Vertiport Operations Modeling, Agent-Based Simulation and Parameter Value Specification

Abstract

Urban air mobility (UAM) is the idea of creating a future mobility market through the introduction of a new mode of aerial transport with substantial travel time advantages. A key factor diminishing travel time savings is vertiport processes. So far, vertiport throughput capacity has only been studied in a static manner using analytical methods, which has been found to be insufficient. This paper wants to increase the level of understanding of operational dynamics on vertiport airfields by being the first to apply agent-based simulation. For this purpose, an existing vertiport model consisting of pads, gates and stands was refined through two means. First, a sensitivity study with over 100 simulations was executed shedding light on the driving processes on a vertiport airfield. Second, an expert interview series with 17 participants was conducted, letting the experts evaluate the model and specify relevant parameter values. Three main results should find mention here: (1) Pad operations were identified to be most impactful on passenger delays. (2) Pad and gate processes have a threshold capacity beyond which delays increase exponentially. (3) A refined vertiport model is presented, including the 27 most relevant parameters and their value specification. In conclusion, this paper finds that optimized vertiport airfield design is crucial to UAM operations, and dynamic passenger and vehicle interactions cannot be neglected.

1. Introduction and Literature Review

Urban air mobility (UAM) has received much attention in recent years, in both academia and the industry, with the potential of introducing a novel mode of transport into urban settings. The trend of urbanization in recent decades [1] has led to increased traffic problems and congestion, which in turn affects travel times, the environment and the overall quality of city life. UAM promises to be a fast [2] and clean [3] mode of transport. Some reports even indicate that UAM could alleviate congestion [4], while most estimations are more careful and do not expect a significant impact on the overall transportation network. A recent study even claims that the opposite could be true: when including access and egress trips to vertiports, the number of cars on the road might increase [5]. Market shares are conceivable to range between marginal significance [6] all the way up to 8% in the long haul [7]. Various market studies have tried to capture the global market potential [8,9,10,11,12,13] in the 2030s; estimates are located in tens of billions USD. This would be a relevant future mobility market, which has the potential to offer a variety of novel, intermodal services. For comprehensive and recent summaries of UAM topics, see publications by Niklaß et al. [14], Straubinger et al. [15] and Garrow et al. [16].

Should inner-city air travel become a reality, and if UAM becomes affordable to a relevant portion of the population, there could be fleets of thousands of aircraft in single metropolitan areas. This development would require highly performant infrastructure. UAM vehicles are envisioned to be fully or hybrid electric and have vertical take-off and landing (VTOL) capability (electric VTOL vehicles (eVTOLs). The infrastructure for eVTOLs, which is commonly referred to as vertiports, will need to cater throughputs in the magnitude of commercial airports [17] on the surface areas of large heliports [18]. Space in cities is costly [19], and airspace management will pose additional difficulties [20]. Vertiports will need to handle this highly constrained environment, which makes vertiport design and operations ambitious undertakings. Early modeling of the impact of UAM on cities and spatial structures was conducted by Straubinger [19]. She found out that the surface area demand of vertiports has a strong negative effect on the overall welfare of the population. Accordingly, designing vertiports with small footprints is an important goal.

Vertiports have been studied to some degree, but there are crucial gaps in the knowledge when it comes to proposed real-life operations. In 2019, the world’s first vertiport was built by Skyports for an exhibition in Singapore [21,22]. Yet, this and similar projects [23] are comparative in size to existing helipads [24] and so far only for exhibitory purposes. Vertiport layout design and current considerations are discussed in detail by Preis [25]. As there are no real-life vertiports operating as of today, there is also no experimental research at this point.

Some work has been conducted concerning the airspace management, looking at equipment to increase safety during landing for small airports [26], airport-specific weather forecasts [27] or dynamic airspace sectorization [28]. These are all valuable considerations on the path to enabling safe air traffic management in the dense airspace surrounding a vertiport. Yet, these studies do not look at VTOL, but rather conventional take-off and landing, and do not consider airfield operations. When looking at VTOL, ground-breaking research was conducted around vertiport capacity envelopes by Vascik et al. [29]. Further, Zelinski’s work on vertiport surface topologies [30] should find mention, as well as Schweiger et al. investigating the level of service a vertiport layout can deliver [31]. What remains is a gap in the current literature around operational dynamics on vertiports and, in particular, the vertiport airfield, which is the main focus of the presented work.

