“Flight 457 to Los Angeles has been delayed…”
We’ve all been there, frustrated and tired asking ourselves, “How is this possible?” as we make our way to the nearest electrical outlet (or bar) to wait for the next available flight.
As it turns out, flight delays are fairly common in the United States, accounting for, on average about 20% of all flights in the last decade according to the US Department of Transportation.
As much of an inconvenience as it is for you, it is an even bigger problem for the airline industry. Considering the scale of operations for some of the largest airlines, if the delay is just a few minutes over the threshold that defines a delayed flight (15 minutes or more), it can have serious financial repercussions. To give you an idea of what is at stake, in 2019, the FAA estimated that flight delays cost U.S. airlines 8.3 billion dollars just in operating expenses. You could say this is a problem worth solving.
The big question for these airlines is, how can we improve our operations to reduce delays, provide a better customer experience and improve the bottom line? The answer, as it turns out, is complex, as it requires the coordination of a number of different planning processes that all affect one another. In other words, it’s an optimization problem that consists of smaller optimization problems. This includes calculating the most efficient and profitable flight schedules, aircraft routes, crew pairings, maintenance schedules, food and drink readiness, and more to be cohesively organized and capable of handling disruption.
To solve these optimization problems, airlines are turning to advanced optimization-based decision support systems to generate efficient flight scheduling solutions as well as many more applications to reach their operational and financial goals. Let’s dive into some of the critical optimization problems in the airline industry and discover how new technologies like MemComputing are delivering significant value.
Optimization in the Airline Industry
The airline industry is a very competitive market consisting of dynamic players who operate on a large domestic and international scale. There are complex logistical operations, many rules and regulations that must be followed, and price wars. From a business perspective, it is a perfect place to employ optimization technologies to try and obtain a competitive advantage. In fact, many optimization techniques have been introduced and deployed in the airline industry over the past few decades. For the sake of keeping this short, we’ll focus on three major optimization areas in the airline industry:
- Flight Scheduling
- Fleet Assignment
- Crew Scheduling
The flight schedule in most airlines is often created months in advance, allowing other departments to evaluate and edit its economic and operational feasibility. The ultimate goal is to maximize its market share and profitability with limited aircraft capacity. To create an efficient, cost-effective flight schedule, airlines must consider:
- Market demand for destinations & time of day
- Aircraft characteristics such as speed, capacity, fuel costs
- Route featuristics
- Operational restrictions
It is tremendously difficult, in fact it is intractable, to render a truly optimal flight schedule that accounts for all of these details and adheres to all operational and legal constraints. That is why airlines include buffers in their schedules to try to minimize excessive delays and offset minor disruptions. Due to the scale of airline’s operations, today’s optimization techniques are forced to include simplified functions of these factors to find feasible solutions. This ultimately results in sub-optimal solutions, but provides a baseline framework for the other planning processes to follow.
Once the flight schedule is designed to the best of their ability, airlines then perform fleet assignment, which aims to find the best assignment of available aircraft to the scheduled flights. This is another difficult optimization problem that airlines face. The goal is to maximize the profitability of each flight based on factors like:
- Seating capacity
- Passenger demand
- Operational costs
- Availability of maintenance at arrival/departure stations
There have been many optimization methods used to solve the fleet assignment problem. However, most are forced to rely on heuristics, which provide approximate solutions to this problem. Nonetheless, a feasible solution is found often through third party optimization software and/or internal systems.
The crew scheduling process typically starts after the flight assignment for the current period comes out. The goal is to optimize the assignment of pilots and flight attendants to a set of scheduled flights to ensure all flights are covered and the safety and labor rules are respected. The airline crew scheduling problem is a famously difficult problem to solve, as there are many variables to consider and delicate constraints to satisfy (plane type, schedule, # of crew needed, individual crew location, crew specific flight hours & required rest periods, payment).
