*2.3. Digital Twins*

The digital twin concept was first used in the manufacturing literature in 2010 as "a digital representation of an asset (e.g., physical objects, processes, devices) containing the model of its data, its functionalities and communication interfaces" [111], providing the elements and dynamics of asset operation throughout its life cycle [112]. Various DT definitions exist in the current literature depending on the domains and industries [113]. A list of DT definitions based on domains is as follows:


DTs in various industries have approximately the same features and application purposes. The main components for generating DT models are physical elements/assets, linked data, and virtual models [113]. DTs can be categorized as follows:


Static DT is the simplest way of implementing DT, and dynamic DT is the most complex one. As the level of details and information increases, the complexity and cost of DTs increase. Figure 6 presents the relationship between DTs and business value. Code green is simple design data, code yellow is the design and manufacturing data, and red is the dynamic DT that also includes operational field data [117].

Figure 7 illustrates the DT complexities (three main complexity levels) and time horizon approximations (three main life cycle stages of a physical system with the related DT applications) [119].

**Figure 7.** Digital twin complexities and time horizon approximations [119].

Before developing and implementing the DT, various research questions must be answered. Semeraro et al. [120] presented Table 3 to summarize the key research questions of DTs answered by the literature so far.


**Table 3.** List of digital twin research questions [120].

It is important to distinguish between the concepts "digital twin", "digital shadow", and "digital model". Figure 8 highlights the differences in these concepts by focusing on the data transfer among physical and virtual twins [119].

**Figure 8.** Data transfer comparison between physical systems and virtual models in a digital twin, digital shadow, and digital model [119].

Figure 9 demonstrates the important risks and challenges when developing DTs [117]. Modeling a digital copy of a physical system to perform real-time validation and optimization is a complex task as it involves sensors, multifunctional models, multisource data, services, etc. A DT requires an accurate model of reality and a large amount of data. It can potentially be used in life cycle assessments; however, the development of standardsbased interoperability is important and challenging for evaluating DT applications along the entire life cycle. A few contributions also focused on DT applications for improving sustainability performance [120].

**Figure 9.** Digital twins' risks and challenges [117].

To comprehensively understand the state of knowledge on the application of DTs in UASs, as well as their benefits and challenges, a synthesis of the literature that integrates various subtopics is crucial. The implementation of DTs has been widely explored in aviation-related scientific literature. For example, the EU-funded project Secure Urban Air Mobility for European Citizens (AURORA) is planning to develop and integrate safetycritical technologies to support autonomous UAS flights in urban environments. Figure 10 presents examples of DT applications in this project [121].

**Figure 10.** The EU-funded AURORA project: (**a**) digital representation of the rotor (digital twin for manufacturing and digital certification) and (**b**) digital representation of the flight path (digital twin for solutions testing) [121].

(**b**)