This paper wants to inform applied operations research in academia and start-ups in the industry who concern themselves with vertiport design; it aims to help develop both use cases and best urban mobility practices. Toward this goal, the fidelity of vertiport research will be raised from previous analytical approaches to an agent-based simulative approach. Next to the classical four-step approach of transport modeling (see for example Ref. [32]), agent-based modeling has become increasingly popular for transport simulation in recent decades [33] and has already been applied to simulating UAM [34]. Therefore, this simulation approach seems fitting. Simulation of vertiports goes beyond existing research and presents the highest possible level of fidelity, as real-life experiments are not an option at the moment, as discussed earlier. A high-fidelity vertiport airfield model will be presented throughout the paper, including a careful selection of relevant parameters and an informed determination of their values. The relevant information is drawn from the literature review and an expert interview series. The agent-based modeling and simulation (ABMS) framework in which the model is implemented was first published by Preis et al. [35] and has since been expanded.

The structure of the paper is as follows. In Section 2, the vertiport model will be briefly re-iterated, including all elements, agents and the assembled simulation environment. The refinement of the model through expert feedback will be highlighted. In Section 3, a sensitivity study of over 100 simulations will be presented with the aim of identifying driving operational processes. These processes encompass vehicle approach and departure, passenger boarding and de-boarding, number of pads and gates, initial fleet size of vehicles and the accumulated daily demand. In Section 4, all conceivable parameters are listed. The driving parameters are identified through expert interviews and insights from the sensitivity study. A final short list of parameters is presented, and each parameter value is determined according to values found in the literature and derived from expert interviews. Lastly, in the Appendix A, Appendix B, Appendix C, Appendix D and Appendix E, additional material on the vertiport model, the expert interview series conducted for this paper, the list of parameters and the statistical interpretation of the parameter values can be found.

2. Vertiport Model and Simulation Method

The vertiport model consists of elements, which are composed into a virtual environment and agents moving within this environment. There are three main elements (pads, gates and stands) and three additional elements (lanes, terminal and airspace). Two types of agents move within the environment: passengers and vehicles. Further, the controller plays a key role in coordinating all agents. In Appendix A, the characteristics of all elements and agents and how they are composed into the virtual environment will be described. Except for the terminal and the airspace, all other elements can only be occupied by one agent at a time. Both the early model and the implementation of the software code have been published before [35] and will, therefore, only be re-iterated briefly. Since the first publication, the model has been expanded and validated.

2.1. Vertiport Airfield Environment

In Figure 1, a schematic sketch of the environment can be seen. All pads, gates and stands are connected through lanes or taxiways and represent the airfield on the ground. The airspace is the system boundary on the side of the pads. Vehicles enter the environment through the airspace and need a certain approach time until they stand on a pad, which is differentiated into multiple steps. When vehicles take off from a pad, they need a certain time of departure while they are in the air before they leave the airspace and thus, the simulation boundary. The terminal is the system boundary on the side of the gates. Passengers arrive at the terminal, which can be imagined as a holding place near the gates, after processing (ticket scan, luggage drop-off, security screening). Passenger processing in the terminal is not considered in the environment, as will be explained in Section 2.2. Instead, passengers enter the environment at the point of arriving at the holding area. From the terminal, passengers walk toward a gate, enter the gate and start boarding a vehicle. After de-boarding a vehicle, passengers immediately walk toward the terminal and leave the simulation environment the moment they arrive at the terminal. Unused and empty vehicles are sent to a stand to clear space for other operations. Processes in the airspace and terminal were not part of the original model and were introduced for this paper. Further, a mechanism was introduced to prevent deadlock situations when the capacity limit of the vertiport is reached.

