Measurement of Driving Conditions of Aircraft Ground Support Equipment at Tokyo International Airport

Abstract

With the global increase in air transport demand, the shortage of ground handling personnel to support ground operations at airports has become a major challenge, impacting airport services and causing considerable flight delays. This study presents a novel method to generate trip data that specify the origin and destination locations as the purpose of travel for each ground support equipment (GSE) vehicle. The proposed method uses data obtained from comprehensive observations of 2234 GSE vehicles over a 24 h × 7 d time interval at Tokyo International Airport. From these observations and trip data, the characteristics of the driving conditions for each GSE vehicle type, the locations where GSE traffic volume increases in the airport, and changes in the time interval are identified. The primary results show that the GSE traffic volume is the highest mainly around passenger terminals and in the vehicle corridors connecting these terminals, which aligns with the airport’s operational status. Investigating GSE driving conditions, such as the traffic flow throughout an airport, can provide valuable data to improve the efficiency of GSE scheduling and facilitate the introduction of automated driving technology.

1. Introduction

The recent surge in global aviation demand—projected to reach approximately 6 billion passengers by 2030 [1] and 10 billion by 2050 [2]—has led to congested airports and significant air travel delays. For instance, the average flight delay in Europe rose to 17.3 min in 2022, up from 9.2 min in 2021 [3]. Moreover, the number of passengers affected by air transport delays exceeding 120 min increased from approximately 4 million to over 14 million between 2010 and 2019 [4,5]. It is estimated that 5–10% of major delays can be attributed to ground handling (GH) [6], underscoring its critical impact on airport performance [7]. Hence, airports must optimize their resources and enhance the efficiency of airport ground support operations to foster sustainable air transportation [8,9].

The GH personnel responsible for airport ground support operations must receive comprehensive training not only in task-specific skills, but also in general airport traffic safety awareness [10]. However, challenges such as labor market shortages, high turnover rates, and frequent employee transfers [10] make it difficult for global GH companies to hire additional personnel at competitive wages, leading to increased costs and delays [7]. For example, as of 2016, 1032 companies operated in Poland in the airline industry and employed 14,500 employees. However, despite an increase in the number of companies and organizational turnover, the number of employees has remained at approximately 15,000 [10], suggesting that labor shortages in airport ground support operations may affect the quality of airport ground services. Additionally, the deregulation of the GH market in European airports in 1996 has produced a marked increase in the number of third-party firms [11]. The labor shortage in GH has become an increasingly important issue because of the COVID-19 pandemic. In Japan, GH personnel left the workforce, leading to a sharp decline in demand for the aviation industry during lockdown. However, the demand for the aviation industry has recovered since the pandemic, and it is crucial for the GH companies to secure the required human resources [12]. Consequently, with multiple ground crews providing multiple services, the efficient scheduling of ground activities has become increasingly important [13]. Thus, there is an immediate need for labor efficiency and increased automation in airport ground support operations, where labor shortages are becoming more acute. With reliable online data on GH service demand and supply, the automation of GH operations and ground support equipment (GSE) is being explored [14]. Autonomous vehicles are already being implemented in airports worldwide [15], with some airports aiming to fully automate their GSE operations [6,16].

In Japan, amidst increasing air transport demand and challenges such as competition from neighboring overseas airports, security threats, and labor shortages owing to a declining working-age population, the government and private sector are collaborating to provide passengers with world-class services. This effort involves promoting innovation in the air transportation industry through advanced technologies and systems such as automation, robotics, biometrics, artificial intelligence (AI), the Internet of Things, and big data [17]. As part of this initiative, there are plans to introduce automated driving technology for the GSE used in airport ground support operations, facilitated by public–private partnerships. The government led the first field operational test of autonomous GSE in an airport’s restricted areas in June 2018 [18]. Subsequently, in June 2019, efforts towards developing common airport infrastructure and operational rules for airport operators were undertaken alongside ongoing operational tests. Understanding the impact of autonomous GSE on the traffic flow within airport vehicle corridors and validating the effectiveness of common infrastructure and operational rules are crucial aspects of studying and implementing these advancements.

Consequently, determining the driving conditions for individual GSE vehicles based on their trip data, which specify their origin and destination (OD) locations within the airport, is essential for the analysis of GH efficiency and automation methods. This is crucial for optimizing GSE deployment, planning travel routes, and enhancing GH management practices, which includes improving schedules, implementing automated driving technology, and reducing environmental impacts.

