*5.2. Aerial Network Model Simulation Scenario*

By considering the digital twin model of the metropolitan area of Bologna—which includes a wider area with respect to the example in Section 3 and also factors and constraints for implementing the 3D-UAN in the city—three vertiports have been identified (Figure 5), which are, respectively, at the airport (vertiport 1), at the main station (vertiport 2) and in a city area with high population, job density and median income (vertiport 3). It is worth noting that this result has been obtained by using still freely available data. More accurate results and possibly different vertiport locations and/or number might be obtained by using a more detailed database.

**Figure 5.** Vertiport location supporting 3D-UAN—case study.

Starting from the vertiport locations, the routes linking them have been set using Equation (1) and the information provided by the digital twin model about local orography and barriers. Where available, data about local environmental features have also been considered. Particularly, some constraints to avoid conflicts with traditional aviation have been introduced in the digital twin model. As for the urban environment, the presence of high buildings—such as historical towers and modern skyscrapers—has prevented the connection of vertiport 1 with the other remaining two with a straight line (see Figure 5).

Figure 6 depicts the scheme of the obtained 3D-UAN network, while link and vertiport features are reported in Tables 2 and 3, respectively. Due to the information contained in the digital twin model (i.e., DTM and building heights) and the introduced constraints to avoids obstacles along the dynamic air corridors, the altitude of the first layer (L1) resulted at 200 m ASL (the highest building roof is at 184 m), while the following one (L2) is at 250 m ASL (distance between L1 and L2 = 50 m).

**Figure 6.** The 3D-UAN model for the case study.



**Table 3.** Vertiport elevation.


To provide a preliminary application of the 3D-UAN for the case study, the following hypotheses have been considered, particularly the existence of j on the *dm*,*L*; *Tg*(*j*, *<sup>j</sup>*−1) is the time gap between ACVs *<sup>j</sup>* and *<sup>j</sup>* <sup>−</sup> 1., i.e., to the upper layers or to the lower layers. If there are multiple ACVs on the same link, i.e., *<sup>i</sup>* > 1, the gap time, *Tg*(*i*,*i*−1) , guarantees suitable separation between two following ACVs along vertical links.

For horizontal links, *Tti* is the running time, *Tri* , if *i* = 1, while if *i* > 1, the gap time is *Tg*(*i*,*i*−1) . To assess the waiting time component, the gap time, *Tg*(*i*,*i*−1) , the following are utilized: (i) advanced communication technologies (V2V, V2I/V2X) [61,62]; (ii) high precision ACVs on-board sensors, such as radars, Detection and Avoidance (DAA) [63]; (iii) data transmission networks (i.e., FANET system [64]).

The average cruising speed allowed on the horizontal links is considered equal to 90 km/h, while on the vertical links (included take-off and landing phases) it is 50 km/h. This average speed value was chosen based on the data distributed by Volocopter on their first prototypes [65], which should also support aspects related to flight safety.

Two scenarios have been investigated, which have been set as follows:


For both scenarios, the dynamic link features are reported in Table 4. Furthermore, for S1 and S2, six ACVs have been considered connecting specific origin/destination pairs, and their shortest paths have been computed (Table 5). Additionally, a *Tg* value has been assigned to fixed nodes, particularly for each scenario: ACV (1) and ACV (2) takes-off at the same time *t*<sup>0</sup> and *Tg*<sup>1</sup> = 0; ACV (3) and ACV (4) takes-off at time *t*<sup>1</sup> and *Tg*<sup>2</sup> = *t*<sup>0</sup> + *Tg*(1,2); ACV (5) and ACV (6) takes-off at time *t*<sup>2</sup> and *Tg*<sup>3</sup> = *t*<sup>1</sup> + *Tg*(3,4). Table 6 reports a detailed analysis of the two scenarios over different periods of time

**Table 4.** Dynamic link features.



**Table 5.** O/D nodes, travelled distances and path costs in the two scenarios.

**Table 6.** Dynamic link status and ACV position in the two tested scenarios.


\* N = node. \*\* = link.

From Tables 5 and 6, it can be seemed that, in a small urban context (such as the city of Bologna), a time gap *Tg* = 180 s between successive departures would not generate air traffic congestion, even if there are crossing routes. A different result is obtained in the case of *Tg* = 120 s. In this scenario, the ACV (5) has to exploit the dynamism of the links and switches to the next layer to ensure flight safety and avoid link congestion. In fact, thanks to the information stored regarding the air traffic conditions, by both the ACV and the control centre, the link (7-5) is disabled for transit, while the vertical links (7-11) and (8-4), and the horizontal links (11-10) and (10-8) are enabled, ensuring a suitable air transport service.

#### **6. Discussion and Conclusions**

The previous sections described two main aspects of UAM systems—i.e., vertiport locations and 3D UAN—that may benefit from the digital twin approach. The presented framework, together with the proposed 3D-UAN model, has been applied to the real context of a medium size city and its metropolitan area. Although this is preliminary research, the results obtained are very encouraging.