2.2. Vertiport Model Refinement

This model was presented to 17 experts, and three main points were criticized (the expert interview series will be explained in detail in Section 4.1). The points are listed below, and an explanation is given on how they will be accounted for:

• Airspace operations. Several experts expected the main operational bottlenecks in the airspace. Additionally, sending eVTOLs into holding loops for extended times, which is done to prevent deadlocks, might not be possible. Currently, these questions are being addressed in cooperation with the German Aerospace Center (DLR) [36]. Bauhaus Luftfahrt (BHL) will focus on vertiport airfield operations. A well-defined interface between both models on the airside (DLR) and groundside (BHL) will allow a holistic view.
• Passenger processing. Some experts criticized the choice to neglect passenger processing and, in particular, ticket scan, luggage drop-off and security screening, because these are essential and time consuming in today’s commercial aviation. First, these processes might and even must change dramatically in their duration to make UAM viable, wherefore they are not yet easily captured in an accurate model. Second, passenger processing has already been studied extensively for commercial airports [37], which allows adding empirical values to the results of the simulation during post-processing steps to account for pre-terminal passenger processing. The need for real-time simulation is not high. Lastly—and this answer comes from one of the experts on passenger boarding—passenger processing follows very different dynamics than vehicle taxiing. Both things may be difficult to harmonize into one unified environment and should rather be simulated independently.
• Energy management. Other experts pointed out the need to include energy management in the model, in particular vehicle batteries, charging ports and an interface to the electric grid. As will be pointed out in Appendix A.1, this feedback was considered and included in the model. A detailed discussion of the model extension through an energy module will be presented in a separate publication to maintain a realistic scope of this paper. The parameter value determination of key energy-related parameters will still be considered in this paper, but the focus is on general vertiport airfield operations, unconstrained through energy limitations.

This section re-iterated the initial vertiport airfield model and highlighted early updates around the areas of the terminal and airspace. Two further aspects will now be considered to refine the model: first, the model improvements as desired from the 17 experts, which were discussed above. Second, a sensitivity study encompassing over 100 simulations, which will be described in the following Section 3. The goal is to identify the driving process and understand which parts of the model need further differentiation to capture all the relevant elements. Combining expert feedback and insights about the sensitivities will inform the refinement of the vertiport airfield model in Section 4.

3. Vertiport Operations Simulation Results and Identification of Driving Processes

With the updated vertiport model and the implementation in the ABMS simulation framework, which are described in Section 2, the first sensitivity study was undertaken. The goal is to understand the driving processes of vertiport operations, which will inform the following refinement of the model. In the course of the sensitivity study, over 100 simulations will be executed, distributed over six parameter variations. The following six processes or characteristics are expected to play a significant role and will be varied systematically:

• Initial fleet size of vehicles parked on the vertiport at the start of simulation;
• Accumulated demand over a day of operation;
• Approach and departure time of vehicles;
• Boarding and de-boarding time of passengers;
• Number of pads;
• Number of gates.

3.1. Definition of Simulation Sensitivity Study

The sensitivity study did not aim to look at realistic scenarios, but rather decouple the effects and variation of input parameters to understand the impact of individual processes. A 16 h day of operations was assumed. Demand profiles were randomly created from a normal distribution, with its peak in the middle of the operational time (see Figure 2). A normal distribution was chosen over a uniform distribution to understand the effects of peak and off-peak operations. Each distribution has the same accumulated number of vehicles arriving through the airspace and passengers arriving through the terminal, which leads to an identical number of parked vehicles at the start and end of each simulation. Asymmetric demand was judged to be an advanced question at the current stage and therefore not considered.

The baseline scenario is composed of 500 arrivals and requests, 20 vehicles of initial population and a vertiport layout with 4 pads, 12 gates and 20 stands. “Plans” is used as a term to describe both vehicle arrivals and passenger requests; there are, accordingly, a total of 1000 plans in the baseline scenario. Approach and departure times for vehicles are set to 60 s; walking time from the terminal to the gate and reverse for passengers is set to 20 s; boarding and de-boarding time is 120 s; and the speed of vehicle taxiing is 2.2 m/s. Five variations of the demand profiles (baseline profile shown in Figure 2, profiles B-E shown in Figure 3) were created in a Monte Carlo approach to account for stochastic effects. Next to the baseline scenario, 30 other sets of parameters were created with each 3–5 random samplings of the demand profile, resulting in a total of over 100 simulation scenarios (see

Table 1. Overview of parameter variations for simulations of the sensitivity study.

