A Critical Review of Information Provision for U-Space Traffic Autonomous Guidance

Abstract

This paper identifies and classifies the essential constraints that must be addressed to allow U-space traffic autonomous guidance. Based on an extensive analysis of the state of the art in robotic guidance, physics of flight, flight safety, communication and navigation, uncrewed aircraft missions, artificial intelligence (AI), social expectations in Europe on drones, etc., we analyzed the existing constraints and the information needs that are of essential importance to address the identified constraints. We compared the identified information needs with the last edition of the U-space Concept of Operations and identified critical gaps between the needs and proposed services. A high-level methodology to identify, measure, and close the gaps is proposed.

1. Introduction

The Single European Sky Air Traffic Management Research Joint Undertaking (SESAR JU) defines U-space as “a set of new services and specific procedures designed to support safe, efficient, and secure access to airspace for large numbers of drones” [1]. SESAR JU expects that U-space services will be supported by advanced automation functions and digitalization in multiple autonomous unmanned aerial vehicles (UAVs) [1].

Unmanned aircraft system traffic management (UTM), named “U-space” in Europe, is a new paradigm of Air Traffic Management (ATM), characterized by a cooperative approach, which provides safe and efficient services for UAVs to allow their operations at the Very Low Level (VLL) [2]. While ATM is a system with a key human role, UTM is based on computing infrastructure with high automation for managing UAVs’ operations. “Ultimately, U-space will enable complex drone operations with a high degree of automation to take place in all types of operational environments, including urban areas” [1].

D.M.K. Zoldi et al. argue that humans can be superseded in any respect of drone operations by machines with further unmanned aircraft system (UAS) autonomy development [3]. A flight without human control requires that a U-space traffic autonomous guidance system is in place. The system can be centralized, decentralized, or hybrid [4,5]. The fundamental difference between these three types is that a centralized system provides a single entity with the functions of management and control over uncrewed air traffic. The decentralized system relies on multiple entities. The hybrid system combines both, depending on the area of operation. For instance, the highly dense uncrewed air traffic may benefit from local centralization, but for remote areas, a decentralized approach can be preferable. The centralized traffic management system needs significant infrastructure, but it allows for the optimization of traffic as a holistic system. A decentralized system has much less demand for infrastructure, but its optimization potential is limited. The authors of the European project “Metropolis 2” suggested the hybrid approach since it is more effective in preventing conflicts from happening rather than trying to resolve them after they occur [5]. From our viewpoint, the hybrid system is highly possible to expect in the future due to its flexibility to address various constraints. Our analysis in this article is accompanied by comments when it is important to note that this information is necessary only for a specific type of system (centralized, hybrid, or decentralized). All other information needs described in the article are universal for all system types. In all cases, it should manage unmanned aircraft 4D (three dimensions and time) trajectories in real time and without human intervention. The decision-making process [6] is a significant part of it, which requires relevant information provision and a combination of AI-based algorithms with, e.g., machine learning (ML).

Eurocontrol predicts significant AI integration in autonomous aviation [7]. We also find that AI implementation will play a crucial role in autonomous guidance. For example, situation awareness is hardly possible without image recognition [8] for many types of missions (surveillance, inspection, etc.). Path planning algorithms, which are a part of AI science, are essential to guide UAVs autonomously in dynamically changing airspace [9]. However, the complexity of some missions and the currently limited progress of AI [10] will imply a step-by-step substitution of human functions with machines [11].

Autonomy in various systems and functions within U-space will first emerge in services where it is easiest to deploy. For example, equal access to airspace relies on a priorities policy that normally does not require complicated decision-making, and thus AI can plan uncrewed traffic according to a priority algorithm without human intervention. Another example is the daily delivery of medical samples by UAVs from one hospital to another. The flight route remains mostly constant, and such an operation is not unique. These missions can be automated more easily than a unique operation like inspecting a new bridge with a drone or pizza deliveries to various addresses in highly congested uncrewed air traffic. The strong side of AI is its reliability in constant and repeatable operations. However, new tasks, new environments, and new conditions can be challenging for modern autonomous systems. Nevertheless, the borders of autonomy implementation in the U-space will change over time until humans’ roles are reduced to the consumption of UAV services.