Although previous studies have addressed the efficient scheduling of GSE [9,13] and GSE travel management systems [7], there remains a gap in the research regarding comprehensive tracking methods for individual GSE vehicles within airports’ restricted areas and understanding their traffic flow across an entire airport through organized trip data, which specify OD locations. Each GH company requires visibility regarding the whereabouts and travel routes of its GSE vehicles, and may use Global Positioning System (GPS) transmitters for some vehicles. However, at airports with several hundred GSE vehicles per company, equipping all vehicles with GPS transmitters is impractical. Although each GH company has a GSE deployment plan according to flight schedules, GSE vehicles do not always run according to plan, but respond flexibly to situations, especially at large airports such as metropolitan airports, because sudden changes in flight schedules and traffic congestion occur frequently. For this reason, each GH company does not have a complete understanding of all the GSE vehicles it owns, and there are unknown aspects of the driving conditions of these GSE vehicles. Furthermore, each company does not know the locations and movements of the GSE vehicles of other companies, and no one gathers information or has a comprehensive understanding of all GSE vehicles traveling at an airport. In recent years, a fleet management system that shares and centrally manages GSE traffic information has been anticipated in Japan for the introduction of automated driving technology at airports. However, numerous challenges remain in information collection and sharing. Furthermore, GSE movement involves not only simple round trips between specific points, but also sequences of consecutive trips from one destination to another, especially in large airports. Therefore, observed GSE movement data must be structured into trip data that identify OD locations in order to accurately assess GSE driving conditions.

GPS is a precise positioning technology. It helps in analyzing the actual vehicle driving conditions or road traffic demand using GPS transmitters installed in vehicles. However, most GSE vehicles at airports do not have GPS transmitters installed, and considerable effort would be required for research on using GPS. Therefore, there are no existing studies that involve the installation of GPS transmitters on all GSE vehicles at an airport to determine the actual driving conditions. RFID and AI-based visual detection are other technologies that can be used to monitor vehicle locations. RFID is widely implemented in the fields of transportation and logistics, as it can read information from several meters away via wireless communication. RFID markers are embedded into the road surfaces of some vehicle corridors at an airport in Japan and are used for the self-position recognition of automated vehicles [19]. The installation of RFID markers is limited to certain areas at the airport owing to the maintenance of vehicle corridors and foreign object debris (FOD) concerns. A previous study [20] involved guiding unmanned vehicle movements by setting barcode markers on the apron surface, but within a limited test field instead of across the entire airport. Conversely, in other studies [21,22], aircraft and GSE vehicles were monitored on the apron through AI-based visual detection. In another study [23], images captured from a running GSE vehicle were used for monitoring. This monitoring was only conducted in a limited area within the apron or to an extent that could be obtained from a specific GSE. Extensive research has yet to be conducted on AI-based visual detection, particularly on the airside of an airport. This is because installing several cameras and monitoring GSE vehicles throughout an airport, particularly a big one, faces limitations in terms of technical requirements and cost. As described in previous studies [22,23,24], the airside area is such a secured area that tracking GSE vehicles were focused on specific targets, and it was difficult to accumulate a dataset throughout an entire airport.

Consequently, this study aims to propose and demonstrate a method to generate trip data that specify the OD of each GSE as the purpose of travel to obtain a comprehensive understanding of the driving conditions of individual GSE vehicles. Data corresponding to the driving conditions of GSE vehicles, such as the traffic flow within the airport, will contribute to other studies conducted on efficient GSE scheduling, the introduction of automated driving technology, and a reduction in the environmental impact of GH, such as fuel consumption.

We used data obtained from observations of more than 2000 GSE vehicles over a seven-day period at Tokyo International Airport (Haneda Airport), the busiest airport in Japan. To address the high costs of equipping thousands of vehicles with standard GPS transmitters and the technical issues involving RFID and AI-based visual detection, we used Bluetooth transmitters on a trial basis, because they are relatively inexpensive and easy to attach to GSE vehicles. Similar to RFID, Bluetooth uses wireless communication to detect vehicles; however, its detection range is wider than that of RFID by several tens of meters. Therefore, it was relatively easy to obtain permission from the airport authorities to install Bluetooth transmitters beside the vehicle corridors and to monitor the GSE vehicles throughout the airport. Based on the observation data and trip data, we comprehensively confirmed the driving conditions of each GSE vehicle, including the time, location, and route traveled within the airport. Additionally, we quantitatively analyzed characteristics such as the frequency of movement (number of moving trips and stopping time per trip) and the distance/time traveled per trip for each GSE vehicle type. Furthermore, we observed overarching time interval trends in the GSE traffic volume within the entire airport’s restricted area. The validity of the proposed method was verified from the perspective of the observation accuracy of the transmitters attached to the GSE vehicles.

The novelty of this study lies in presenting a method to empirically understand the driving conditions of GSE vehicles within airports’ restricted areas, which have yet to be sufficiently investigated. By comprehensively observing the movement of each GSE vehicle in an inexpensive and straightforward manner and organizing the traffic flow of GSE vehicles into trip data, we provide valuable insights.