Aspect to Be Varied	Demand Variations Included	Varied Values (Baseline Marked)	Description
Fleet Size	A-C	0, 8, 12, 16, 20, 24, 32	Number of vehicles initially parked at vertiport
Demand	A-E	750, 1000, 1200, 1250, 1300, 1500	Accumulated demand over day of operations
Approach and Departure	A-C	30, 60, 75, 90, 105, 120	Time vehicle needs to land or take off (in seconds) before next operation is possible
Boarding	A-C	1, 2, 3.5, 4, 4.5, 5	Time passenger needs to board or de-board a vehicle (in minutes)
Pads	A-C	2, 3, 4, 5	Number of pads included in airfield layout
Gates	A-C	6, 7, 8, 9, 10, 12, 16	Number of gates included in airfield layout

). In the following, the average delay time per passenger over variations of parameters will be discussed.

3.2. Performance Indicator of Average Passenger Delay

The vertiport performance of any given simulation scenario in the sensitivity study will be expressed through a value, which is labeled “average (passenger) delay”. This value is defined as the involuntary or idle waiting time of n passengers, which is caused by non-optimal operations. In other words, the delay of passenger i is the difference between actual operations t_i(actual operations) and optimal operations t_i(optimal operations) (see Equation (1)). Delay can occur while a passenger is waiting in the terminal to have a vehicle assigned to them or while waiting at the gate when the assigned vehicle has not yet arrived at the gate. Delay can also occur while the passenger is sitting in the vehicle and the vehicle is forced to wait, both on the ground and in the air.

$$average delay=\frac{1}{n}\ast {\displaystyle \sum }_i^n\left[t_i\left(actual operations\right)-t_i\left(optimal operations\right)\right]$$

(1)

Throughout the sensitivity study, it will be observed that there are, generally speaking, two regions for each variation. First, there is a region of low average passenger delay (less than 5 min), where a variation of parameter has little impact. Second, there is a region of high passenger delay (more than 5 min), where variations of the parameter lead to an exponential increase in passenger delay. The first type of region will be considered “stable” operations, the second type of region will be considered “unstable” operations.

3.3. Simulation Results of Sensitivity Study

3.3.1. Fleet Size

The fleet size or initial population at the start of simulation is set to 20 vehicles in the baseline scenario. A quadratic fit approximates the data well (see Figure 4), showing an optimal spot around the center (16 vehicles) when half of the vertiport parking capacity is used initially (32 vehicles maximum capacity, 20 at stands and 12 at gates). Depending on the demand profile, the optimum shifts more toward an initially fuller vertiport (shift to the right) or an emptier vertiport (shift to the left). The explanation might be a temporal excess of requests or arrivals, respectively. Average delays of passengers vary between 0 and 5 min, showing that the initial population has very little impact on the delay.

3.3.2. Demand Magnitude

The accumulated demand of the baseline scenario consists of 500 vehicles arriving through the airspace with each one passenger on board and 500 passengers arriving through the terminal, resulting in 1000 passengers over the course of one day. For each demand magnitude from 750 to 1500 passengers, five random samples were created (see Figure 5). In particular, for an accumulated demand of 1200 passengers and more, the variation is strong, which suggests the interpretation that with increased demand, the system becomes unstable. Further, temporal peaks in demand probably affect the average delay of passengers more strongly than the accumulated demand over a day of operations. Average delays of passengers vary between 0 to 10 min. In summary, it can be said that the demand has a moderate impact on the delay, but the accumulated number of passengers is not a reliable indicator to predict passenger delay.

