1. Introduction
The aerospace industry has always been at the forefront of innovation, pushing the boundaries of human exploration and technological advancement. As we embark on the journey towards the future, a new concept known as “digital twins” is poised to revolutionise the aerospace landscape [
1]. Digital twins, virtual replicas of physical assets, have the potential to transform how we design, build and operate aerospace systems [
2]. By integrating real-time data from sensors installed on physical assets, digital twins can monitor equipment health, predict maintenance needs and proactively address potential issues [
3]. This predictive maintenance approach minimises downtime, improves safety and maximises the lifespan of aerospace systems.
Cities are ill-prepared for the integration of UAM (Urban Air Mobility) into their existing transportation infrastructure, which presents a significant challenge. UAM has the capacity to revolutionise a city economy, generate employment opportunities, reduce emissions compared to conventional road vehicles and aircraft and lower infrastructure costs [
4,
5]. It heralds the beginning of a new era of multimodal transportation. While UAM has been considered a luxury mode of transport since the pandemic, AAM (Advanced Air Mobility) may play a crucial role in the efficient transport of time-sensitive cargo, including vital medical supplies, manufacturing equipment and other packages where saving 30 min to an hour or more justifies the use of airborne vehicles [
6]. The focus of the study was specifically on the application of UAM and air taxis within an urban setting [
7]. In [
8], the authors explore the application of 3D GIS environments for Advanced Air Mobility route planning operations, demonstrating the role of data analytics in decision-making by visually representing key influential factors. Ref. [
9] propose an innovative method for modelling future Urban Air Mobility (UAM) for Middle-Mile Delivery (MMD) using Systems-of-Systems (SoS) methodologies and real-world datasets, presenting a framework that allows for analysis of resources, operations, policies and economics involved in the operation of future UAM fleets for MMD. However, their research lacks empirical validation in real-world contexts, signifying a need for future studies that apply and test these models in diverse locations and scenarios. A comprehensive review of the advancements, standards and regulations related to major unmanned aircraft systems can be found in [
10], evaluating the specific technologies required for urban air mobility and exploring operational scenarios based on lessons learnt from remotely piloted aviation and novel unmanned traffic management systems.
While the advent of Urban Air Mobility (UAM) and Advanced Air Mobility (AAM) paints a promising picture for the future of transportation, it concurrently introduces new complexities in air traffic management. Unmanned Traffic Management (UTM) systems are emerging as an essential solution to manage and control the growing traffic in low-altitude airspace, particularly in urban environments [
11,
12,
13]. UTM systems are designed to integrate manned and unmanned aircraft operations in the same airspace, enhancing safety and efficiency while ensuring minimal human intervention [
14,
15]. The adoption and success of UAM and AAM highly depend on effective UTM systems that can dynamically adapt to diverse and high-density traffic scenarios while ensuring safety, security and regulatory compliance. As we dive deeper into this new era of aviation, it becomes increasingly crucial to develop sophisticated UTM systems that can fully unlock the potential of UAM and AAM [
16,
17,
18].
In this context, the AoF project holds the potential to advance UAM initiatives by offering a detailed case study and a framework for cities to address the challenges associated with UAM and UTM implementation [
19]. The vision for the Airspace of the Future (AoF) project is to enable routine operational drone services in a safe coordinated environment on a regional and national basis in cognisance of realistic end user requirements; validated by robust business cases, simulation, stakeholder and public engagement; underpinned by an integrated transportation model with aviation at its core and an exploitation roadmap for the UK [
20]. The Airspace of the Future study encompasses a wide range of interconnected domains, including the development and compliance of regulatory frameworks, enabling technologies and tool-sets such as digital twins and simulation. Additionally, it also addresses crucial societal concerns such as the validation of public acceptance of commercial drone operations. The key objectives are:
Develop the rules, system of systems and operational safety cases to allow mixed use airspace by manned and unmanned traffic.
Establish a national test and evaluation facility for commercial unmanned vehicles with representative operational environments which are digitally trusted and secure.
Develop customer use cases for large-scale virtual and live demonstration in an expanded and open access environment.
Develop a blueprint for the future national airspace structures and ground infrastructure.
Develop a virtual experimentation environment and digital twins to test new rules, processes, systems, technology and operating concepts rapidly at scale.