The German Aerospace Center (DLR) has also established an internal project to identify techniques, technologies, and processes for DTs [44]. Liu et al. [45] reviewed the overall framework for creating a DT in combination with the industrial Internet of things (IIoT) to enhance the autonomy of aerospace platforms. Liao et al. [46] presented the findings of research conducted at the National Research Council of Canada (NRC), which included a review and evaluation of DT concepts and digital threads, particularly the airframe digital twin (ADT) framework used by the United States Air Force (USAF), as well as a feasibility and adaptability study of the ADT for use with Royal Canadian Air Force (RCAF) aircraft. Aydemir et al. [47] reviewed the available approaches, technologies, and challenges of DTs for aircraft applications. Mendi et al. [48] evaluated DT applications and their advantages in military aviation. Ibrion et al. [49] presented DTs' risks and challenges in the marine industry by learning from the aviation industry. DTs can be effectively utilized in any stage of the aircraft life cycle, encompassing the design, manufacturing, operation, and maintenance phases. DTs enable engineers to create virtual prototypes and simulate various scenarios, allowing for the efficient optimization of aerodynamic performance, structural integrity, and overall aircraft functionality in the design stage. DTs can facilitate real-time monitoring and quality control, ensuring that components are produced to precise specifications and tolerances during manufacturing. DTs, based on their level of complexity, have the potential for real-time data collection and analysis, offering insights into the operation phase, including aircraft performance, fuel efficiency, and operational safety. DTs can also support predictive maintenance by continuously monitoring the health of aircraft systems and components, as well as detecting potential issues before they lead to failures or disruptions. Leveraging DTs throughout the aircraft life cycle can enhance decision making, improve safety, reduce costs, and ultimately maximize the overall performance and lifespan of the aircraft. For instance, Tuegel et al. [50] proposed the airframe DT structural modeling concept to design and maintain airframes (which has the potential to improve US Air Force aircraft management over the life cycle) by creating a tail number computational model and structural management plans for each aircraft. Seshadri et al. [51] suggested employing DTs to manage the structural health of damaged aircraft using guided wave responses. A genetic algorithm (GA) optimization evaluates the cumulative signal responses at preselected sensor locations to estimate the size, position, and orientation of the damage. Mandolla et al. [52] implemented a DT for additive manufacturing in the aerospace industry by utilizing blockchain solutions. This work highlights how businesses utilizing the blockchain can create secure and connected manufacturing infrastructure and provides a conceptual solution to securing and organizing the data generated by an end-to-end additive manufacturing process in the aerospace industry. Zhang et al. [53] established a digital-thread-based modeling digital twin (DTDT) framework for an aircraft assembly system, enhancing the controllability and traceability of the manufacturing process and product quality through improved data management. Tyncherov et al. [54] proposed DT modeling of aircraft operational life cycle by presenting aircraft systems' DTs with operational and maintenance environments as a cloud of data considering machine learning (ML) methods to improve prediction and planning accuracy. Tuegel et al. [55] reengineered the aircraft structural life prediction process to high-performance digital computing, presenting a conceptual model of DTs for predicting aircraft structure life and assuring its structural integrity. Ríos et al. [56] discussed an aircraft avatar implication through an industrialization-focused perspective while reviewing the various topics involved in an aircraft's digital counterpart development (i.e., product identification, product life cycle, and product information). Strelets et al. [57] created a DT in a uniform information environment of the product life cycle, which, as the virtual copy of a product, is convenient to use at all stages of the life cycle. Liang et al. [58] presented a real-time displacement detection DT in aircraft assembly. Zhang et al. [59] proposed an effective simulation and optimization containing heuristic algorithms and applied them to a DT-based aircraft part production workshop. Singh et al. [60] presented an information management (IM) framework for DTs in aircraft manufacturing, with a case

study for aircraft structure damage tolerance, demonstrating the different phases of IM (from identification to retrieval and retention).

The existing body of aviation-related scientific literature extensively explores the potential of DTs and highlights their versatile applications, including their effectiveness not only in system-level implementation but also at the individual-component level. Employing DTs at these different levels can unlock new insights and ultimately advance the state of knowledge in the field. For example, Lei et al. [61] modeled a DT for tooth surface grinding, considering the low-risk transmission performance of non-orthogonal aviation spiral bevel gears. Zakrajsek et al. [62] developed a DT for a specific aircraft tire at touchdown to improve tire touchdown wear prediction. Xu et al. [63] suggested DT optimization with several DT modules for a system to virtually simulate as well as optimize the parameters, performance, and manufacturing. The DT modules make corrections during the optimization using real-time feedback data from manufacturing measurements and performance testing. Borgo et al. [64] presented a DT of a ground steering system and systematically analyzed the effect of uncertainties and sensor faults with estimation algorithms (least squares estimation and soft computing approach) under several scenarios. Hu et al. [65] developed a DT decision-making approach to generate reconfigurable fixturing schemes optimization for the trimming operation of aircraft skins. Peng et al. [66] provided an online fault diagnosis system for the TFE-731 turbofan engine and used model-based and data-driven approaches to create DTs of the engine parameters. Li et al. [67] used the concept of dynamic Bayesian networks (DBNs) to develop a health monitoring model for aircraft. An example of the proposed method is also illustrated on an aircraft wing's fatigue crack growth [68]. Kosova et al. [69] developed a DT and used ML for a health-monitoring system (limited to aircraft hydraulic systems) to diagnose system failures in the early stages using 20 failure scenarios. Laukotka et al. [70] implemented DTs for civil aviation, aircraft, and aircraft cabins, based on modular product family design and model-based systems engineering.

