Team 2053 1of30 Boarding at the Speed of Flight Team 2053 February 12, 2007 Executive Summary After mathematically analyzing the aircraft boarding problem, our modeling group would like to present our conclusions, strategies, and recommendations We examined the mathematical effects of waiting in line to board. sending in different groupings of seat assignments, and the interaction between various components of the boarding process to determine the time required to board an aircraft. We developed a detailed simulation methodology to test our ideas and to quantify the differences between boarding strategies. Our simulation models all of the critical factors at play in a boarding scenario, and is easily modified to support different plane dimensions and interior configurations as well any as- sortment of passenger characteristics depending on average demographics and other statistics. We believe that further collaboration with your company and access to your internal business data would provide us with the capability to more accurately determine results and to tune our parameters specific to your airline Our analysis began by determining what factors impact boarding speed the most across all boarding algorithms. Our conclusions are presented in the list below along with strategies that can be implemented to mitigate their impact Passenger entry speed: The faster passengers enter the plane, the faster boards. This means fight check-in procedure(ticket checking )should be optimized to ensure the correct number of gate agents are present. This is particularly important on large planes with multiple aisles or levels. Flight attendants should be stationed at critical junctions(such as entrances to aisles in a multi-aisle plane) to direct each passenger to the correct row and thereby maintain throughput Baggage stowage time: The faster passengers put their bags away and sit down, the faster the plane boards. The impact of storage time can be mitigated by changing or enforcing carry-on baggage limits and by having fight attendants assist passengers with particularly large bags that they cannot easily lift. Another possibility to consider is a redesign of the overhead bins to make them more easy to load For airlines interested in further decreasing average boarding time we have fur
Team 2053 1 of 30 Boarding at the Speed of Flight Team 2053 February 12, 2007 Executive Summary After mathematically analyzing the aircraft boarding problem, our modeling group would like to present our conclusions, strategies, and recommendations to the airline industry. We examined the mathematical effects of waiting in line to board, sending in different groupings of seat assignments, and the interaction between various components of the boarding process to determine the time required to board an aircraft. We developed a detailed simulation methodology to test our ideas and to quantify the differences between boarding strategies. Our simulation models all of the critical factors at play in a boarding scenario, and is easily modified to support different plane dimensions and interior configurations as well any assortment of passenger characteristics depending on average demographics and other statistics. We believe that further collaboration with your company and access to your internal business data would provide us with the capability to more accurately determine results and to tune our parameters specific to your airline. Our analysis began by determining what factors impact boarding speed the most across all boarding algorithms. Our conclusions are presented in the list below along with strategies that can be implemented to mitigate their impact: • Passenger entry speed: The faster passengers enter the plane, the faster it boards. This means flight check-in procedure (ticket checking) should be optimized to ensure the correct number of gate agents are present. This is particularly important on large planes with multiple aisles or levels. Flight attendants should be stationed at critical junctions (such as entrances to aisles in a multi-aisle plane) to direct each passenger to the correct row and thereby maintain throughput. • Baggage stowage time: The faster passengers put their bags away and sit down, the faster the plane boards. The impact of storage time can be mitigated by changing or enforcing carry-on baggage limits and by having flight attendants assist passengers with particularly large bags that they cannot easily lift. Another possibility to consider is a redesign of the overhead bins to make them more easy to load. For airlines interested in further decreasing average boarding time we have fur- 1
Team 2053 ther analyzed the merits of different boarding algorithms. Through our simula tions we have developed a generic classification of boarding methods: Best No assigned seats Better Outside in boarding Mediocre back to front We understand that the proposition of no assigned seats may be problematic from a customer service perspective. If this is the case outside in boarding(win dow seats first, in towards the aisle) provides significant advantages over back to front, particularly when our previously mentioned optimizations are incorpo- rated into the system. The exact numbers depend on the aircraft dimensions and other factors, but in general outside in boarding provides a 10-30% advan- tage over back to front. Similarly, foregoing assigned seats results in a 10-30% age over outside in. We know that for m es, this magnitude of aprovement could provide the margin necessary for an extra run in the course of a day, resulting in additional revenue. However, our analysis does not stop at determining mere speed increases; we also analyzed the reliability of each boarding method in order to determine the deviation between the longest and hortest possible delays for each boarding algorithm. In order to schedule an extra flight, you have to be sure the tightened timetable will always be met, not t most of the time. We found that the faster methods are also considerably more reliable: outside in has a time deviation range 50% smaller than back to front. For more specific numbers, examples on varying sizes of planes, and in general a more complete description of our work, please refer to our in-depth re- ort, attached. With our insights and your business expertise, we can cooperate to benefit the customer, your business, and your shareholders