A digital twin is the virtual copy or model of any physical entity (physical twin) both of which are interconnected via exchange of data in real time [
21]. Digital twin technology saw its origins in the aerospace industry and it is expected to revolutionise other industries [
22]. The digital twins trend is gaining momentum thanks to rapidly evolving simulation and modelling capabilities, better interoperability and IoT sensors and more availability of tools and computing infrastructure. As a result, capabilities of digital twins are more accessible to organisations across industries [
23,
24]. UAV digital twin technology holds immense potential in areas like real-time infrastructure monitoring, precision agriculture, smart city construction and intelligent security [
25]. The digital twin concept of UAV on-board systems consists of several subsystems, including communication, navigation, surveillance, DAA (Detect And Avoid), geofencing and the autopilot [
26]. The communication subsystem enables the transmission of data to and from the Unmanned Aircraft (UA), including direct communication with the Remote Pilot Station (RPS) and, in some cases, communication with the U-space service provider [
27]. Navigation involves measuring or estimating the state vector of the aircraft and is integrated into the Flight Management System (FMS) within the Remotely Piloted Air Systems (RPASs) on-board system architecture [
28]. The FMS also handles waypoint management, guidance priority, envelope protection and emergency procedures. The surveillance subsystem enables the transmission of the identification and position of the UAV. Geofencing involves on-board management of constraints and utilises DAA to ensure that the specified areas are not violated [
26].
Existing literature presents the advances in UAV, UTM and UAM digital twin technologies. Ref. [
29] introduce a simulation environment and digital twin support for shared drone infrastructure in smart cities. This system allows comprehensive pre-deployment testing and real-time malfunction detection, addressing safety and privacy concerns. While the research offers a simulation environment and digital twin support for shared drone infrastructure, the system lacks integration with U-space services, an essential aspect for managing drone flights effectively and efficiently. Ref. [
30] focus on developing an early-stage digital twin framework for ground-to-air emissions using small unmanned aircraft systems, highlighting the need for environmental sensing and efficient plume behaviour replication. Ref. [
31] present a comprehensive review on recent developments in Internet of Drones (IoDs), emphasising the potential of machine learning and deep learning algorithms to enhance IoD functionalities, including navigation, battery scheduling, object tracking, collision avoidance and security. They also identify existing challenges and areas for future research. However the research falls short of providing a practical demonstration of a digital twin implementation that could have helped to validate these theoretical concepts. Ref. [
32] propose the concept of a digital twin city as a transformative solution for smart cities. Their study examines characteristics, key technologies, application scenarios, theories and research directions associated with digital twin cities, aiming to tackle challenges of urban governance due to global warming, population growth and resource depletion. Nevertheless the paper does not address the complexity and practical challenges involved in creating and validating a high-fidelity digital twin of an entire city. Ref. [
33] combine spatial digital twins with a convolutional neural network algorithm to investigate the airspace structure and safety performance of UAV systems. Their study indicates that this approach can improve safety performance, decrease packet loss rates and enhance network availability, providing valuable insights for future UAV applications. Nevertheless, the limited scope of this research in evaluating parameters presents a shortcoming, underscoring the need for future work to consider a wider array of parameters integral to UAV network operations. Ref. [
34] develop a digital twin to analyse and optimise vertiport capacity management in electric air mobility networks, underscoring the significance of network design and maintenance policies in enhancing service provision, passenger satisfaction and asset utilisation. Despite this, the paper falls short in demonstrating how the model can interact in real-time with physical drones and other elements of a UTM system, thereby limiting its practical utility. Furthermore, the research heavily relies on an idealised simulation and lacks real-world empirical validation, thereby restricting the generalisability of the findings to actual, on-the-ground air mobility networks. Ref. [
35] presents a digital twin model for designing and developing Urban Air Mobility (UAM)/UTM applications, such as vertiport location problems, airspace and air vehicle management. The paper discusses a digital twin of a case study for a 3D Urban Air Mobility Network. The authors noted that the data used for the model was limited and lacked validation, which could affect the accuracy and completeness of the model.
