Written by Christiaan Hen, Chief Customer Officer, Assaia International
Together with sustainability, capacity was probably the hottest topic in aviation in the pre-COVID era. Limited air space and airport capacity were bottlenecks for the ever-increasing demand for air travel.
As the COVID-19 virus swept across the globe and almost all airlines grounded their fleets, the world seemed to have changed. Logically, the main focus of the industry has changed from growth to survival. However, capacity problems and the need for optimization solutions might come back sooner than many might anticipate today.
There has been a very generic reaction pattern for all airports and airlines across the globe. Initially as the virus manifested in China, its potential impact was underestimated. Memories of the SARS outbreak in Asia acted as a frame of reference to guide reactions and decision making. Airlines reduced and eventually stopped flights to China as a reaction to the outbreak.
As the virus spread from Asia to Europe, it became clear that the magnitude of the situation was going to be much bigger than originally anticipated. But even as the first areas in Europe started reporting high numbers of COVID-19 cases, the reaction was initially very much related to these areas and resulted in airlines cutting specific destinations from their schedule. Only after local/regional containment of the virus failed, it became apparent that the industry might come to an almost complete standstill. With the US and Canada practically closing its borders, international flights dropped to around 5-10% from its pre-COVID levels. Even though it appears as if we are past the peak of infections in Asia and Europe, the same cannot be said with certainty for the US and definitely not for South America.
In their efforts to get legislators to allow international travel again and to regain passenger trust, airports and airlines have been working around the clock to come up with all sorts of measures in order to minimize the risk of infection and virus spreading while travelling. These measures include (but are not limited to): physical distancing, temperature monitoring, enhanced cleaning regimes, and more touchless travel opportunities.
Pre-COVID, airspace was regarded as one of the main capacity bottlenecks. With airlines flying only a fraction of their pre-COVID schedules, airspace capacity is not expected to be scarce in the short term. Especially since the usage of airspace will not be different from before the crisis. It is, however, good to keep in mind that this relief will only be temporary of nature (especially since solutions for this issue have long lead times). After full recovery with a return to pre-COVID flight numbers, the amount of airspace (without any changes to the use of this airspace) will regain its status as capacity bottleneck. Especially if we consider that the number of flights might increase faster than the number of passengers. This especially holds true once the temporary waiver on slot regulation is abandoned and airlines will have a strategic incentive to keep serving destinations, even with low load factors.
The other scarce resource, airport capacity, is a different story. Initially, mainly airside capacity (e.g. stand, taxiway or runway) was the main bottleneck. Intuitively, we expect these bottlenecks to disappear as a result of the lower number of passengers and flights. If we look at terminal capacity, ACI has estimated that due to COVID measures implemented by airports, terminal capacity will be reduced by as much as 20-40%. It is easy to understand that if every passenger needs more space, you can process less passengers with the same amount of available space. Also, process times are expected to significantly increase. Therefore, it is also expected that aircraft turnaround times will be much longer mainly due to longer boarding processes.
The issue described above obviously has a spillover effect from terminal capacity to airside capacity. Longer turnaround times mean lower stand utilization rates and lower stand capacity. Whereas pre-COVID an airport might have been able to do as much as 10 turnarounds per day per stand, this number will now be significantly lower. What makes matters even worse is that many airports will still be accommodating many aircrafts that are not back into service yet. So, the number of stands available for operational duties is less than before. Initially the amount of traffic might be so low that this situation will not be problematic. As traffic starts to increase, however, available stand capacity will very quickly become a problem again.
Coming out of the current crisis, airports and airlines will be forced to operate as efficiently as possible in order to be as operationally profitable as possible. With a stand capacity bottleneck this will unfortunately not be fully possible. Airlines will especially not be able to fly their desired schedules if they do not have the required stands available at origin and destination airports. For the airports, less flights also mean less revenue (both aeronautical as well as commercial).
So the challenge is, given the current circumstances, how can we optimize stand capacity in order to maximize the potential for airlines and airports to recover from the COVID crisis?
There are several factors that make stand allocation a difficult task.
As you can imagine, the task of stand allocation very quickly demands more computational power. Therefore, most airport stand planners already use software that helps them to create a stand plan. All of these applications, however, work with static business rules and require data inputs from third-party systems (providing input-data of unknown quality). They will calculate an outcome, which is as optimal as it can be, considered that it is based on incomplete input variables and static rule sets that are often not 100% up-to-date. In practice this means that the outcomes of these systems are almost never optimal and manual changes are required to optimize the outcome. This is both time consuming and as we have seen it is very likely that the human will also not find the optimal outcome because of the complexity of the task.
One could compare the task of stand allocation with a game of Chess or the Chinese game of Go. Traditional computer systems have not been able to beat humans because of their limitation to consider many different possible outcomes. Artificial intelligence software has proven to be the game changer in this field. It has successfully been applied to exactly these kinds of problems where many variable inputs need to be considered and many possible outcomes need to be evaluated to decide on an optimal solution. If we want to maximize the utilization of available stand capacity, we should consider these kinds of technologies because they provide superior results over their predecessors.