Team 2053 2 of 30 ther analyzed the merits of different boarding algorithms. Through our simulations we have developed a generic classification of boarding methods: • Best No assigned seats • Better Outside in boarding • Mediocre Back to front We understand that the proposition of no assigned seats may be problematic from a customer service perspective. If this is the case outside in boarding (window seats first, in towards the aisle) provides significant advantages over back to front, particularly when our previously mentioned optimizations are incorporated into the system. The exact numbers depend on the aircraft dimensions and other factors, but in general outside in boarding provides a 10-30% advantage over back to front. Similarly, foregoing assigned seats results in a 10-30% advantage over outside in. We know that for many routes, this magnitude of improvement could provide the margin necessary for an extra run in the course of a day, resulting in additional revenue. However, our analysis does not stop at determining mere speed increases; we also analyzed the reliability of each boarding method in order to determine the deviation between the longest and shortest possible delays for each boarding algorithm. In order to schedule an extra flight, you have to be sure the tightened timetable will always be met, not just most of the time. We found that the faster methods are also considerably more reliable: outside in has a time deviation range 50% smaller than back to front. For more specific numbers, examples on varying sizes of planes, and in general a more complete description of our work, please refer to our in-depth report, attached. With our insights and your business expertise, we can cooperate to benefit the customer, your business, and your shareholders. 2
Team 2053 3of30 1 ntroduction Short of a single minor detail the airplane boarding problem would be easily solved using a very simple algorithm. Given his performance in the summer blockbuster" Snakes on a Plane we know Samuel L. Jackson is an optima de-boarder of snakes from planes [1]. Assuming that he maintains equal effe tiveness with people, simply invert his role and you have an optimal passenger oarding algorithm. We could then simply model people as snakes and play the film in reverse and determine the effective boarding time. The only potential challenge would be scaling our results from the boeing 747 used in the movie to planes of varying sizes. Ignoring the only detail that there is only one Samuel L Jackson(maybe cloning could help here), the idea of an airplane boarding problem is still an ambiguous oncept. After all, people want to board the plane quickly and the geometry of the plane is fixed. How much is there to modify that could potentially lead to any speedup in boarding time? Upon first observation, it is not obvious the true multitude of factors that mesh to determine airplane boarding time. However, after a closer look the true num- ber of degrees of freedom appear. Then the problem becomes one of determining ich factors significantly contribute to the problem. The complexity required for this analysis is daunting and in many cases the problem would be relegated into the category of“ not worth the time.” However, like many problems orphaned into this category, it is often the market economy that comes to the rescue. The competitive nature of the marketplace continually redefines the differential that determines what is within the bounds of a marginal gain. For the airlines this marginal difference of even a few minutes per flight can represent millions of dollars in revenue over a fiscal year. Consid- ering the number of airlines currently operating under bankruptcy protectio with federal subsidies, this is no small matter. It is this demand for revenue that has thrust the airplane boarding problem into the forefront of modern in- dustrial problem capable of being solved with mathematics. With this in mind we embark on our journey to tackle the airplane boarding problem with Samuel L Jackson as our inspiration 2 Problem restatement We start our journey by concretely stating what we wish to examine. We would ke to finish by having an efficient method for boarding a commercial airplane that accommodates for editable human behavior and a framework that allows us to compare and contrast between different procedures. In the process we would like to gain a"deeper"understanding into the fundamental issues of airline boarding, both to understand the reasons why certain procedures behave