Building on this, our research develops an Airspace of the Future (AoF) digital twin for the National Beyond Visual Line of Sight (NBEC) Experimentation Corridor [
36]. The digital twin represents the airspace realistically, incorporating elements such as terrain and buildings and facilitates drone flight simulations to identify potential issues and optimise flight plans. Additionally, the digital twin acts as a tool for comprehensive synthetic testing and Live Virtual Constructive (LVC) testing when integrated with live components, providing a secure and safe platform for UTM concept evaluation [
37].
The paper highlights two main contributions. The first contribution is the detailed design and successful implementation of the Digital Twin for the Airspace of the Future (AoF) project. This digital twin not only emulates the physical characteristics of the airspace, including terrain and built environments, but it also simulates drone flights within this realistically represented environment. Through these simulations, potential issues such as collision risks, optimal flight paths and environmental factors can be proactively identified and mitigated. This comprehensive modelling and proactive problem-solving can lead to safer and more efficient flight plans, greatly improving the operational safety and efficiency of drones.
The second contribution of this research is the practical validation of the AoF Digital Twin through extensive flight trials, encompassing both simulated and real-world drone operations. These trials demonstrated the effectiveness of U-space services in securely and efficiently managing drone flights. They provided insights into the impact of increasing flight submissions on acceptance rates, the need for alternative deconfliction strategies, optimising airspace utilisation, implementing safety measures during takeoff and landing, balancing safety and efficiency and improving data management for high traffic loads. These findings highlight key areas for improvement in drone operations.
Consequently, the AoF Digital Twin serves as a reliable and effective tool for securely and efficiently managing drone flights. It provides real-time monitoring, risk detection and route optimisation, enabling users to enhance safety and maximise operational efficiency. The users can confidently manage their drone operations with improved effectiveness.
In
Section 2, the concept of the AoF Digital Twin is presented.
Section 3 describes the digital twin design. In
Section 4, the proposed use cases are summarised. Flight trials were performed for the digital twin, some of the results are presented in
Section 5 and the results are analysed. Finally some conclusions are drawn.
5. Flight Trials
An imperative step in our research involves advancing from theoretical postulations and thrusting into the intricate practicalities of real-world operations. For this purpose, we orchestrated a series of flight trials, envisaged to serve as an instrument of practical validation for our Airspace of the Future (AoF) Digital Twin. These trials serve to challenge and affirm the robustness of our digital twin model within a range of realistic operational settings, thus enhancing the generalisability and applicability of our findings. Each trial acts as a rigorous proof-of-concept, demonstrating the capacity of U-space services to securely and efficiently manage drone flights.
The trials focused on testing the system behaviour, AoF processes, agreed and available CORUS functionality and performance measures such as air traffic load, traffic proximity etc., when the volume of Drone Operations increased from one to the maximum concurrent flights in airspaces defined during high and low peak periods.
The following trial runs were conducted during flight trials
Trial Run#1: The purpose of this trial run was to test single drone operations in X Airspace.
Trial Run#2: The purpose of this trial was to test three concurrent flights in X and airspace in Synthetic Environment only prior to loading the system with increased number of flight plans.
Trial Run#3: The purpose of this trial was to test 10 concurrent flights in X and airspace in Synthetic Environment at low peak volumes.
Trial Run#4: The purpose of this trial was to test all concurrent flights in airspace in Synthetic Environment at low peak volumes.
Trial Run#5: The purpose of this trial was to test all concurrent flights in X and airspace in Synthetic Environment at high peak volumes.
Trial Run#6: The purpose of this trial was to test all concurrent flights in X and airspace in hybrid environment at high peak volumes.
The analysis of the trials was conducted according to the defined performance requirements. These requirements include the total number of active flight plans, the total number of accepted flight plans, the total number of rejected flight plans, the total number of drone positions and tracks, the number of drones per airspace volume, traffic proximity and message throughput. By examining these performance metrics, we can gain a comprehensive understanding of the system capabilities and identify any areas that require improvement.
5.1. Flight Trial Observations in Synthetic Environment
Observations made only for Digital Twin Simulation and Guardian UTM. The architecture processes developed for each phase of flight were conformant and as expected.
5.1.1. Air Traffic Load for an Airspace
The analysis included trial runs with both flights requiring strategic deconfliction and those in
X airspace not needing it. Trial run #1, consisting of flights solely in
X airspace, was included because it had multiple concurrent flights. As shown in
Figure 19, the Digital Twin system managed flight plans, providing users with approval or rejection of their submitted plans without any details of conflicts.