The remainder of this paper is organized as follows. Section 2 summarizes previous studies. Section 3 describes the GSE survey and the analytical method. Section 4 presents the results of the GSE survey. Section 5 discusses these results. Finally, Section 6 concludes the paper.

2. Literature Review

GH operations are the airside operations of an airport, involving the handling of passengers and cargo, as well as the management of facilities and supplies in and around parked aircraft [7,25]. Most of these operations are performed by a variety of service providers and involve the extensive use of specialized vehicles and equipment known as GSE, the management of which is central to GH operations [7]. GH comprises the activities, operating procedures, equipment requirements, and personnel necessary to prepare an aircraft for its next flight [13]. Consequently, any individual operation can be a potential source of delay, which can impact other ground operations and airport processes owing to the inherent interdependence among these operations [13].

Research on GH operations has been conducted from several perspectives. One perspective is investigating the sequence of ground service processes, from aircraft arrival to departure, to effectively manage milestones. This approach aims to reduce flight delays and enhance airport operational capacity [26]. Owing to increasingly high-density operations, any small interference could lead to flight delays [27], making the accurate management of milestones a considerable challenge [26]. With the increasing use of AI technology, numerous deep learning models have been developed [28,29]; however, none of the existing prediction methods can handle missing values [26].

The second perspective is that developing a system for monitoring GSE and aircraft movements can support efficient GH operations. Drawing from the concept of airport collaborative decision making (ACDM), methods for enablinb the online allocation of GSE fleets by exchanging information with ACDM systems at the operational level have been studied [7]. ACDM was a concept introduced by Eurocontrol [30], emphasizing cooperation and real-time data sharing among airport operators, aircraft operators, GH operators, and air traffic control [9]. Conversely, a previous study also employed neural networks and camera calibration techniques to detect when an aircraft landed on the runway, tracked its movement to the apron and gate, monitored the turnaround process in the apron area, and predicted the pushback time [24]. Another study [23] detected aircraft and estimated the aircraft movement context by using images captured by a running GSE vehicle. Additionally, studies conducted in the apron area utilized deep learning and computer vision methods to detect and analyze the key ground service behaviors of aircraft parked on an apron from a single fixed camera image [21,22]. Some studies [31,32,33] have involved improving the management of apron movement control through the observation of GH in the apron.

Moreover, research on legal systems, such as the deregulation of and discretionary authority for GH operations [34], and systems and methods to evaluate GH performance [35] has also been conducted.

However, there has been insufficient research on empirically understanding the driving conditions of GSE vehicles in airports’ restricted areas by comprehensively measuring individual GSE movements and organizing the traffic flow into trip data that specify OD locations as the purpose of travel. The empirical data obtained in this study are expected to serve as foundational data for various issues noted in previous studies.

3. Methodology

3.1. Survey Scope and Data Collection

Table 1. Summary of surveys based on driving conditions.

Location	Haneda Airport’s restricted area
Implementation period	21–27 November 2019: 24 h × 7 d
Vehicles surveyed	All self-propelled GSE (except for some vehicle types such as forklifts and lighting vehicles), details are listed in Appendix A Figure A1
Companies surveyed	22
Transmitters attached	2234 vehicles (approximately 74% of the vehicles surveyed)
Receivers installed	53 locations
Data obtained	Vehicle ID, passing position, and detection time data for GSE (hereinafter referred to as detection record data)

summarizes the GSE driving conditions surveys. Haneda Airport, the busiest airport in Japan, with an annual passenger volume of 87.41 million and a domestic market share of 26.3% (as per 2019 data) [36], is among the airports conducting field operational tests for GSE automation to enhance GH efficiency. Therefore, Haneda Airport was selected as this study’s focus. The study was conducted in late November 2019, because Haneda Airport has recorded one of the highest passenger volumes in the world over the last 20 years, except for the period from 2020 to 2022, owing to the COVID-19 pandemic [37]. The airport is extremely busy throughout the year, and there are no significant differences in the numbers of departures and arrivals per month [36].

A diversity of GSE vehicles is essential for aircraft operations, with each type performing specific ground support roles at airports. At Haneda Airport, approximately 7000 GSE vehicles are registered, with around 40% being self-propelled (as of June 2019). The remainder are non-self-propelled vehicles towed by self-propelled GSE vehicles. In this study, GSE vehicles towing air cargo or passenger baggage containers are referred to as towing tractors (TT), while those towing aircraft are called aircraft-towing vehicles. The survey included all types of self-propelled GSE vehicles registered to operate in Haneda Airport’s restricted area, excluding vehicles such as forklifts primarily used in cargo sheds and lighting vehicles used exclusively at night.

