Artificial Intelligence (AI) and Operations Research in Aviation
Artificial Intelligence (AI) is transforming the way we do business. Industries are racing to adopt AI technology to learn about their market, make better decisions, and improve their operations.
At Lean Systems, we specialize in Operations Research (OR), a branch of AI that we define as the use of mathematical modeling and optimization to arrive at optimal or near optimal solutions to complex decision making problems. OR lends itself well to solving complex logistical problems inherent in transportation, and in particular, the aviation industry.
Broadly speaking, civilian aviation industry can be divided into two categories: commercial and business aviation. Commercial aviation is generally scheduled in advance (~2 months or more), while business aviation is generally demand responsive, scheduled 1 month to 24 hours in advance, and adjusted until the last minute.
Until recently, only large commercial aviation companies were able to afford costly OR technology to solve their planning, and scheduling problems. But advancement in the field, coupled with creative and new business models, are lowering the costs of these solutions, making them more accessible to small- and medium-sized business aviation companies. We can attribute these advancements to the development of high quality open source OR libraries, market adoption of cloud-based services and web APIs, and recent breakthroughs in machine learning and data science which can be integrated with OR applications.
This article explores the role of AI (OR in particular) in addressing the complex logistical challenges of commercial and business aviation. OR provides useful tools for making strategic (long-term), tactical (medium-term), and operational (short-term) planning decisions.
Making Decisions with Operations Research
The first order of business for an aviation company is to decide on strategic planning decisions like which markets to target, and their expected volume. These decisions require making complex trade-offs that can only be fully explored using OR’s mathematical optimization techniques. Here are some examples of strategic planning decisions for aviation companies:
Should we open new bases? if yes, then where and when?
How many people should we hire?
Should we shift from renting aircrafts to owning and managing them?
What is the ideal composition of our fleet based on the last 3 years of growth?
What is the ideal composition of our fleet for the next 1-3 years?
Tradeoff is a common theme among these decisions. For instance, consider the first decision of when and where to build new bases. While deciding on the location of a new base, managers must take into account the activity of each market to identify a midpoint that minimizes the overall distance covered by their aircraft. In OR, the problem is known as the facility location problem and can be modeled and solved very efficiently.
The number of new employees to hire involves a tradeoff between opportunity cost and paying salaries.
Aircraft renting versus owning decision involves a tradeoff between large upfront cost versus recurring costs. The ideal composition of the fleet involves a tradeoff between different market needs, and so on.
Using the right data, OR practitioners quantify these tradeoffs and solve optimization problems that determine the best possible decisions.
After the strategic decisions are sorted out, decision makers have to make a series of tactical decisions. An example of these decisions in the commercial aviation industry is fleet assignment.
When faced with a fleet assignment decision, managers must decide which aircraft types to assign to different flight legs. The trade-off is straightforward. If the aircraft has more seats than the number of available passengers, the company incurs extra operational cost, while if the the number of passengers exceed the capacity, the company ends up with an overbooking situation and loses the opportunity cost of serving more passengers.
Using OR tools, the problem can be solved to optimality, yielding a solution that maximizes revenue and minimizes cost for the company. In business aviation, fleet assignment is often intertwined with other planning decisions such as crew scheduling, but the principles are the same.
While strategic and tactical decisions set the tone for efficient operation of an aviation company, it is the operational level decisions, such as crew scheduling, that are most challenging. Salaries of crew members are among the largest costs incurred by both commercial and business aviation operators. Therefore, an optimal assignment of crews to flight legs results in substantial savings. Merely finding a feasible solution for the crew scheduling problem is challenging, let alone finding an optimal one.
The reason for this complexity arises from the immensely large number of possible ways to assign crew to airlines. Consider the case of a small business aviation operator trying to assign a crew of 20 members to 5 flights, each requiring 4 crew members. Without considering any rules (such as legal requirements), and assuming all members can operate on the all airplanes, there are 3.2 million possible assignments!
This phenomenon is known in mathematics as combinatorial explosion, where a seemingly small number of options result in an extremely large number of combinations. OR provides advanced techniques to leverage existing business rules to weed out undesirable assignments and produce a tractable solution. Crew scheduling decisions can be automated and solved in a relatively short period of time while producing solutions that minimize crew costs and deadheads.
These brief examples show how AI in general, and OR specifically, have the power to tackle complex logistical challenges in the aviation industry. While these techniques were traditionally reserved for large-scale commercial aviation operators, the technology is quickly becoming available to small- and medium-sized companies in the business aviation sector. In a future article, we will discuss why these technologies are becoming affordable, and how your business aviation company can take advantage of them.
Get our Artificial Intelligence in Transportation e-book here.
Get a PDF version of this blog post here:
 Cynthia Barnhart, Peter Belobaba, Amedeo R. Odoni, (2003) Applications of Operations Research in the Air Transport Industry. Transportation Science 37(4):368-391. https://doi.org/10.1287/trsc.37.4.368.23276
 Schön C. (2007) Market-Oriented Airline Service Design. In: Waldmann KH., Stocker U.M. (eds) Operations Research Proceedings 2006. Operations Research Proceedings, vol 2006. Springer, Berlin, Heidelberg
 Gopalakrishnan, Balaji, and Ellis L. Johnson. "Airline crew scheduling: state-of-the-art." Annals of Operations Research 140.1 (2005): 305-337.