3.3.3. Approach/Departure and Boarding Times

The approach and departure time of the baseline scenario is each 60 s. An exponential fit approximates all three randomly sampled demand profiles nearly perfectly (see Figure 6). It can be seen that approach and departure times of up to 75 s have almost no impact on the average passenger delays. For more than 75 s, the average passenger delay then increases rapidly up to around half an hour for 120 s, which is twice the approach and departure time of the baseline scenario. Vertiports operate stably under a stretch of approach and departure times, but operations become unstable and delays increase rapidly once a threshold is crossed. Average delays of passengers vary between 0 and more than 30 min, which shows the major impact approach and departure times have on the delay.

The boarding and de-boarding time of the baseline scenario is each 2 min. An exponential fit approximates the results well (see Figure 7). Similarly to the approach and departure time, there is a long stretch of stable operations up until around 3.5 min. Afterward, the average passenger delays rise exponentially. The effect is not as strong as for the approach and departure time; at twice the boarding time of the baseline scenario of 4 min, the delays are still below 5 min. An interpretation can be that increased boarding time will lead to unstable operations and exponentially increasing delays, but the boarding time is not as critical as approach and departure time. Average delays of passengers vary between 0 and 20 min, showing a substantial impact on the delay.

3.3.4. Number of Pads and Gates

The number of pads is four for the baseline scenario. An exponential fit approximates the results nearly perfect (see Figure 8). Only discrete variations of the number of pads are possible, which limits the resolution along the x-axis. At three pads, the average passenger delay times have increased, but operations seem to be stable. At two pads, a tipping point in delays is seen, pointing to a highly critical relationship between stable vertiport operations, expressed in low average passenger delays, and a minimum number of pads. Average delays of passengers are below 5 min for three or more pads and are around 45 min for two pads. From the observed factors, the number of pads has the strongest impact on the delay.

The number of gates is 12 for the baseline scenario. An exponential fit approximates the results for individual demand profiles nearly perfectly (see Figure 9). This concludes that the particular demand profile has an effect on how efficiently the number of gates operate. For 10 or more gates, operations are stable; between 8 and 9 gates, average passenger delays start increasing and then rise exponentially for lower numbers of gates. Similarly to the number of pads, the number of gates shows a region of low delays and then experiences an exponential increase after a threshold toward a region of unstable operations. Average delays of passengers are below 2 min for 10 or more gates and increase to over 20 min for as few as 6 gates. The number of gates has a substantial impact on passenger delay.

3.4. Summary of Driving Processes in Vertiport Operations

Vertiport operations are highly sensitive to a variation in the presented driving parameters, typically showing a region of stable operations with low average passenger delays and an unstable region with exponential increase in delays. The initial fleet size at the start of a simulation has an optimal point at half capacity (half of all stands and gates are occupied by available vehicles) of the vertiport and has very little impact on the average passenger delay. The accumulated demand over a day results in scattered results, suggesting that temporal peaks in demand have a larger impact than the overall demand. This finding suggests that vertiports need to be designed toward a peak hour, not toward accumulated demand.

Both processing times and vertiport sizes show exponential growth in average passenger delays after crossing a certain threshold. Vehicle approach and departure time and the number of pads appear to be more sensitive than passenger boarding and de-boarding time and the number of gates. The central focus in designing vertiports should therefore be placed on optimal placement and handling of pads. Selecting the ratio between the number of pads and the number of gates, also accounting for the processing times, is also critical to plan an efficient vertiport layout. Once a vertiport layout is planned, it can handle a certain amount of demand, but not much more; due to the tipping point characteristic, it will quickly enter unstable operations, and average passenger delays will increase exponentially.

In this section, the previously published software implementation of the ABMS [35] was applied. Over 100 simulations were executed to understand the sensitivities of vertiport airfield operations. The focus of the initial publication lay on the software architecture, whereas this paper focuses on the application of the framework to generate new insights.

4. Discussion of Vertiport Parameters and Values

In this section, the list of parameters that are incorporated in the refined vertiport model are introduced, including the evolution of the list and the current short list of most relevant parameters. The model was refined based on the responses of experts presented in Section 2.2 and the insights from the simulation sensitivity study presented in Section 3.4. Initially, a list of parameters was formulated with a total of 82 entries assembled into 10 groups. Through expert interviews and literature review, the list was expanded and individual parameters were differentiated or aggregated where necessary, resulting in a full list of 95 parameters. Irrelevant parameters were dropped to form a long list of 55 parameters, and from these, the most relevant 27 parameters were chosen and assembled into 5 new groups. The full list of parameters can be seen in Appendix C and the short list of parameters in Appendix D. In the following sections, the expert interviews and selection process will be discussed in further detail.

