A New York Times piece last year addressed the question, “Are you ready to fly without a human pilot?” which sparked a vigorous conversation on Facebook. Many users responded with comments that were quite emotional, and other users also volunteered their expertise on the aviation sector. Others claimed that people were more amenable to accepting autonomous automobiles than pilotless aircraft, maybe because driving feels safer than flying.
Technologies like computer vision power the automation and self-driving cars, but are we close to commercial planes being controlled by AI systems instead of pilots? Prototypes of self-piloted passenger air vehicles by Boeing and Airbus have already done their first test flights. But for now, we can put aside futuristic scenes in which commercial airplanes fly without human control. It will take years of certification and testing before air travel becomes fully pilotless. The good news is that airlines do use AI or rather data science and machine learning to automate pr speed up operations. So we will discuss real-life use cases that don’t aggravate your aerophobia with AI phobia.
AI and Data Science in Aviation
Revenue Management and Route Planning
A direct flight between Chicago O’Hare and the Polish city of Krakow by a lot of polish airlines may strike you as overreach. Chicago and Krakow are both large airport hubs but is that enough to run cross-Atlantic flights between noncapital cities. Well, yes, the Chicago metropolitan area is home to an estimated 1.5 million people of Polish descent. The real demand for this city pair is high. One of the key issues airlines must solve to thrive is how to price flights and determine traveler demand for particular city pairs. Carriers must evaluate data while taking into account thousands of parameters to do this.
Analysts can still make use of conventional statistical methods. Demand analysis can now be done in more complex ways, thanks to data science. The International Air Transport Association (IATA) advises that airlines can leverage traveler behavior data and abandon searches on online travel agencies, Meta Search engines, or social media chatter to define leisure demand. Data from professional networking sites, recruitment, and procurement activities, may signal emerging business travel destinations. In a 2017 showcase for airlines, Skyscanner used machine learning-based clustering to the group about 50,000 origins and destinations by similarities. They considered about 30 parameters, like the month of travel, the time a reservation is made, how long people stay at the destination, and more. Some results are quite surprising. For instance, the cities that are traditionally considered best for romantic trips were as popular with single travelers, and the proximity to a destination may be more important than the city itself. On top of that, events like festivals, conferences, or Expos drive short-term spikes in demand. So, revenue teams can rely on event data to raise fares for specific routes and dates to benefit from rising demand.
To determine how much a specific event may impact traveler demand, predict HQ’s aviation ranking system compares previous flight bookings with event data using ranking algorithms. Qantas One of the company’s clients is the national airline of Australia.
In-Flight Sales and Food Supply
Imagine you’re having an early morning flight. Challenging stuff! Once you’ve gone through airport chucks and finally taken your seat, you might think about a cup of coffee and a sandwich, and a carrot cake looked tasty on a menu. But some people never order airplane meals. So, the airline supply management specialists must estimate how many snacks and drinks they onboard to deliver to food eaters without being wasteful. Cabin waste is a serious environmental issue.
In 2018, Airlines generated 6.1 million tons of waste. Most of which were incinerated or moved to a landfill. In spring 2008, easyJet CEO John Lundgren tossed the data science team to analyze the demand for food items depending on a route. The team learned that the demand for items on a 6:00 a.m. flight to Edinboro is very different from that of a Friday night flight to Ibiza. So, EC jet threw three fresh food items in the trash after each flight or nearly 800,000 annually. John Lundgren noted that such a mistake cost the carrier millions of pounds. Eventually, data scientists created a new algorithm for demand prediction. The insights helped the airline save a significant amount of money and do the right thing for the environment.
Fuel Consumption Optimization
Commercial aviation contributed 2.4 percent of global co2 emissions from fossil fuel use in 2018. The percentage doesn’t seem significant, but here is another fact carbon emissions increased by 32% over the past five years. That’s why aircraft manufacturers and airlines are looking for ways to improve their fuel efficiency. The second big reason for carriers to reduce carbon emissions is a financial one. In 2018 airlines spent 23.5% of total expenses on jet fuel. That’s a lot! To become more fuel-efficient, an airline must accurately predict how much fuel it needs for every scheduled flight to supply a plane. The best scenario is to have a single analytical tool.
Southwest Airlines worked on such a solution in its fuel consumption project. The team developed eight predictive models that included time series algorithms and neural networks. The system could produce 9600 fuel consumption forecasts for each month and each Airport the carrier flies to. Previously the team generated 1,200 monthly forecasts, and each analyst spent up to three days making one. The new solution does it in five minutes. It generates forecasts for a 12-month horizon and considers such influencing factors as fuel price, number of trips, and time period. The predictions also became much more accurate.
Boarding and Checking Bags With Facial Recognition
Facial recognition technology is about analyzing a person’s facial landmarks for given purposes. Airlines use this biometric technology as a boarding option. The equipment scans travelers’ faces and matches them with photos stored in border control agency databases. These can be photos from passports, visas, or other travel documents. Here’s how it works, travelers first gain themselves and then their passports at self-service kiosks, then proceed to check in their backs with a scanner – and go through another facial scanner.
Why is it needed? Government agencies such as the US Customs and Border Protection stress that the technology allows for creating a seamless traveler experience. That’s faster and safer. Numerous airlines are either piloting or already using biometric gates in selected airports. Delta is one of them. In November 2018, the airline opened a biometric terminal at the Atlanta Airport. The airline claimed the terminal is the first of its kind in the US. Delta also noted it would start testing the technology in all 14 international gates at Detroit’s McNamara Terminal by mid-December 2018. And added that in 2019 customers will be able to use biometric boarding from curb to gate. In the summer of 2019, Delta introduced facial recognition in another Atlanta Airport terminal as well as in Minneapolis, Detroit, and Salt Lake City airports. That’s 49 more gates equipped with facial recognition software. Travelers seem to like the new boarding option. Delta surveyed customers at Atlantic terminal F and discovered, for instance, that 70% of them found the biometric boarding experience appealing, and 72% favored it over the traditional one.
Preparing a Plane for the Next Flight
Passengers sometimes have to wait at the gate to get on a plane when an aircraft isn’t ready for boarding and departure. Maybe a catering truck is late, or the cleaning team is busy with a different jet. The time between an airplane landing and the next departure is called a ‘turnaround.’ In 2018 US passenger airlines were losing an average of 74 dollars and 20 cents per minute due to delays. the US Department of Transportation calculated those delays caused by plane servicing accounted for 5.8 percent of all delayed flights. That’s nearly six times more than whites delayed because of extreme weather.
Zurich-based startup aside provides software that processes video streams from airfields using image recognition algorithms and neural networks that power the software and recognize objects, movements, and interactions. Airline employees can monitor in real-time how the plane is being prepared for the next flight, including fueling, cargo loading, or catering delivery. So, they can decide whether they need to take measures or not.
Lufthansa systems address the same problem with its deep turnaround solution. The IT service provider for the aviation industry opened its subsidiary zero-g to work on its AI initiatives. The solution also analyzes video data and updates users on what’s going on during airplane services in real-time.
Airlines use data science and machine learning to evaluate passenger demand across different routes. Use data insights to optimize aircraft ground handling and fueling or redefine passengers’ airport experience with biometric boarding. With new distribution technologies, we may expect airlines to start personalizing offers for individual travelers based on their preferences and willingness to pay. Personalization is one of the priorities that iota outlines. As for the AI autopilots, airlines are very conservative when it comes to new technologies that can directly impact the safety of flights. Carriers will likely wait until AI matures enough and until the traveling public is ready to trust it.