The actual driving conditions were surveyed using a transmitter attached to a GSE vehicle (beacon) and a receiver (smartphone) installed near a vehicle corridor in Haneda Airport’s restricted area. This setup recorded data such as the vehicle ID, passing position, and detection time for each GSE (referred to as detection record data). When a GSE equipped with a beacon passed by a receiver, the receiver detected the Bluetooth low-energy (BLE) signal emitted by the beacon, recording the vehicle ID and the time of detection. Beacons are cost-effective compared to standard GPS transmitters and are relatively easy to attach to GSE. BLE signals can be detected by receivers within a maximum range of 30 m, allowing for the capture of approaching GSE IDs within an approximately 30 m radius around each receiver. By consolidating these data with location information from the receivers, a detection record was compiled, detailing the vehicle ID, passing position, and detection time for each GSE vehicle. We used Bluetooth technology on a trial basis; however, this method can be replaced if other technology can acquire the vehicle IDs, passing position, and time data for each GSE vehicle equivalent to that of Bluetooth. Twenty-two companies that owned GSE vehicles were asked to cooperate in the survey, and beacons were attached to each of the 2234 GSE vehicles. The companies surveyed included airport managers, airlines, GH operators, catering companies, and refueling companies.

In this study, Haneda Airport’s restricted area was divided into four areas—Terminal 1 (including the West Domestic Cargo Area), Terminal 2 (including the East Domestic Cargo Area), Terminal 3 (including the International Cargo Area), and the Maintenance Area—as depicted in Figure 1. Receivers were strategically placed at 53 locations near major vehicle corridors and intersections to cover the main OD locations and travel routes of the GSE vehicles across nearly all restricted areas. This setup ensures that the movement of GSE and aircraft remains smooth and safe without impediment. The locations of these receivers are illustrated in Figure 1.

The number of attached beacons and detection records obtained for each vehicle type are listed in

Table A1. Number of attached beacons and number of detection record data.

Vehicle Type Classification for Analysis	Number of Vehicles (Units)	Number of Detections (Cases)	Average Number of Detections (Cases/Unit)
Towing tractors	455	694,205	1526
Cargo vehicles	354	555,813	1570
Maintenance-related vehicles	353	312,618	886
Liaison vehicles	295	495,383	1679
Handling vehicles	156	279,791	1794
Refueling vehicles	135	189,303	1402
Catering vehicles	116	79,318	684
Aircraft towing vehicles	105	165,902	1580
Passenger transport buses	71	105,903	1492
Airport maintenance related vehicles	62	46,258	746
Passenger step vehicles	53	26,995	509
Non-handling freight transport vehicles	40	2507	63
Other vehicles	39	57,715	1480
Total of all vehicles	2234	3,011,711	1348

of Appendix A.

3.2. Trip Data Creation Method

The obtained detection data were arranged chronologically to construct continuous travel trajectories for each GSE vehicle, depicting the movement from one destination point to the next in the sequence. These trajectories were further segmented into individual trips, each representing a single unit of travel from an origin to a corresponding destination point. When generating the trip data, Haneda Airport’s traffic rules governing the allowable routes, directions, and speed limits for each vehicle type were meticulously considered. To accurately reflect the total traffic volume at Haneda Airport, an expanded estimation of trip data was conducted, incorporating direct measurements from video recordings. The methodology is outlined below.

3.2.1. Extraction of Target Trips

The average speed ($V$) of the GSE trips between points can be calculated using the detection time from the detection record data and distance traveled between the detection points (Equation (1)). The trip distance ($l$) is the shortest vehicle path between two points (Equation (2)).

$$V={\displaystyle \frac{l}{t}}$$

(1)

$$l=\mathit{min}\left(L\right),\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}l\in L$$

(2)

$$t=\left|T_1-T_2\right|$$

(3)

where $L$ denotes the set of the trip distance between the two points; $t$ denotes the traveling time between the two points; and $T_1$ and $T_2$ denote the detection times at Points 1 and 2, respectively. The calculated average trip speeds for each vehicle type can be presented as a histogram; a histogram of vehicle type T is shown in Figure 2.