4.1. Expert Interview Approach

As UAM, in its proposed magnitude, is a novel field of transportation, the scope and confidence of parameters for vertiport ground operations are strongly limited in current literature, wherefore expert interviews were chosen as a method for gathering data. The interviews followed a semi-structured approach with a list of questions related to UAM, the vertiport model and vertiport parameters. Each interview lasted 30–60 min and was conducted through video calls. Experts were selected according to their expertise and experience; each expert was interviewed in person and filled out a questionnaire about specific parameter values after the interview. The questions of the interview questionnaire can be found in Appendix B. The interview series took place between October 2020 and July 2021. All responses are anonymized for protection of privacy and confidentiality of the answers. For this reason, only the aggregated values can be shown in this paper.

In total, 17 experts took part in the interview series with backgrounds in research, the aviation industry, aircraft operations and architecture. The experts were from the USA, Great Britain and Germany and had an average experience in their field of 10.6 years (median experience 7 years). A total of 19 qualitative remarks toward the vertiport model were given, which are incorporated in Section 2.2. Forty-nine qualitative responses concerning the definition of parameters were gathered during the interviews. This information was incorporated in the refinement process of the parameter list presented in the following section.

4.2. Identification of Relevant Parameters

For the vertiport model, a total of 97 parameters in 10 groups were considered, out of which a short list of the most relevant parameters was formed, including 27 entries separated into 5 groups. There were four stages of the parameter list: the initial list (82 entries), the expanded full list (97 entries), the refined long list (57 entries) and the selective short list (27 entries). The responses from the expert interviews helped to expand the initial list toward the full list through adding new parameters or differentiating existing parameters. Further, the expert responses aided in forming the long list by aggregating multiple parameters into single parameters and dropping irrelevant parameters. The short list (see Section 4.3) was finally formed by looking at the expected driving processes (discussed in Section 3.4) on a vertiport and reducing the list of parameters to account for the most relevant aspects.

In

Table 2. Evolution of list of parameters and division into initial groups.

ID	Initial Group	Initial List	Full List	Long List	Short List
A	Vertiport Elements	9	9	5	4
B	Passenger Processing	9	12	7	2
C	Passenger Boarding	16	16	4	4
D	Vehicle Taxiing	6	8	6	5
E	Take-Off and Landing	5	5	3	2
F	Flight Approach and Departure	11	14	14	6
G	Battery Charging	10	11	9	3
H	Battery Swapping	8	10	3	1
I	Vehicle Maintenance	6	8	2	0
J	General Vertiport Operations	2	4	4	0
	Total	82	97	57	27

, the initial 10 groups of parameters are shown and the number of parameters from the initial list, the full list, the long list and the short list.

4.3. Final Short List of Parameters

A short list of 27 parameters, which capture the most relevant processes in a vertiport, is shown in Appendix D. The IDs are coherent over the evolution of the parameter list (see all parameters considered in Appendix C). The short list of parameters is separated into five new groups and given a new index:

• Vertiport elements (no index)
• Flight approach and departure (indices A and D)
• Passenger boarding and de-boarding (index B)
• Energy management (index E)
• Vehicle taxiing (index T)

Group vertiport elements was shown to be impactful in Section 3.3.4, where the number of pads and gates in a vertiport layout impacted the average passenger delay significantly. Groups boarding and approach and departure were both highlighted by the experts in Section 2.2 and shown to be relevant for vertiport operations in Section 3.3.3. Group energy was also highlighted by the experts in Section 2.2. Lastly, group taxiing was added due to the potential impact of different taxiing concepts, which have not yet been studied in depth.