Table 2 presents the number of accepted and rejected flight plans for each trial run. Flights in
X airspace were considered accepted. Trial runs #1 and #2 had acceptance rates of 100% because they did not require strategic deconfliction and all intended flights were flown.
During trial run #3, 90% of ten flights were accepted, although it is possible that the airspace was not fully loaded. In contrast, trial run #6 had a 24% acceptance rate, indicating that current processes were at capacity. The current deconfliction approach blocks other drones from flying through any point on the reserved path, which has a significant impact on flights from delivery hubs. Both strategic and non-strategic deconfliction flights were conducted in all trial runs. As the number of submitted flights increased, the acceptance rate decreased, with a maximum of 18 accepted flights per hour. Trial run #4 had a 70% acceptance rate, which was lower than trial run #3, due to the randomisation of drone operations and route selection.
To understand the reason for the rejections in trial run #4, a plot of the operations plan was generated, which showed rejected plans in red and accepted plans in blue (
Figure 20). The plot indicated that once a flight departs a given base, no other flight can depart or arrive at the base for the duration of that flight. For example, flights that planned to depart the hub during the duration of an accepted flight were rejected. The trial used a “light goods delivery” use case, where all flight plans started and ended at a single point, the supermarket.
The UTM rejected flight plans if conflicts were identified along the planned route. Trial run #4 had a lower acceptance rate due to conflicts at the start and end points. Alternative deconfliction methods and rule sets could address these issues. Optimised trial run #5 split flights into subsections and required approval for each subsection. Further optimisation of airspace use is possible as even low-density runs had flight plan rejections. Next, safety and efficiency of airspace management will be explored.
5.1.2. Traffic Proximity/Nearest Approach
Airspace management aims to prevent drone collisions by analysing traffic proximity data. Results from post-trial analysis show how close drones come to each other, which is a safety and efficiency metric. Separation that is too great reduces the number of drones flown, potentially failing to meet user demand. Proximity data was obtained through a discrete and approximate method, resampling position data at
s intervals to calculate the distance between concurrent drones. A conflict or incursion occurs if drones come within
or
of each other. Results for each trial run can be seen in
Table 3.
The majority of trial runs showed effective separation of drones with no conflicts within defined thresholds, demonstrating the efficacy of the airspace-management processes. However, Trial Run #4 had a near miss and collision in the horizontal plane, corresponding to
of operations. Vertical separation was
and
, respectively, despite the drones flying at the same cruising altitude, due to altitude changes during takeoff and landing. The affected drones belonged to the same operator and were synthetic without detect-and-avoid. Lack of buffer time for landing in some operation plans likely caused the conflicts.
Figure 21 shows a proximity violation, with drones 1 and 2 having blue and red tracks, respectively, occurring at Base 1. Both drones were concurrent in flight time. Tactical deconfliction was not implemented in the trials.
Therefore, a drone from a previous operation may still be nearby in a location when another drone is about to takeoff. This issue was corrected in later trials by increasing the buffer time at the end of operations. This observation further stresses the need to take extra care during takeoff and landing procedures.
For most of the trial runs, the separation between the drones is large, reaching up to in Trial Run #5. This indicates there is room to optimise the airspace efficiency by exploring more ways to support more drones per unit time. This is especially the case if we consider that some flight plans were rejected even though the separation distance between any pair of the drones was large.
5.1.3. Number of Drones per Airspace Volume
A crucial measure of the efficiency of the airspace management is the number of drones that can be safely flown per airspace volume. By analysing the timestamped position reports, this measure can be obtained for each trial run. That is, for each time interval (one minute in this case), the number of drones in operation in the airspace of interest was counted. The best-case scenario is to keep the number of drones per time close to the maximum capacity that can be supported or the airspace. That way, the airspace is used more efficiently.
Table 4 summarises the relevant results for each of the Phase 2 trial runs.
Trial Run #4 had the most drones, with a support of seven drones per minute, but suffered from reduced separation and a collision between two drones. The highest number of drones with good separation was five drones per airspace in Trial Run #1, which had a large minimum separation of
. Therefore, the optimal number of concurrent drones for good airspace use and sufficient separation lies between Trial Run #4 and Trial Run #5. To analyse the results further, a plot of the number of active drones over time for Trial Run #5 is shown
Figure 22.