Here, T are passenger transport buses (BUS). The speed at the minimum valley of the histogram was set as the lower limit ($V_{min}^T$) for vehicle type T, with trips falling below the lower limit being set as “stopped trips”. Because each GSE vehicle spends a relatively long time traveling at low speeds (≈stop) or high speeds (≈move) and a relatively short time traveling at the transition between low and high speeds, the transition speed can be assumed to be the minimum valley in the histogram of the average speed distribution. The lower limit ($V_{min}^{BUS}$) for the BUS in Figure 2 is 5 km/h. An upper limit ($V_{max}^T$) for vehicle type T was set with reference to the speed limit for vehicle corridors in Haneda Airport’s restricted areas. Trips in which the average speed exceeded the upper limit were excluded from the trip data as “incomplete data acquisition”. The upper limit ($V_{max}^{BUS}$) for the BUS in Figure 2 is 60 km/h, which is twice the speed limit in the restricted areas. In other words, the set of trips for which the average speed was above $V_{min}^T$ but below $V_{max}^T$ was a moving trip ($A^T$) for vehicle type T (Equation (4)), which is the subject of this analysis. The set of trips for which the average speed was above 5 km/h but below 60 km/h was the moving trip ($A^{BUS}$) for the BUS in Figure 2. We considered the running speeds as the reference values in this study, owing to the deviation reported in Section 4.3. To analyze a large number of trips, we set the upper speed limit for moving trips to twice the official speed limit (60 km/h for BUS) on a trial basis, considering the distribution of the calculated running speeds, without using the official speed limit of the airport.

Similarly, each moving trip could be organized for each vehicle type. It should be noted that $V$ calculated using the aforementioned method (Figure 2) is not the average of the actual running speed of each GSE vehicle, as there is always the possibility of discrepancies between the time of detection record and the actual time of the GSE passage.

$$A^T=\left\{X^T|V_{min}^T<V^T<V_{max}^T\right\},\hspace{1em}\hspace{1em}\hspace{1em}{A}^T\in X^T$$

(4)

where $X^T$ denotes the set of all trips of vehicle type T and $V^T$ denotes the average speed for vehicle type T.

Considering scenarios where GSE vehicles may temporarily exit the surveyed area (Section 3.1) or where GSE vehicles operating outside the surveyed area may be detected depending on the receiver placement, trips with more than three consecutive missing points were excluded from the trip data as “leaving” or “survey error”. However, trips with one or two adjacent missing points were supplemented using the most frequent routes, ensuring that these trips were classified as valid “moving trips”. Figure 3 outlines the rules applied for analyzing the moving trips in this study.

3.2.2. Turnaround Trips

A moving trip that returns along the same route can be considered to have a destination around the turnaround point, irrespective of the stop time at that point. The moving trip can then be segmented at the turnaround point. Figure 4 illustrates an example of how a moving trip is divided in this manner.

3.2.3. Expansion Estimates of Moving Trips

To accurately reflect the total traffic volume at Haneda Airport, an expanded estimation of moving trips was conducted. This estimation compared the actual number of GSE vehicles recorded by a fixed-point video camera in the section from location ID No. 20 to Nos. 8 and 9 (as shown in Figure 1) with the number of GSE vehicles obtained from the detection record data acquired (Section 3.1). The results of the comparison between the two numbers of GSE vehicles are listed in

Table 2. Calculation of expansion coefficients for moving trips.

Classification of Vehicle Type	Number of GSE Vehicles Based on Detection Record Data (a)	Number of GSE Vehicles Recorded by Fixed-Point Camera (b)	(b)/(a)
Special heavy-duty vehicles	7	15	2.14
TT *	559	837	1.50
BUS/MB	96	112	1.17
Other vehicles	348	147	0.42
Total	1010	1111	1.1

Note: Special heavy-duty vehicles include aircraft-towing vehicles (WT), high-lift loaders (HL), belt loaders (BL), and passenger step vehicles (PS) in Appendix A Figure A1. TT * refers to TT other than towing tractors for maintenance.

. Precisely classifying the vehicle types was impossible, as distinguishing between them using the fixed-point camera recordings was too difficult. The ratio of the detection record data to the fixed-point camera recordings varied for each vehicle type (Table 2). Consequently, setting an expansion factor for each vehicle type was challenging, and the expansion factor of 1.1 for moving trips was set based on the ratio of the total values across all vehicle types. The moving trips to be added using the expanded estimates were randomly extracted from ${OD}_{day\_n}$ (the OD table of moving trips between the 53 locations shown in Figure 1) for each vehicle type category and replicated. The trips added to ${OD}_{day\_n}$ were extracted from ${OD}_{day\_(n-1)}$ by shifting one day back; however, ${OD}_{day\_7}$ was shifted from the data of 27 November 2019 and added to ${OD}_{day\_1}$ of 21 November 2019. This is because the same moving trips were not generated simultaneously. ${OD’}_{day\_n}$ (the OD table of moving trips between the 53 locations after expansion) can be expressed as shown in Equations (5) and (6).

$${OD}_{day\_n}^{’} = {OD}_{day\_n} + 0.1 \times {OD}_{day\_(n - 1)}\hspace{1em}\hspace{1em}\hspace{1em}\left(n = 2, 3, 4, 5, 6, 7\right)$$