Team 2053 3 of 30 1 Introduction Short of a single minor detail the airplane boarding problem would be easily solved using a very simple algorithm. Given his performance in the summer “blockbuster” Snakes on a Plane we know Samuel L. Jackson is an optimal de-boarder of snakes from planes [1]. Assuming that he maintains equal effectiveness with people, simply invert his role and you have an optimal passenger boarding algorithm. We could then simply model people as snakes and play the film in reverse and determine the effective boarding time. The only potential challenge would be scaling our results from the Boeing 747 used in the movie to planes of varying sizes. Ignoring the only detail that there is only one Samuel L. Jackson (maybe cloning could help here), the idea of an airplane boarding problem is still an ambiguous concept. After all, people want to board the plane quickly and the geometry of the plane is fixed. How much is there to modify that could potentially lead to any speedup in boarding time? Upon first observation, it is not obvious the true multitude of factors that mesh to determine airplane boarding time. However, after a closer look the true number of degrees of freedom appear. Then the problem becomes one of determining which factors significantly contribute to the problem. The complexity required for this analysis is daunting and in many cases the problem would be relegated into the category of “not worth the time.” However, like many problems orphaned into this category, it is often the market economy that comes to the rescue. The competitive nature of the marketplace continually redefines the differential that determines what is within the bounds of a marginal gain. For the airlines this marginal difference of even a few minutes per flight can represent millions of dollars in revenue over a fiscal year. Considering the number of airlines currently operating under bankruptcy protection with federal subsidies, this is no small matter. It is this demand for revenue that has thrust the airplane boarding problem into the forefront of modern industrial problem capable of being solved with mathematics. With this in mind we embark on our journey to tackle the airplane boarding problem with Samuel L. Jackson as our inspiration. 2 Problem Restatement We start our journey by concretely stating what we wish to examine. We would like to finish by having an efficient method for boarding a commercial airplane that accommodates for unpredictable human behavior and a framework that allows us to compare and contrast between different procedures. In the process, we would like to gain a “deeper” understanding into the fundamental issues of airline boarding, both to understand the reasons why certain procedures behave 3
Team 2053 differently, but also to make well-informed and theoretically-justified recommen- dations to our industry patrons We approached the problem by first mathematically analyzing different factors that contribute to delays in airplane boarding. Mathematical analysis of blocks which prohibit smooth flow was carried out using techniques from stochastic processes We also developed a computer simulation which modeled the airplane boarding process while accounting for different boarding methods and individual variation of passengers. We also used our simulation to learn an ideal boarding proce- dure, which we refer to as Parabola for the parabola-like zone assignments that it uses. We then pitted our boarding scheme against other standard boarding schemes to see how it fared 3 Conventions This section defines the basic terms used in this paper 3.1 Terminology Passenger: A passenger is an individual traveling on the plane who is not part of the crew Boarding Scheme: a boarding scheme refers to an assignments of zones or groups according to which passengers board the plane. Depending on the modeling options, it could be exactly deterministic or the general assignment before random mixings Interference An interference is an in which a passenger cannot progress towards their seat because ther passenger blocking their way. Variables efine the following variables here as they are used widely throughout Additional variables may be defined later, but will be confined to particular section C refers to the number of columns in the plane which is also the number of seats in a row R refers to the number of rows in the plane. For the most part we ignore or treat in a different manner distinctions between classes. for details see section 4
Team 2053 4 of 30 differently, but also to make well-informed and theoretically-justified recommendations to our industry patrons. We approached the problem by first mathematically analyzing different factors that contribute to delays in airplane boarding. Mathematical analysis of blocks which prohibit smooth flow was carried out using techniques from stochastic processes. We also developed a computer simulation which modeled the airplane boarding process while accounting for different boarding methods and individual variation of passengers. We also used our simulation to learn an ideal boarding procedure, which we refer to as Parabola for the parabola-like zone assignments that it uses. We then pitted our boarding scheme against other standard boarding schemes to see how it fared. 3 Conventions This section defines the basic terms used in this paper. 3.1 Terminology • Passenger: A passenger is an individual traveling on the plane who is not part of the crew. • Boarding Scheme: A boarding scheme refers to an assignments of zones or groups according to which passengers board the plane. Depending on the modeling assumptions, it could be exactly deterministic or the general assignment before random mixings. • Interference: An interference is an event in which a passenger cannot progress towards their seat because of another passenger blocking their way. 3.2 Variables We will define the following variables here as they are used widely throughout our paper. Additional variables may be defined later, but will be confined to a particular section. • C refers to the number of columns in the plane which is also the number of seats in a row. • R refers to the number of rows in the plane. For the most part we ignore or treat in a different manner distinctions between classes, for details see section 4. 4