The plot reveals that, for most of the time, the number of active drones was lower than the peak capacity of five drones. Ideally, the number of concurrent drones should remain at the peak capacity of the airspace throughout the trial run. Therefore, increasing the number of drones in the airspace per unit time to the peak values for a larger fraction of the trial run would be beneficial. However, this observation is based on the assumption that there was good separation during the trial. To increase the number of concurrent drones, more sophisticated airspace-management strategies could be employed. For instance, flight plans could be segmented and deconflicted, as demonstrated in the optimised Trial Run #5.
5.1.4. Total Number of Drone Positions and Tracks
For these trial runs, the number of drone tracks was equivalent to the number of flight plans submitted. This is because relatively ideal conditions were considered. This meant that most approved operations were flown and were not cancelled.
In addition, the drone positions were regularly submitted to the Altitude Angel Surveillance API. This meant that any stakeholder that has access to the API could be aware of drone traffic in their vicinity. This ensures safety and the drone position reporting service is a requirement of the CORUS-based systems used in this design.
Figure 23 shows a screenshot of the Guardian UTM dashboard displaying some tracks for planned drone operations during trial run #5. This type of information is useful to airspace users and stakeholders for informing adequate drone separations and drone awareness.
5.1.5. Message Throughput
An important interface covered in this analysis was the interface between the GCS and strategic conflict resolution interface (GCS-SCR) through which deconfliction requests are submitted. A summary of the analysis for Phase 2 runs can be seen in
Table 5. The table also shows the results for the interfaces between the GCS and Surveillance API (GCS-SURV) and those between the Remote ID and surveillance API (Remote ID-SURV).
The results for the GCS-SURV interface showed that the message traffic increased as the number of concurrent drones increased. This was expected because the interface handled position reports. This was because each drone transmitted its position regularly and these reports eventually got to the GCS-SURV interface. Note message data logging was not available during run #2. A similar result was obtained for the Remote ID-SURV interface because it also handled drone position reports.
Figure 24 shows an example plot of the variation of the message rate for the GCS-SURV interface (Trial Run #5). The figure could be compared with the respective plot for number of drones in the airspace to confirm the observation that message throughput varied with number of drones in the airspace.
In contrast, the results for the GCS-SCR interface showed a small number of messages. The messages correspond to the number of flight plans submitted to the strategic deconfliction API. Therefore, this interface handled messages mostly before operations and did not experience much loading, compared to the GCS-SURV interface. Because the messages handled by the GCS-SURV and Remote ID-SURV interfaces grew with the number of drones, the interfaces were more likely to be bottlenecks in the LVC (Live Virtual Constructive) environment. This is in comparison with the GCS-SCR interface that had low message load.
5.1.6. Discussion
Based on the conducted trials, it can be concluded that a rise in flight submissions inversely affects the acceptance rate, indicating a need to explore alternative deconfliction strategies and airspace optimisation methods. The near misses and collisions witnessed in Trial Run #4 emphasise the critical need to implement rigorous safety measures during takeoff and landing, particularly with respect to ensuring sufficient buffer times.
Moreover, despite the maximum drone support in Trial Run #4, the occurrence of a collision underlines the necessity to strike a balance between safety and efficiency for optimal airspace utilisation. The robust operation of the Altitude Angel Surveillance API was demonstrated in the successful tracking of drone positions and routes.
Furthermore, the trials revealed the potential for bottlenecks in the Live Virtual Constructive (LVC) environment due to the rise in message traffic correlated with an increase in drone concurrency. These findings necessitate focused attention on improving data management techniques to sustain high traffic loads.
5.2. Flight Trials in Hybrid Environment
The conducted flight trials in a hybrid environment refer to the testing and evaluation of the system and processes in a combination of simulated and real-world operational conditions.
5.2.1. Air Traffic Load for an Airspace
During the hybrid operations at scale run, 75 target operations were conducted, some of which required submission to the strategic deconfliction service due to their flight paths passing through
or
airspace.