U-space traffic autonomous guidance will likely rely on robotic navigation, which is a well-developed discipline. The usage of robotic navigation in airspace is just a case of operating in a 3D (three-dimensional) environment [12] with the constraints described in this paper (see

Table 1. Physics of flight constraints.

Physics of Flight Constraints	Needed Information	U-Space Service	Note
UAV performance	UAV performance	✓Common Information Service	U3
	Angles during flight	-	Gap
Battery charge/fuel available	Available energy on board	-	Gap
Weather and turbulence	Visibility	✓Weather Information ✓Monitoring	U2
	Ordinary and hyper-local
	wind velocity, direction, gusts
	Ordinary and hyper-local
	precipitation
	Temperature
	Humidity
	Atmospheric pressure
	Natural turbulence map	-	Gap
	Area and time of operation for heavy aircraft generating wake vortices at VLL	-	Gap
	UAV wake vortex category	-	Gap
	Lightning strike threat areas	-	Gap

,

Table 2. Trajectory computation constraints.

Trajectory Computation Constraints	Needed Information	U-Space Service	Note
Time complexity	Selection among algorithms to plan a trajectory with a suitable time complexity for real-time operations	-	Gap
4D trajectory optimality	Drone user’s preferences in 4D trajectory planning	-	Gap
Scalability, adaptability, learning capability, robustness	Selection among software solutions to allow suitable scalability, adaptability, learning capability, and robustness	-	Gap

,

Table 3. Collision avoidance constraints.

Collision Avoidance Constraints	Needed Information	U-Space Service	Note
Static obstacles	Static obstacles’ location and height	✓Geographical Information Service	U2
	Terrain map with altitudes	✓Drone Aeronautical Information Management	U1
Dynamic obstacles	Coordination with ATM for flight in controlled airspace	✓Procedural Interface with ATC	U2
	U-space traffic with 4D trajectories or temporarily occupied airspace	✓Traffic Information	U2
		✓Strategic Conflict Prediction
		✓Strategic Conflict Resolution
		✓Monitoring
		✓Dynamic Capacity Management	U3
		✓Tactical Conflict Resolution
		✓Tactical Conflict Prediction
	ATM traffic with 4D trajectories	✓Collaborative Interface with ATC	U3
	Position and classification of on-surface dynamic obstacles	-	Gap
	Position and classification of airspace intruders and wildlife	-	Gap
Separation	Demands to separation	✓U-space Separation Management Service	Potential U-space service (proposed by the BUBBLES project)
Obstacle uncertainty	3D map of known and unknown environment	-	Gap
Vertiport availability	Vertiport capacity	✓Vertiport Resource Allocation Management	U3
		✓Vertiport Dynamic Information Service
	UAV flight characteristics	-	Gap
	UAV size	-	Gap
	Requirements for charging and fueling	-	Gap
	Ground handling needs	-	Gap
	UAV noise category	-	Gap

,

Table 4. Communication navigation surveillance constraints.

Communication Navigation Surveillance Constraints	Needed Information	U-Space Service	Note
Communication network availability, coverage, and load	Map of communication network availability at VLL	✓Communication Coverage Information	U2
	Status of communication network availability at VLL	✓Communication Infrastructure Monitoring
Electromagnetic interference	Electromagnetic interference	✓Electromagnetic Interference Information	U2
Navigation network availability	Map of GNSS coverage and navigation network availability	✓Navigation Coverage Information	U2
	Status of link with navigation networks	✓Navigation Infrastructure Monitoring
Aircraft position	Real-time UAV position	✓Tracking	U2
		✓Vertical Alert Service	U3
		✓Vertical Conversion Service	U3
Surveillance for guidance	Surveillance data	✓Surveillance Data Exchange	U2

,

Table 5. Institutional constraints.