Team 2053 5of30 B refers to the time it takes for a person to stow their baggage into the overhead bin. B is assumed to be constant for our preliminary mathemat- ical analysis; it is allowed to change in the simulation. Refer to section 7.5 for more information about variation of B v refers to the walking speed of the passengers. It is assumed to be constant throughout the model. See section 4 for an explanation. s refers to the time it takes for an already seated passenger to get up and get out to let another passenger pass. It is assumed to be constant for nathematical analysis but is allowed to vary in the simulation A refers to the rate at which people enter the plane through the main door This value is assumed to be constant as any deviations in time between passengers is mitigated by walking down the jet-bridge to the plane 4 Assumptions We make the following assumptions about airplane boarding process in this Passengers with physical limitations, families with infants, and passengers extremely advanced in years board the plane before other passengers for their own safety and comfort. We assume that these passengers might perhaps with the assistance of fight attendants. The time taken for this pre-boarding is assumed to be a constant overhead that airlines cannot First class passengers are boarded separately. The existence of a first class in our view means that they require first class treatment: a first class section where passengers have to fight through the proletarian masses is antithetical to the very idea of a"first"class. We can either assume a single-class plane, or model the first class separately( see section 11) All passengers boarding the plane during general boarding walk at ap- proximately the same speed. Since we assume passengers of extremely limited mobility are already aboard the plane, this is plausible. Further more, the walking speed is limited more by the environment (aisle size people in the way) than the persons innate maximum physical capacity Passengers board independently and walk independently, that is, we have no groups waiting for each other or slowing in line. For families we might ssume that they are assigned seats next to each other, which satisfies their bonding and closeness desires We confine our analysis to the interior of the plane. That is, we ignore ter minal effects(anything outside the plane door)) beyond requiring that the gate agents can supply us with passengers at a certain(typically constant
Team 2053 5 of 30 • B refers to the time it takes for a person to stow their baggage into the overhead bin. B is assumed to be constant for our preliminary mathematical analysis; it is allowed to change in the simulation. Refer to section 7.5 for more information about variation of B. • v refers to the walking speed of the passengers. It is assumed to be constant throughout the model. See section 4 for an explanation. • s refers to the time it takes for an already seated passenger to get up and get out to let another passenger pass. It is assumed to be constant for mathematical analysis but is allowed to vary in the simulation. • λ refers to the rate at which people enter the plane through the main door. This value is assumed to be constant as any deviations in time between passengers is mitigated by walking down the jet-bridge to the plane. 4 Assumptions We make the following assumptions about airplane boarding process in this paper. • Passengers with physical limitations, families with infants, and passengers extremely advanced in years board the plane before other passengers for their own safety and comfort. We assume that these passengers might need the plane to be relatively empty to successfully reach their seat, perhaps with the assistance of flight attendants. The time taken for this pre-boarding is assumed to be a constant overhead that airlines cannot avoid. • First class passengers are boarded separately. The existence of a first class in our view means that they require first class treatment: a first class section where passengers have to fight through the proletarian masses is antithetical to the very idea of a “first” class. We can either assume a single-class plane, or model the first class separately (see section 11). • All passengers boarding the plane during general boarding walk at approximately the same speed. Since we assume passengers of extremely limited mobility are already aboard the plane, this is plausible. Furthermore, the walking speed is limited more by the environment (aisle size, people in the way) than the person’s innate maximum physical capacity. Passengers board independently and walk independently, that is, we have no groups waiting for each other or slowing in line. For families we might assume that they are assigned seats next to each other, which satisfies their bonding and closeness desires. • We confine our analysis to the interior of the plane. That is, we ignore terminal effects (anything outside the plane door)) beyond requiring that the gate agents can supply us with passengers at a certain (typically constant) 5