Another way in which AI technology can help airports improve their stand utilization rates through better planning is by dynamic stand planning rules. Today, most stand planning tools work with static rules. The rules are defined during the implementation of the tool. However, as time progresses, often these stand planning rules change without being updated in the tool. As a result, airports either end up with suboptimal stand plans and/or require more manual adjustments. An AI based stand allocation algorithm has the ability to learn from user feedback. This means that once the algorithm has proposed a stand plan, it will record any manual changes that are made to it. These changes are compared to the most recent version of the rules to see if the rules might need to be updated. If the algorithm finds a pattern in the human alterations to the automatically generated stand plan, it will propose the user to change or add this as a rule. After the user (or system administrator) has accepted the system’s suggestion, it will use this new rule to create future stand plans.
A second way in which we can enhance the outcome of stand planning exercises is by improving the quality of the data inputs. Any optimization is only as good as the input data used. The most critical data points for an optimal stand plan are aircraft arrival to the stand (expected in-block time) and aircraft departure from the stand (expected off-block time). We have already determined that both of these data points are not static which creates the need for continuous updates.
Before we look at how we could improve, let’s look at the status quo. Within the A-CDM framework, airlines share updates about the Expected In-Block Time (EIBT) with the airport. These updates are unfortunately not always very accurate and in any case not continuous. Typically, the airline will push EIBT updates to the airport at fixed moments in time. The best estimate of when the aircraft will leave is the Target Off-Block Time (TOBT) which is provided by the ground handler. Our studies of the quality of this manually created estimation have shown that the absolute quality (precision) of this data is typically very low. Furthermore, this data is not continuously updated and typically updates only arrive very shortly before the Scheduled Off-Block Time (SOBT), which means that the proactive power of these updates is often lost.
A resolution for the shortcomings described above can again be found in using Artificial Intelligence technologies. Both for the EIBT and TOBT we can use a predictor that will continuously calculate a Predicted IBT and OBT value that we will call PIBT and POBT. To calculate an accurate PIBT both historical as well as real time data about a flight can be used. Past arrival punctuality performance will already give a good indication of future performance. Just think of early flight arrivals in Europe from the US. Most operational employees will know that these flights typically arrive early as a result of tailwinds. This kind of history can be taken into account for a PIBT. Furthermore, real time data like weather data (is that tailwind expected to be strong today?) and real time aircraft location and speed data can be used to calculate a very accurate and continuous prediction about the arrival of the aircraft to the airport and even to the stand.
The POBT is slightly more challenging than the PIBT as there is typically no accurate and real time data available about the progress of the turnaround. However, our research has shown that an OBT predictor can be created based on AODB, weather and other public data. This relatively simple predictor has proven to already easily outperform the TOBT. One of the key success factors is using data from different airports and airlines to learn the inherent patterns of airport and airline operations. In order to further boost the accuracy of these predictors one can use unambiguous and stakeholder-agnostic real time turnaround data. It is easy to imagine that knowing whether the fueler has been late or whether catering takes longer than usual, might have an impact on the OBT. These two turnaround stakeholders, however, are often not integrated into any system and their current task status is therefore unavailable. Companies like Assaia use computer vision technology and (preferably) existing cameras to capture this kind of data out of video feeds.
At the end of the day, stand optimization is like cooking a recipe. You need a good recipe but maybe even more important is that you have high quality ingredients to achieve the optimal result.
Artificial Intelligence technology has already been around since the 1960s. So why is it that these technologies are such a good fit for the current crisis situation?
First, the technology has matured very rapidly over the last few years. It has clearly stepped outside of the research domain and there are already many applications that we use on a daily basis (Google Maps, Auto Correct, email spam filters, etc). The reason for this development is actually twofold. On one hand, the technology has become better over time in order to be usable in these applications. But maybe even more important, the cost of using AI has dropped significantly. Therefore, it has become available in more and more applications in our daily lives.
The second reason why AI technology should be considered by airports and airlines to emerge as better companies from this crisis, is because it has the ability to provide employees with “superpowers”. Already before the crisis many airports and airlines had a difficult time in finding and retaining highly-skilled and trained staff. After the crisis, many will be in a position where their workforce had to be reduced even further in order to survive. This situation will have a negative impact on an organization’s overall productivity. AI can enhance employees’ productivity by automating time consuming tasks and provide important information that can be used to significantly improve decision-making. Employees with superpowers who are making the best decisions will eventually help airports and airlines to regain passengers’ trust, overcome the current crisis and become more profitable and resilient organizations.
Christiaan Hen has spent his entire career in the aviation industry. He worked for over 8 years at Amsterdam Airport Schiphol in different functions like Airport Development, Capacity Management and Operations Management. The last three years at Schiphol Christiaan was Head of Innovation. During this time he met Swiss startup company Assaia. Discussions about a partnership between the airport and Assaia evolved into a partnership between Assaia and Christiaan personally.
At the end of 2018 Christiaan joined Assaia to become their Chief Customer Officer. In this role he is involved in both business as well as product development. Christiaan uses his extensive knowledge of the aviation industry to further enhance the Apron AI product to make aviation more efficient, safer and more sustainable.
The article was provided by a third party and, as such, the views expressed therein and/or presented are their own and may not represent or reflect the views of ACI, its management, Board, or members. Readers should not act on the basis of any information contained in the blog without referring to applicable laws and regulations and/or without appropriate professional advice.