Table 6 displays the traffic load results. In trial run #8, only 19 out of 75 operations were flown, resulting in an acceptance rate of 25.33%. This is similar to the acceptance rate of 24% obtained in synthetic run number #5, which also had 75 target flights. The slight difference in acceptance rate may be attributed to the variation in randomised routes between the two runs. For instance, trial run #8 included an operation at Cranfield Airport, which was not present in trial run #5.
The UTM system used in the Airspace of the Future study offered a limited range of services and did not support tactical deconfliction, resulting in the work flow mechanism not accounting for the exact location of each participating drone. Thus, when flight plans were submitted, if the plans physically or time ’overlap’ the first flight plan will be accepted and all other conflicting flight plans will be rejected until the first flight plan has ’timed out’.
5.2.2. Traffic Proximity/Nearest Approach
This analysis aimed to determine the proximity of active drones to each other during the trial run. Three proximity thresholds were used for this analysis:
for collisions,
for near-misses and
for incursions. The results are presented in
Table 7.
The table shows that all drones in the trial run remained within the defined proximity thresholds with no violations recorded. The minimum separation between drones was , indicating safe separation and a low probability of conflicts. This result is similar to that of run #7, which had the same number of drones and no proximity violations, but with a higher separation distance of . However, trial run #8 had a slightly higher number of accepted flights (19) compared to trial run #7 (18), indicating that it could support a higher number of flights without compromising safety.
Given the safe separation distances observed, there may be opportunities to optimise airspace utilisation by increasing the number of drones in the airspace. By doing so, airspace managers can enhance the efficiency of drone operations while maintaining safety.
5.2.3. Number of Drones per Airspace Volume
Figure 25 shows the number of drones per airspace volume as a function of time. it can be observed that the maximum number of drones per minute recorded during this trial run was 5.
Throughout most of the run, the number of drones in the airspace was below its peak capacity of 5. Ideally, the airspace utilisation should be kept as close to its peak as possible to support more drones per unit time, thus increasing the number of services and operations available to airspace users. For example, if the number of concurrent drones was maintained at 5 per minute at all times, the drone traffic would increase to 300 drones per hour (5 multiplied by 60).
However, while large separation distances ensured safety during the run, they may have had a negative impact on airspace efficiency. This highlights the need for more CORUS services such as dynamic deconfliction and Detect and Avoid (DAA) to improve efficiency in the future.
5.2.4. Total Number of Drone Positions and Tracks
Like previous runs, the number of drone positions and tracks was equivalent to the number of flight plans, as discussed in the Air Traffic Load for an Airspace section above. This was because only the ideal (“sunny day”) conditions were within the scope of these trials. Therefore, there were no cancellations and so on.
5.2.5. Message Throughput
Table 8 summarises the message throughput analysis for important interfaces in the Digital Twin.
For the GCS-SCR interface, the throughput was low with just 38 messages sent per minute. This is because the interface was used to handle the few flight plans that needed strategic deconflictions. This process is usually done before operations. Thereafter, the interface was not needed. On the other hand, the GCS-SURV interface handled a larger number of messages throughout the trial run with a peak message rate of 48 messages per minute. The results indicate that the GCS-SURV interface was more likely to be a bottleneck compared with the GCS-SCR interface. The GCS-SURV interface would, therefore, require more resources and redundancy in operational deployments compared with the GCS-SCR interface.
Figure 26 showed the plot of the message throughput variation with time. The message throughput for the GCS-SURV interface increased as the number of drones in the Digital Twin increased. This can be seen by comparing the figure with the number of drones in the airspace. The increase is because each drone transmits its position periodically and this data are then sent by the GCS to the SURV interface (surveillance API). Therefore, the message rate for a single drone is multiplied by the number of drones in the airspace to give the total throughput.
5.2.6. Discussion
The flight trials in a hybrid environment provided valuable insights into system functionality and efficiency. With an acceptance rate of 25.33% (19 out of 75 operations flown), the trials revealed limitations of the existing UTM system, emphasising the need for tactical deconfliction. Safe drone separations of at least suggested potential for optimising airspace utilisation. The peak capacity of five drones per minute highlighted the importance of dynamic deconfliction and Detect and Avoid (DAA) strategies for improved efficiency. Matching drone positions to flight plans indicated ideal conditions without cancellations, while message throughput analysis identified the GCS-SURV interface as a potential bottleneck compared to the GCS-SCR interface.