(5)

$${OD}_{day\_n}^{’} = {OD}_{day\_n} + 0.1 \times {OD}_{day\_7}\hspace{1em}\hspace{1em}\hspace{1em}\left(n = 1\right)$$

(6)

$$n = date - 20,\hspace{1em}\hspace{1em}\left(date = 21, 22, 23, 24, 25, 26, 27, \right)$$

(7)

where ${OD’}_{day\_n}$ refers to the OD table of moving trips between the 53 locations after expansion on the n^th day, ${OD}_{day\_n}$ refers to the OD table of moving trips between the 53 locations on the n^th day, $0.1\times {OD}_{day\_(n-1)}$ is randomly extracted by 0.1 from ${OD}_{day\_(n-1)}$, and $0.1\times {OD}_{day\_7}$ is randomly extracted by 0.1 from ${OD}_{day\_7}$.

The moving trips included in the expansion estimates were randomly extracted from discrete OD trips and may not represent the actual service paths of the vehicles. The service paths of the vehicles that could not be observed in this study are unknown and cannot be considered equivalent to the observed service paths. Further research must be conducted on discrete ODs, as they were included in this study on a trial basis.

4. Results

4.1. Results of the Detection Record Data

Through a survey of actual driving conditions (Section 3.1), 3,011,711 detection records were obtained for 1856 vehicles (83.1% of the 2234 vehicles with beacons attached) over a seven-day period. The distribution of detection records by vehicle type is detailed in Table A1 of Appendix A. As an example of the GSE vehicle travel patterns, Figure 5 shows plots of detection points in a time interval for various GSE types, as follows: (a) towing tractor, (b) food loader, (c) sewage truck, and (d) aircraft-towing vehicle. The vertical axis represents location IDs (refer to Figure 1), while the horizontal axis shows the date and time over the seven survey days. Different colors represent the following four areas from Figure 1: Terminal 1 (location ID No. 1–10) in red, Terminal 2 (location ID No. 11–27) in blue, Terminal 3 (location ID No. 28–42) in green, and the maintenance area (location ID No. 43–51) in yellow. Location IDs 52 and 53 denote points on the two vehicle corridors connecting Terminal 3 with other areas. The data illustrated in Figure 5 demonstrate that each of the four types of vehicles primarily travels within specific areas and infrequently moves to or from other areas. The sewage truck primarily travels within Terminal 2, although it also travels in Terminal 3 occasionally. Additionally, the sewage truck is detected several times on location IDs 8 and 9. The sewage truck does not operate in Terminal 1, and instead passes along the route to Terminal 3 or beyond toward an exit outside the airport’s restricted area.

4.2. Results of Trip Data Organization

The trip data from the seven-day survey period (Section 3.1) were generated using the method outlined in Section 3.2. The data included 194,922 moving trips (41.0%), 240,340 stopped trips (50.5%), 33,439 incomplete data acquisitions (7.0%), 3091 survey errors (0.6%), and 4198 leavings (0.9%) for a total of 475,990 trips. Of the detection record data obtained, 91.5% (41.0 + 50.5%) for moving trips and stopped trips could be categorized as valid. For moving trips, aggregate results for each vehicle type are presented in Figure 6 for the following: (a) the total number of moving trips, (b) the average travel time per trip, (c) the average distance traveled per trip, and (d) a box-and-whisker diagram of the stopping time per trip. The vehicle type classifications in Figure 6 correspond to the symbols listed in Figure A1 of Appendix A.

In Figure 6a, the total number of moving trips is highest for the TT, more than five times that of the BUS and cargo trucks (CT), which follow closely. This indicates a notably high frequency of TT traffic. Conversely, belt loaders (BL), high-lift loaders (HL), passenger step vehicles (PS), sewage trucks (LS), and water supply trucks (PW) have fewer trips, with PS vehicles having the least.

The average travel time per trip is the longest for BUS, FL, MB, and WT vehicles; moderate for CT, TT, LS, PW, and SC vehicles; and shortest for BL, HL, and PS vehicles, as depicted in Figure 6b. Figure 6c shows that the average travel distance per trip is the longest for BUS, FL, MB, CT, TT, LS, PW, and SC vehicles and shortest for BL, HL, and PS vehicles.

Figure 6d shows that the stopping time per trip exhibits significant dispersion for BL, HL, and PS vehicles, with the maximum values for BL and PS vehicles exceeding 10 h. Conversely, dispersion is minimal for all other vehicle types, with the maximum values being relatively short, less than two and a half hours.