The first step, i.e., the location of the vertiport close to the main train station in the city area, confirmed some of the suggestions and preliminary results in the literature. The identified area has some interesting features, such as high levels of population and job densities—which would generate demand levels suitable for supporting UAM services—and good ground connections—which would assure great accessibility to the vertiport from the remaining part of the city.

As for the 3D-UAN structure, in the case study the metropolitan area has been considered, which is more suitable to this aim. Three vertiports have been identified and, by using the digital twin information on the most important factors and constraints to set safe aerial routes, a preliminary 3D-UAN structure has been identified. Furthermore, a preliminary simulation of the aerial traffic flows has been provided, based on Equations (2)–(4). In a real operational context, detected traffic data could also be added to the digital twin of the system as time series data useful for figures and off-line scheduling purposes. In fact, in the case of scheduled services, the computation of the shortest path and assignment of ACVs at specific enabled links is performed before the departure of each aerial vehicle, based on pre-trip information regarding the origin and destination points of ACVs. Scheduled minimum paths and vehicle separations may be computed based on Equation (2).

It is worth noting that the data used for feeding the digital twin of the case study were limited, because only freely available data have been used at this stage, and some other data should be added for improving the nature of the information provided by the digital twin, both static and dynamic data. For example, information regarding static and dynamic populations could be relevant for dynamically adapting the 3D-UAN in order to avoid overflying crowded locations. Similarly, information on dynamic and static population density could be used to adjust the dynamic corridors in order to reduce the externalities produced by ACVs, e.g., noise emissions during the day or night, respectively. In the case study of Bologna, the information about the dynamic population is particularly relevant because the city hosts one of the most important universities in Italy, with a student population of about 90,000, which is a high number compared to the whole population of the city. Particularly, many of them often live in Bologna for a limited number of months. In this context, dynamic population density data are probably the best option for safety evaluations of aerial corridors. Moreover, information on how much a location is busy—which might be obtained by several sources (e.g., Google popular times)—can also be useful to design and adapt the 3D network, while information on areas subjected to urban canyon effects may help in refining the optimal routes.

Another important aspect that affects the location of the vertiports and the design and management of the 3D-UAN is the energy consumption of each ACV. While for small urban areas—as, for example, the city of Bologna—this kind of analyses is not necessary because current aerial vehicle prototypes could realize several trips with a single charge; however, for larger cities such as Rome, Paris or London the cost function of the 3D-UAN model should include the energy consumption factor so that the computation of paths between vertiports will also consider the vehicles autonomy. The maximum autonomy range can then be included as relevant information for setting suitable locations of vertiports and routes between them, as ACV energy autonomy affects the length of the links, also depending on the vertiport (fixed node) where the recharging facilities has been located [66].

To summarize, the opportunity to use a digital twin approach at a high detail level will help system designers and urban planners to evaluate and implement procedures to realize successful aerial networks which can support the existent ground transportation system.

For the effective use of digital twin models in UAM scenarios, several aspects should be considered. First of all, it is important to validate the accuracy and the precision of the gathered data before their integration into the digital model. For example, inaccuracies in the measurement of building heights would compromise safety and produce problems in the risk management process as well as vertiport location, which strongly depends on obstacle clearance. Secondly, continuous digital twin data update is required to reproduce the actual conditions of the represented system. For example, if cranes are introduced inside the city for construction aims, which are possible obstacles to drone operations, the 3D network should be verified and changed, if necessary, in order to avoid these hindrances.

To conclude, this paper proposed a coherent framework based on a digital twin approach in order to deal with the vertiport location problem and the aerial network setting. The digital twin model may significantly support the proposed 3D-UAN, characterized by a high degree of dynamism. Particularly, dynamic links may be enabled or disabled according to real time conditions that are expected to be included in advanced digital twin models of the territorial system. In this perspective, data regarding link traffic volumes, also transmitted in real time among both ACVs and traffic control centres (e.g., UTM), should be included, which will allow one to compute new routes, even in real time. The dynamic requirement of the links represents a considerable advantage and can be utilized to better manage the system and guarantee adequate separations and fast and competitive transport services, especially for medium-long distances.

Further studies will integrate other variables in the localization process, such as environmental capacity, to limit the impacts on population as well as the use of an energy consumption factors in the cost function of the 3D-UAN model.

**Funding:** This research received no external funding.

**Data Availability Statement:** The datasets used in the current study to realize the digital twin are publicly available and can be accessed at the following links: DTM is available https://geoportale.regione. emilia-romagna.it/catalogo/dati-cartografici/altimetria/layer-2 (accessed in June 2022) whereas the other types of data for the city of Bologna are available at https://opendata.comune.bologna.it/ pages/home/ (accessed in June 2022).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