Institutional Constraints	Needed Information	U-Space Service	Note
Registration and identification	UAV registration and identification	✓Registration	U1
		✓Network Identification
Airworthiness	Status of airworthiness	-	Gap
Emergency	Emergency status of UAV (including status of onboard systems)	✓Network Identification	U1
		✓Emergency Management	U2
Risk assessment	Flight risk evaluation	✓Operation Plan Preparation/Optimisation Service	U2
		✓Risk Analysis Assistance
		✓Flight Authorisation Service
	Population density map	✓Population Density Information	U2
	Location of suitable landing areas in case of emergency	-	Gap
	Runway and Vertiport surface conditions	-	Gap
	Incident and accident data, legal recording	✓Legal Recording	U2
		✓Incident/Accident Reporting
		✓Digital Logbook
Geofencing	Restricted areas’ coordinates and duration, controlled airspace map, NOTAM	✓Geo-awareness (Geo-fence Provision)	U2
		✓Geo-awareness	U1
The spatial limits	The spatial limits of the U-space airspace	✓Common Information Service	U3
Security	UTM security breakthrough (status and level of threat)	-	Gap
Noise reduction	Map of noise-sensitive areas	-	Gap
Regulation demands	Rule awareness in terms of robotic algorithms	-	Gap

and

Table 6. Mission type constraints.

Mission Type Constraint	Needed Information	U-Space Service	Note
Delivery Transportation Inspection Surveillance Photography Search and rescue Watering, sowing, and spraying Filming Mapping Communication network deployment UAV shows Leisure flights with passengers	Take-off and landing locations	✓Operation Plan Preparation and Optimization Service	U2
	Classification/identification and coordinates of object/subject/area of interest	-	Gap
	Applicable in case of dynamic missions
	Mission’s purpose and specialty inputs, matching with available algorithms	-	Gap
	Applicable in case of dynamic missions

). To tackle the constraints, it is essential to collect the corresponding information. This issue has been partly solved or already included in the deployment phases of U-space services. For example, the Drone Aeronautical Information Management Service provides terrain maps with altitudes to aid with addressing the issue of static obstacles. However, there are still multiple gaps that the current paper highlights with an analysis.

The European consortiums DREAMS (DRone European Aeronautical information Management Study) and IMPETUS (Information Management Portal to Enable the integration of Unmanned Systems) have contributed significantly to investigating the information requirements for U-space [13,14]. The required information includes aeronautical, meteorological, terrain and obstacles, surveillance, communication, and other information [15]. Both studies were highly influenced by the ATM heritage, where manned aviation has a prime role. However, the information needs of U-space traffic autonomous guidance were not addressed completely.

There are four phases of U-space development: U1, U2, U3, and U4 (see Figure 1) [16]. It is expected that at each stage, the level of connectivity and automation will be increased. The first three phases U1–U3 are well defined already [17]. U1 is a set of foundational services, which include Registration, Network Identification, Drone Aeronautical Information Management, and Geo-awareness. The U2 phase unites initial services such as Tracking, Weather Information, Operation Plan Preparation/Optimization Service, etc. The U3 phase should deliver suitable services assisting with more complicated operations specifically in dense zones where capacity management and detect and avoid (DAA) will play a significant role. U4 is expected to include full services; it is the final phase with the highest level of automation and digitalization. The set of U4 services has not been defined yet. However, this paper contributes to establishing a significant basis for U4 understanding and formalization by identifying essential information provision gaps that will be unavoidable in defining and developing U4.

For a more informative analysis, it would be better to compare how U-space [17] and Federal Aviation Administration (FAA) UTM [18] address the information needs. However, such a detailed study is beyond the scope of this article. Generally speaking, both approaches are founded on the services, procedures, legal framework, and infrastructure to ensure flight safety and efficiency for UAV operations in uncontrolled and controlled airspace for Visual Line of Sight (VLOS), Beyond the Visual Line of Sight (BVLOS), and Extended Visual Line of Sight (EVLOS) operations [17,18,19].

U-space Concept of Operations (ConOps) defines three categories for U-space services: mandated, recommended, and optional. Similarly, FAA UTM proposes the following categories for the services: required to be used, may be used, and add assistance to a drone operator. U-space and FAA UTM rely on a Remote Pilot in Command (RPIC) to operate the UASs with established responsibilities. Manned aviation has a higher priority than uncrewed air traffic. Both concepts identify the three layers of separation: strategical, tactical, and collision avoidance, where DAA equipment plays a significant role. A continuously accessible information network assists in airspace coordination for U-space and FAA UTM.