Various research efforts have been diligently conducted to explore and harness the potential of DTs in UASs. The application of DTs in UASs has emerged and prompted researchers to utilize the benefits of this technology, aiming to enhance design, operation and mission planning, and maintenance practices, leading to more reliable, efficient, and capable UASs. However, after reviewing DTs throughout the entire life cycle of the aviation system, Xiong et al. [71] concluded that while aviation DTs are frequently utilized in manufacture and maintenance, more effort and attention are required for UAV DT applications. Lv et al. [72] also reviewed AI applications in DTs in aerospace, intelligent manufacturing, unmanned vehicles, and smart city transportation. Salinger et al. [73] presented a hardware testbed for a self-aware UAV to advance dynamic data-driven application system (DDDAS) development. Self-awareness refers to a vehicle's ability to collect information about itself and utilize that knowledge to complete missions through dynamic decision making on board. Kapteyn et al. [74] combined reduced-order models with Bayesian estimation to create a data-driven DT for a 12 ft wingspan UAV to enable the aircraft to adjust its mission plan in the event of structural damage or deterioration. The authors further advanced the methodology using interpretable ML [75]. Alaez et al. [76] modeled a DT of a VTOL UAV using the Gazebo robotics simulator, compared the UAV's take-off, hovering, and landing operation with and without a wind physics model, and tested it in different wind speeds and directions. Yang et al. [77] proposed a DT for a multirotor UAV with a simulation system, a physical UAV, and a service center for advanced capability training as well as algorithm verification. The authors also demonstrated a DT simulation platform for verification that further simulates and tracks the life cycle of a multirotor UAV [78]. Lv et al. [3] analyzed the effects and limitations of UAVs in 5G/B5G wireless communication and developed a UAV DT 5G communication channel model using deep learning (DL) to further reduce UAV limitations. Moorthy et al. [79] designed a UAV network simulator focusing on high-fidelity UAV flight control by using two simulators they developed in prior years: UBSim (a Python-based event-driven simulator) and UB-ANC (a simulation