The 194,922 moving trips were organized into an OD table for the 53 locations (Figure 1) and then expanded using the method described in Section 3.2.3 to reflect the total traffic volume at Haneda Airport. The total number of moving trips after the expansion was 194,922 × 1.1 = 214,414. The OD table after the expansion is shown in Appendix A Figure A2. ODs with a higher number of trips are highlighted in red. Higher trip counts along the diagonal of the OD table indicate fewer trips to/from other ODs, suggesting that GSE vehicles predominantly travel within the same area of the airport.

Figure 7 shows the GSE traffic volume after the expansion from 6:00 to 20:59 on 21 November (totaling 15 h) for each vehicle corridor. The traffic volume is notably concentrated around each passenger terminal (Nos. 1, 2, and 3) and the vehicle corridors linking these terminals. The total number of moving trips during this period totaled 23,018.

The traffic volume between location IDs 38 and 53 is considerably high, owing to the concentration of cargo vehicles for domestic and international transfers. This is attributed to the fact that this route is the vehicle corridor that connects the domestic (Terminals 1 and 2) and international (Terminal 3) terminals within the shortest possible time.

4.3. Discrepancy Between the Detection Time and Actual Passage Time of GSE

There may have been discrepancies between the recorded times of GSE passage at each location in the detection record data acquired during this survey and the actual times at which the GSE vehicles passed through those locations. The following factors could account for these discrepancies:

• The physical distance between the receiver’s location and the center of a nearby intersection.
• Individual differences and fluctuations in the BLE signal strength for each beacon.
• The BLE signals are detected at 2 s intervals; however, only detection data with the highest signal strength at 55 s intervals are actually recorded as the detection record data.

To assess discrepancies, we selected four test beacons (transmitters A, B, C, and D) of the same product used in this survey. These were used to measure the differences between the detection time at each receiver-installed location and the actual time the vehicle passed the intersection center point (the deviation time). Figure 8 illustrates the distribution of deviation times, calculated by subtracting the actual passage time from the detection time. Each receiver could detect BLE signals emitted from approaching GSE vehicles within a radius of approximately 30 m. Consequently, deviation times are distributed as both positive and negative values. Of the instances where the test beacons passed near the receiver, 18% went undetected, while 82% were detected. The deviation times ranged from −3 s to +11 s with an average of 2.7 s. The most common deviation times were +2 and +3 s, and 67% of the total deviation times fell within the range from ±0 to +5 s.

Figure 9 illustrates the fluctuations and individual differences in the deviation times for each beacon at various locations. The deviation times for each beacon show uniformity only at three points (location IDs 7, 10, and 20), while deviations between beacons vary by location, highlighting the fluctuations and individual differences among them.

Figure 10 depicts the relationship between deviation time and the distance of the receiver from the intersection center point at each location. Assuming a GSE speed of 15 km/h or 30 km/h, Figure 10 shows the time difference generated as the GSE passes from the receiver installation position to the intersection center point based on their distance. The discrepancy between the detection time and the actual passage time at the intersection center point ranges approximately from 0 to 8 s, derived by subtracting the time difference from each point’s deviation time value.

At location ID No. 37, where the deviation time is notably large, it is possible that the beacon was detected and recorded after passing through the intersection and making a U-turn, as illustrated in Figure 11.

Given this discrepancy, as discussed in Section 3.2.1, the average speed of the GSE calculated from the detection record data can be expected to differ from the actual average running speed of each GSE.

5. Discussion

The aforementioned results quantitatively show that the GSE vehicles traveling in the airport’s restricted area are characterized by their frequency of movement (number of moving trips and stopping time per trip) and the distance traveled per trip, depending on the type of vehicle, and are consistent with the role of each type of vehicle. TT—which transport air cargo between passenger terminals and aircraft spots and between passenger terminals—travel frequently, making a significantly higher number of moving trips and demonstrating shorter stopping times than other vehicle types. The distance traveled per trip was relatively long. CT—which also transport air cargo—do not have as many moving trips as TT; however, their average travel time, distance, and stopping time per trip were similar to those of TT. BUS—which transport passengers between passenger terminals and open spots and between passenger terminals—had shorter stopping times, because BUS with passengers have priority over other vehicles owing to the operation of GH and are likely to travel relatively smoothly without long stoppages, although the average travel time and distance were long. FL vehicles mainly travel between catering factories located outside of the restricted areas and aircraft spots, resulting in longer average travel times and distances per trip. In contrast, BL, HL, and PS vehicles—which often remain on the right-hand side of open spots as a vehicle storage area and work within neighboring spots when an aircraft parks—had shorter average travel times and distances, suggesting that they travel at low speeds within a relatively small area. Additionally, for these types of vehicles, the dispersion of the stopping times per trip and the maximum values were large, exceeding 10 h for BL and PS vehicles. Finally, LS, PW, and SC vehicles had long average travel distances and medium average travel times, suggesting that they travel at high speeds over relatively wide areas.