Although many approaches have been introduced to solve the crew scheduling problem, it is difficult to solve in its entirety, which is why it is commonly separated into two processes; crew pairing and crew assignment. Still, these sub problems are too large to solve optimally at scale for current optimization methods, so airlines are forced again to rely on heuristics.
For most airlines, employing the crews to operate and service their planes is the second largest operating expense, fuel being the first. Just small improvements to crew scheduling can result in millions of dollars of cost savings, which supports the heavy interest within the airline industry to evaluate new optimization tools, like the Virtual MemComputing Machine (VMM).
Many airlines today generate schedules, manage resources and control operations using fragmented systems. For example, the system that creates aircraft routes does not consider the complexity in generating the corresponding crew pairings. This limits their ability to collect and integrate data across systems to solve many of these problems efficiently and reliably. Because of this, managers still rely on manually-generated optimization solutions to make planning and operational decisions. Although some commercial airlines have some optimization systems embedded in their decision making processes, they generate sub-optimal solutions that are unable to address these problems at scale.
This results in increased operating expenses in fuel, maintenance, crew, and more. Buffers in the schedule result in longer lead times, which inevitably leads to poorer aircraft utilization and the need for larger fleets (extra planes to swap in for real-time schedule interruptions). In addition, moderate or excessive delays have a ripple effect and disrupt flights down the line sometimes for days or weeks, resulting in significant additional operational costs.
According to some (pre-pandemic) calculations done by All things On Time Performance, the three largest U.S. airliners combined have 35,000 delayed flights per month on average. The average delay is 12 minutes. The average cost per delayed minute is roughly $74 ( $74 * 12 * 35,000 = $31,080,000 per month).
More and more Airlines are reaching out to MemComputing to take a huge bite out of the $millions they are losing each month. The Virtual MemComputing Machine, or VMM, represents a new computing technology that solves optimization problems at the scale required by the airline industry much faster and more accurately than today’s best in class approaches. The tremendous value that is achieved through improved scheduling and planning results in significant cost savings, higher levels of customer satisfaction, and stronger resistance to disruption.
More important than just developing schedules, MemComputing is the only technology that can truly provide optimal solutions based on interruptions. The VMM has the power and performance necessary to reschedule an entire fleet based on interruptions due to individual aircraft maintenance issues, severe weather delays, etc. Not only can the VMM reschedule all affected routes, aircraft, crew, passengers and cargo in minutes, the solution MemComputing provides will help the airline recover quicker and get back on the expected schedule. A solution such as this can save individual airlines $millions for each interruption.
MemComputing has been working with Airlines and Aircraft scheduling companies since 2019. Much of the work that we perform is held as confidential due to the proprietary information associated with each Airline. However, one example of a hard optimization problem in the airline industry that we are able to publicize is the optimal cargo loading onto an aircraft. The goal is to calculate the optimal configuration of cargo to be loaded on an aircraft under specific capacity and operational constraints in order to minimize costs, while, and this is critically important, ensuring that the maximum allowed weight of the cargo is not exceeded and that the plane’s center of gravity and flight characteristics will not be impacted. This problem is not possible to solve at scale for current methods. There is hope that Quantum Computing may one day be able to solve this. In that regard, Airbus included this problem in its Quantum Computing Challenge in 2019. We, at MemComputing, participated in this challenge and successfully rendered solutions at the scale industry faces today, while adhering to all constraints. This problem cannot be solved by current Quantum Computers, and, at best, they hope to be able to solve this in 10 or more years. MemComputing’s solution is ready to be deployed to the field today. Details can be found in this case study.
As you can see, optimization plays a huge role in the airline industry across a variety of areas. It helps companies plan and manage their resources to drive operational efficiencies across the entire lifecycle of the business. However, companies are still struggling to solve these problems optimally at scale, making the need for more powerful optimization technologies like MemComputing more and more apparent.
Those who adopt such technologies early will be better positioned in operational capacity, disruption, and achieve significant competitive advantages. If you’d like to learn more about the Virtual MemComputing Machine and how it can address Airline Scheduling or other problems, please visit our website or create a free account to get started today.