lot of repercussions because, for example, 43 passengers were flying from Paris, through Zurich, to get to New York. And if they arrive late in Zurich, they won't be able to make the connection because the flight to New York was supposed to depart at 5:20pm.
So operations controllers need to take a couple of questions into consideration. Should the passenger stay in Zurich until the next day, or is there maybe a later flight available to transport these 43 passengers and is there actually availability on this next flight?
Or is there an option to reroute our passengers through the Lufthansa Group network, or even through the worldwide network of airlines? Or should we just wait for the passengers and delay the flight to New York, which will then have a bigger effect on all the other passengers traveling to New York?
Further challenges, he added, could be that the aircraft that was coming from Paris was supposed to fly to London Heathrow straight afterwards. With the aircraft being delayed, the flight to London is going to be later. Should that flight be canceled? Or should a different flight be canceled and the aircraft from that flight be used? Or is there a reserve aircraft around or perhaps an aircraft from one of the Group’s other airlines? You get the idea, lots of complexity and plenty of implications for each decision.
Crew also adds complexity. Duffner added:
The crew situation is quite relevant, because the crew coming from Paris were supposed to continue the day flying to Rome. And with the crew arriving delayed, they will also not be ready for the flight to Rome. They won't be able to prepare the aircraft, to check the aircraft and make sure that their flight to Rome can be operated in a safe manner.
So is there maybe a reserve crew around that we could use? And are we then maybe wasting this standby reserve pool? Or should we maybe keep them around because there might be a bigger, more critical effect upcoming? So all of these questions are in the minds of the operations controllers.
Duffner makes the point that this is just one example of a challenge, from one flight path that could arise in the midst of a whole flight schedule, on any given day that operations controllers are responsible for. In the context of all flights and all situations that arise, the challenge can be quite large for controllers.
A single panel
Understanding this context, Lufthansa Group is working with Google Cloud to build an AI-enhanced, unified and integrated ops decision support suite. Duffner says that this suite will be the driver for passenger punctuality, technical availability of Lufthansa aircraft, flight regularity, cost improvements, and also the complexity reduction of all the operations. Duffner explained:
The ops database that we are building up in Google Cloud, where we are gathering the data from all our source systems and creating this ops cloud layer on top - the exact process of how this looks is that we first ingest the data and harmonize it, so that it works across the globe. We then store the data in different technologies in Google Cloud. And lastly, we also enhance the data with different factors.
All of the state of the world data is coming from our source systems. Then we need to apply a lot of business rules and cost functions on top of it, to be able to assess whether certain options are possible at all.
And we also use the machine learning capabilities of the Google Cloud to give our operations controllers not just the data that they already know that they could find in their own systems, but to also give them better data, better predictions of what might be happening to them. To enable them to make better decisions.
Once Lufthansa has all of this data processed and created in its cloud layer, it then sends this to an optimizer, which has been built by the Google Operations Research team, and is provided via API. Although the project has started as a pilot within the SWISS Airline subsidiary, it is already delivering measurable benefits to the Group - particularly in terms of Lufthansa reducing its carbon footprint. Duffner said:
We will of course also optimize for the overall cost of our operation. And the good thing with an airline is that a big part of the costs are our fuel. And so saving costs also always means saving fuel - and saving fuel also means saving CO2.
So within SWISS alone, where the use case is already in production, we are saving more or less 9,000 tonnes of CO2 per year with this better assignment from the optimizer results.
This is equivalent to, more or less, 61 flights one way from Zurich to New York. And even though this is just a small number of the whole CO2 that we produce as an airline, it is still a significant number that could be achieved through a rather small and easy use case.
But cost and CO2 aren’t the only benefits being achieved, as the optimizer for controllers is having a knock on effect all the way through to passenger satisfaction. Duffner added:
Operations controllers have these different systems where all the data is coming from. And when they need to make a decision they will have to look into all of these different systems to make the decision. But with our tool we are actually giving them all the information at one glance - they can make the decision more easily and will therefore have a reduced complexity in their daily work.
But in the end, we will then have not only a more sustainable and cost efficient fleet assignment, like we're having right now, but we will also have robust schedules, decreased complexity for operations controllers, higher crew satisfaction and efficiency - and we will have less passenger irregularities and lower costs. And through this we will also have higher passenger satisfaction overall.