The time interval detection record data and the OD table of moving trip data showed that each vehicle mainly traveled within the same area (e.g., Terminal 1) and rarely traveled to/from other areas. It was empirically confirmed that each GSE vehicle traveled within its pre-assigned area of responsibility, which is consistent with the GH company’s efforts for efficient operation. The traffic volume of GSE vehicles in each vehicle corridor, calculated from the moving trip data, indicated that the traffic volume is high, mainly around passenger terminal buildings and in vehicle corridors connecting the buildings. The number of moving trips in a time interval by travel start time—which indicates the GSE traffic volume—tended to increase or decrease in tandem with the number of aircraft flights operated at Haneda Airport. Thus, the overarching or time interval trend of GSE traffic within the entire airport’s restricted area was consistent with the airport’s operational status.

However, the observation method of GSE traffic proposed in this study is a simple method using Bluetooth technology and, thus, is prone to some discrepancy in its results. In particular, the running speed of each trip, which can be calculated using Equation (1) (given in Section 3.2.1), is considered to be affected by this observation deviation. In contrast, the points and routes where the GSE vehicles travel are not affected by this observation deviation, and the amount of discrepancy is considered to be limited for the trip data and the traffic volume in each vehicle corridor presented in this study.

6. Conclusions

In this study, we conducted a survey of GSE operations in a restricted area of Haneda Airport to empirically and comprehensively understand GSE driving conditions, including traffic flow, and generate trip data.

This study involves direct observations of GSE movements to understand driving conditions. The detection record data obtained in this study incorporate the effects of airport traffic rules for each vehicle type and operational events such as aircraft maintenance delays or traffic congestion, including taxiing aircraft crossing vehicle corridors. This detection record data, along with the trip data generated from them, could be used for GH management in the future. This includes improving scheduling efficiency, introducing automated driving technology, and reducing environmental impacts, such as fuel consumption, by analyzing differences from the GSE deployment plan. Additionally, this data can serve as inputs for traffic simulations. For example, traffic simulations based on these detection record data and trip data could be used to understand the impact of introducing autonomous GSE vehicles on the traffic flow in an airport’s vehicle corridors. A preliminary evaluation of the impact of introducing autonomous GSE vehicles on the traffic flow within the airport could be performed, and the measures required at the airport could be identified without conducting large-scale field operational tests. These measures could include the development of common infrastructure and operational rules for autonomous GSE vehicles, for example, traffic signals and priority traffic control at intersections.

However, the detection record data and trip data obtained in this study were acquired during a survey, and the actual GSE traffic flow could vary owing to changes in airport conditions, such as modifications to flight schedules and the expansion of airport facilities. For instance, at Haneda Airport—focused on in this study—the volume and routes of GSE traffic are expected to change owing to increased flight numbers, alterations in flight schedules accompanying airport capacity expansion, the introduction of the South Tunnel, and the enlargement of GSE parking areas. While it would be ideal to gather new data on GSE traffic volume and routes through travel surveys each time airport conditions change, this is impractical, because changes can occur frequently. Therefore, developing a method to estimate the changes in trip data resulting from shifts in flight schedules and airport facilities has become crucial. Estimating changes in trip data would enable more generalized use of this information.

We consider the running speeds calculated in this study as the reference values based on the discrepancies between the calculated and actual speeds reported in Section 4.3. We set the lower and upper speed limits for moving trips on a trial basis to analyze a large number of trips. These speed limits must be analyzed and the observation accuracy must be improved in future works. Additionally, if the accuracy of the analysis of the travel speed can be improved, the methods employed in this study can contribute to the development of analysis methods for safe vehicular traffic in the future. We used Bluetooth technology on a trial basis in this study; however, future studies must develop other methods that are more accurate and can easily acquire data equivalent to that of Bluetooth.

We included the randomly extracted discrete OD trips for the expansion estimates of the moving trips on a trial basis. Future research must be conducted on determining the actual service paths of all GSE vehicles.

We performed the data observation and analysis over a 24 h × 7 d time interval, including during peak and off-peak hours, and acquired consistent results, even with daily fluctuations in the traffic volume. Therefore, the proposed methodology can be applied to periods other than that from the 21 to 27 November. Analyses conducted during other periods can help in evaluating the GSE driving conditions in all seasons.

The methodology outlined in this study can be applied to other airports, facilitating GSE driving condition surveys and enabling the determination of actual GSE trip conditions through the procedure outlined herein. It is essential to account for the layout of facilities and the traffic rules governing GSE vehicles at the specific airport under study. These findings are anticipated to benefit aviation authorities, airport managers, operators overseeing GSE vehicles, and vehicle manufacturers alike.