All groups that have operational character (all except for vertiport elements) will be considered in the parameter value determination in Section 4.4. For group vertiport elements (which includes the dimensions of pads, gates, stands and taxiways), the values can be extracted with good confidence from existing heliport standards (see heliport manuals by FAA [38] or ICAO [18]). In addition, this group does not have any operational character in the same sense as the other four groups, but rather determines the possible vertiport layouts, which have to be designed prior to simulating a day of operations. Parameter groups boarding, taxiing, approach and departure and energy will be explained in the following and are visualized in Figure 10.

4.3.1. Flight Approach and Departure

Parameter group approach and departure starts with a vehicle being in the controlled airspace of the vertiport and starting final approach. It enters the physical airspace above the pad (A1), transitions into a final hover over the center of the pad (A2), touches the ground and shuts down the engines (A3), if the taxi mode is other than hover taxi. The pad needs a time to “cool down” after an operation (A4) to account for wake vortices and separation minima between vehicles. In reverse order, the vehicle starts its engines (D1), unless the taxi mode is hover taxi, lifts off and hovers (D2), and transitions into forward flight leaving the physical airspace of the pad (D3). Analog to the approach process, the pad needs a “cool-down” time (D4) before the next operation can be initiated.

4.3.2. Passenger Boarding and De-Boarding

Parameter group boarding starts with a passenger arriving or waiting at the holding place in the terminal. Once a vehicle is assigned, the passenger walks toward the gate (B1) and enters the gate (B2), which can be imagined as stepping through a door into the immediate proximity of a vehicle, in order to step into the vehicle and board it (B3). In reverse order, the passenger de-boards the vehicle (B4), leaves the proximity of the vehicle, steps through the gate door (B5) and subsequently walks toward the holding place in the terminal (B6) to exit the simulation.

4.3.3. Energy Management

Parameter group energy contains parameters for both charging or swapping vehicle batteries. As in the case of swapping, the batteries still need to be charged. Charging is essential for both types; for battery swapping, the charging process occurs remotely. The battery swapping time (E4) is aggregated into one parameter encompassing the entire swapping procedure. The battery has a certain capacity (E2), which corresponds to the usable capacity, not the full physical capacity of the battery. Therefore, a depth of discharge reserve needs to be subtracted beforehand. The batteries can be charged with a certain charging speed or charging power (E3), which is assumed constant as a first approximation. During the charging process, some energy is lost due to inefficiencies in the charging process. The losses are given as relative losses to the overall processed energy (E1).

4.3.4. Vehicle Taxiing

Parameter group taxiing encompasses three modes of taxiing: “hover” taxi, “passive” taxi and “active” taxi. The “passive” taxi mode facilitates a taxiing device, which is first mounted to the vehicle (T1), then the device moves over the airfield (T2), and at the destination, the device is de-mounted (T3). It could be pushing, pulling or carrying the vehicle, performed by a movable platform, through ropes or a bot, respectively. The “active” taxi mode facilitates auxiliary electrical motors at the wheels of the vehicles (skids are not possible in this scenario) through which the vehicle can taxi (T4) without using the main engines or an external device. The “hover” taxi mode facilitates the main engines to hover near the surface (eVTOLs with skids) or touch the ground and roll (eVTOLs with wheels) (T6). If the engines are off at the start of taxiing or have to be off after taxiing, the engines are started (T5) or stopped (T7), respectively.

4.4. Parameter Value Determination

After selecting and defining the essential parameters for the vertiport model in the previous sections, the parameter values are now going to be determined. As no vertiports in the envisioned dimension are operating at the time of writing, no direct experimental data are available. Some parameter values can be borrowed from similar applications, for example, the parameter values in group elements can be determined according to existing heliport guidelines [38]. Yet, for most parameters, value determination poses a substantial challenge. In order to attempt an initial value determination of the presented vertiport model and its parameters, two sources will be included in the following: parameter values determination from other authors’ statements captured in the literature and the estimations given by the experts during the interviews presented in Section 4.1.

From the literature review, a total of 135 datapoints were identified of which a subset of 47 parameter values is included in this paper. The other datapoints were either values in an aggregated form, had differently defined parameter boundaries or corresponded to parameters, which are not included in the final short list of parameters (see Section 4.2 for the selection of parameters). The experts answered a total of 186 datapoints, of which 77 correspond to one are another parameter of the final short list.