framework used to design, implement, and test various UAV networking applications). Wu et al. [80] addressed the security concerns that arise when a drone system is attacked and investigated the computational intelligence of drone information systems and DTs of drone networks based on DL. Shen et al. [81] proposed a DT with deep reinforcement learning (DRL) (in which a DT of a multi-UAV system is built into a central server to train a DRL model) to solve the flocking motion problem of multi-UAV systems. Lv et al. [82] developed a UAV DT to provide medical resources quickly and accurately to analyze the feasibility of UAV DTs during COVID-19 prevention and used DL algorithms to construct a UAV DT information forecasting model. Fraser et al. [83] used DT and data-driven approaches to investigate the general susceptibilities of UAVs against contemporary cyber threats. Kapteyn et al. [84] suggested a probabilistic graphical model representing the DT and its physical asset for a UAV using experimental data to calibrate the DT. The UAV encounters an in-flight damage event and the DT is updated using sensor data. Riordan et al. [85] presented a DT to evaluate UAS-mounted LiDAR ability to detect small-object air collision risks, considering the Hamburg port with its aerial hazards (e.g., birds, drones, helicopters, and low-flying aircraft). Iqbal et al. [86] presented a DT with a runtime trust assessment for an autonomous food delivery drone system to evaluate the trusted execution of intelligent agents (autonomous drones or other vehicles). Grigoropoulos et al. [87] employed DTs and simulations to support offline validation and runtime checking in a platform as a service (PaaS) system for drone applications. Lee et al. [88] proposed a DT with a model-based system engineering methodology for a UAS capable of route selection in a military case study, where the route optimization module suggests an optimal path based on inputs such as potential damage. Lei et al. [89] created a DT to define the physical entity of a UAV swarm and track its life cycle. The UAV swarm's behaviors are investigated using an ML-based decision model. Wang et al. [90] combined DTs and convolutional neural networks (CNNs) for a UAV autonomous network to explore the airspace structure and safety performance of the UAV system. The presented literature emphasizes the significance of exploring and utilizing DTs in UASs. These case studies highlight the significance of DTs in addressing various challenges and opportunities of UASs associated with topics such as driving technological advancements, decision-making processes, and operational efficiency within this dynamic and evolving field. Digital twin technology has the potential to address some of these challenges and complement existing measures in UAV management. By modeling a digital copy of UASs and their operational environment, DTs can provide real-time monitoring, analyses, and optimization of UAS operations. This can enhance situational awareness, enable predictive maintenance, improve traffic management, and support decision-making processes. DTs can also facilitate data integration and interoperability across different systems, enabling a more comprehensive and coordinated approach to UAV management. However, it is important to note that DTs are not a standalone solution but should be integrated into a holistic framework that considers regulatory, technical, and operational aspects. Overall, the unique role of using a DT to facilitate UAV certification and regulation lies in the ability to model a digital copy of a physical system for real-time validation and optimization. However, this task is inherently complex and presents several challenges, as depicted in Figure 9, which offers an overview of the risks and challenges associated with the overall DT process. One challenge is the requirement for an accurate model of reality, which necessitates a deep understanding of the physical system and its operational characteristics. Additionally, as demonstrated in Figure 8, the creation of DTs necessitates the transfer of data between a physical system and a virtual model. Depending on the complexity level of a DT, this process involves handling a large amount of data from various sources, including sensors and sometimes even multifunctional models. Ensuring the accuracy and reliability of these data is crucial for the effectiveness of a DT. Furthermore, integrating a DT into UASs to assist the certification process requires careful consideration of legal and regulatory requirements. These challenges highlight the need for careful planning, robust data management, and

close collaboration between experts in UAV certification and DT technology to successfully utilize DT in the context of UAV certification.

#### **3. Results**

UAVs are becoming popular. Autonomous (artificial intelligence applications) and automatic UAVs are expected to conduct safe operations, and they will enter UAM to transport goods and individuals in the near future. A wide range of literature is published to answer the research questions of "how to adapt UAV applications to regulations" and "how to adapt DT applications to UAV". However, it is fair to state that there is not much literature considering the use of DT applications in UAVs for certification and regulation. This lack of literature is inevitable in the early stages of new, emerging concepts. In order to fill this gap, we conducted a literature review considering a total of 121 references. Table 4 provides a comprehensive collection of references along with the keywords that are closely aligned with our research concepts. They serve as concise descriptors that capture the essence of the paper's content and help identify its key focus areas. The inclusion of these relevant keywords allows for a focused exploration and clear navigation of the existing literature, facilitating the identification of common themes, connections, and relationships across the literature. By including associated keywords in the table, we aimed to provide additional information and context about the content of each reference. We have systematically identified and classified the references into key focus areas: DTs, general aviation, UAVs/UASs, UAM/AAM, and regulation. By organizing the references under these categories, the table allows for a clear understanding of the primary themes and topics covered in each reference, enhancing the clarity and structure of our research with a more organized exploration. While the references consider multiple topics and overlap across the key focus areas, we have made an effort to present the primary purpose of each paper and provide associated keywords to highlight key themes and connections that contribute to a more comprehensive understanding of our research concepts and emphasize the various aspects explored in the literature.

**Table 4.** Compilation of references and their associated keywords relevant to our research concepts.