In

Table 3. List of parameter values in the refined vertiport model, including references.

ID	Parameter	Value	Unit	# Experts	# Literature	References
B1/B6	Terminal to/from gate	31.9	s	3	0
B2	Enter gate	19.7	s	5	0
B3	Boarding	73.0	s	4	5	[8,30,39,40,41]
B4	De-boarding	65.8	s	4	4	[8,30,40,41]
B5	Leave gate	26.7	s	2	0
T2	Passive taxiing	2.63	m/s	6	0
T1/T3	(De-)mounting passive taxiing device	27.0	s	3	0
T4	Active taxiing	2.15	m/s	5	1	[42]
T6	Hover taxiing	3.25	m/s	3	3	[30,36,43]
T5/D1	Start engine	4.50	s	2	0
T7/A3	Stop engine	5.00	s	2	0
A1	Enter airspace	46.3	s	6	2	[8,44]
A2	Final hover	22.9	s	5	5	[43,45,46,47,48]
D2	Initial hover	13.5	s	5	5	[43,44,45,46,48]
D3	Leave airspace	28.7	s	6	1	[44]
A4/D4	“Cool-down” after landing/take-off	30.0	s	3	3	[49,50,51]
E3	Charging speed	311	kW	4	7	[42,44,45,48,52,53,54]
E2	Battery capacity	133	kWh	4	3	[46,48,55]
E1	Energy loss	7.17	%	3	3	[44,48,52]
E4	Swapping time	349	s	2	5	[11,52,56,57,58]

, the value specification of all parameters identified as crucial for the vertiport model are shown (please find the definition of all parameters in the final short list in Appendix D). In Appendix E, a more detailed discussion of the statistical aggregation and confidence of value determination is presented.

5. Conclusions and Future Work

Urban air mobility promises to enrich the current transportation system with a new and fast mode of transport. Advances in the required infrastructure, in particular, vertiports, are necessary to realize the promises UAM offers. As of today, no vertiports are operating, and experimental data are, except for related research on heliports, non-existent. Through expert interviews and agent-based simulation, this paper attempted to raise the fidelity of vertiport operations modeling to a higher level.

A vertiport model, consisting of pads, gates and stands, was introduced. The model was implemented in an ABMS framework, with passengers and vehicles as agents. Experts were asked to evaluate the model and recommend refinements. Their answers on airspace operations, passenger boarding and energy management were discussed. Next, a simulation sensitivity study was conducted, showing that vertiports have a region of stable operations and a threshold beyond which operations become unstable, causing passenger delays to increase exponentially. In particular, the number of pads and length of approach and departure time of vehicles were identified as design drivers. This study further guided the focus of model refinement. Lastly, through expert interviews (n = 17) and results from the sensitivity study, the early model (initially presented by Preis et al. [35]) was refined, and 27 driving parameters were identified. These parameters were grouped into vertiport elements, vehicle approach and departure, passenger boarding and de-boarding, energy management and vehicle taxiing. Through the literature review and expert estimations, the parameter values were specified.

The work of this paper can aid the wider scope of research in at least three ways. First, the updated vertiport model (Section 2) may be used as a reference for general research around infrastructure for UAM. Second, the results of the sensitivity study (Section 3.4) give valuable insights into the design drivers for vertiport operations and will be helpful when planning vertiport layout and operations in highly constrained urban environments. Third, the list of parameters (Section 4.3 and Appendix D) and, in particular, the specified parameter values (Section 4.4) can be used in part or as a whole for studies in the area of UAM. There is not much consensus in the UAM community on the parameter values, which makes this specification especially valuable. A further strength is the combined approach of literature review and expert estimations.

In the future, the ABMS framework will be used to further investigate the effects on vertiports, in particular the questions of taxiing modes, energy sources and charging vs. swapping of batteries. An ongoing refinement of the parameter values will also be performed as new research is conducted throughout the community. Lastly, a stochastic extension to the vertiport model is planned to account for variations, in particular, between passengers in real life.