From a general perspective, it is possible to state that both approaches propose comparable services. For instance, U-space includes a collaborative interface with Air Traffic Control (ATC), while FAA UTM addresses a similar function with the Flight Information Management System (FIMS). Other examples of commonality of the services are weather, registration, e-identification, surveillance, monitoring, geo-fencing, terrain data, aeronautical information, data regarding mission intent, etc. Both systems will be interconnected with the Urban Air Mobility (UAM) airspace system [20,21]. In case of emergency, the drone operator must announce the issue immediately to the National Aviation Authority (NAA) and other relevant airspace users.

U-space will be deployed in four phases: U-space foundation services (U1), U-space initial services (U2), U-space advanced services (U3), and U-space offering full services (U4). Similarly, FAA UTM uses a spiral concept where the services will be deployed according to the complexity of the operations—the less complex first.

The last major similarity is that both approaches plan UAV operations on a VLL. FAA UTM defines the height Above Ground Level (AGL) as 400 ft. For U-space, the height is defined differently: 120 m above the Take-Off point.

Considering the operational risk level, U-space defines three categories for operations: open, specific, and certified. Instead, FAA UTM relies on performance authorization provided by UTM participants. This authorization guarantees that a particular operation complies with safety and regulatory requirements.

This article is organized in the following way. In the first section, we start with an introduction. The second section proposes an analysis of the existing constraints, corresponding information needs, which must be addressed to tackle the constraints, and the information needs’ matching the U-space services. The analysis identifies the existing gaps, which must be filled to allow the U-space traffic autonomous guidance.

In the third section, we propose a prioritization of the identified gaps that reflect the different levels of necessity for further research and development.

The fourth section contains high-level methodology to identify, measure, and close the gaps. The fifth section outlines the authors’ vision of promising approaches and technologies that have the potential to assist in filling the extant information gaps. Finally, we offer some conclusions and avenues for further research.

The main contributions of the paper are the identification of important constraints, the classification of essential information required, and the identification of information provision gaps, achieved by comparing the information requirements with Eurocontrol’s Concept of Operations [17] to enable UAV autonomous guidance within U-space.

2. Constraints, Information Needs, U-Space Services

This section is divided into six groups of constraints: flight physics, trajectory computation, collision avoidance with static and dynamic obstacles, communication navigation surveillance, institutional ones, and mission type. While the division is conditional, it reflects the impact of the problem statements obtained from different areas of knowledge and sciences. For instance, flight physics imposes unique constraints on the aircraft, while trajectory computation relies on optimization techniques from pathfinding and computer science.

2.1. Flight Physics

2.1.1. Aircraft Performance and Maneuverability

A UAV is a robot that operates in a 3D environment bearing constraints of the physics of flight that any aircraft has. For example, fixed-wing aircraft are limited in their performance (required take-off/landing distance, velocity, fuel efficiency, ceiling, range, climb rate, and controllability speeds). Aircraft also have limitations regarding gross weight, maximum horizontal velocity, minimum velocity for fixed-wing (due to stall threat), and maximum descent velocity due to flatter hazard [22]. In practice, it means that geometrically the shortest path cannot be feasible at all for most fixed-wing aircraft if the trajectory planner proposes too high pitch angles, for example. Additionally, aircraft normally have limitations regarding turning radius, flutter susceptibility, load factor, and payload [23,24]. A rotorcraft has similar limitations, but it does not have a minimum speed, since it can hover. Also, a rotorcraft needs a suitable take-off/landing area instead of a runway. Typically, the UAVs have an onboard energy storage, such as batteries or fuel tanks [25]. This fact implies a limitation on the flight time with various regimes of flight, including climb, flight with maximum speed on constant altitude, maximum endurance, gliding or hovering (for rotorcraft), descent, etc. The Common Information Service will address this issue in phase U3 [17].

To plan a feasible 4D trajectory in U-space, there is a need for uncrewed aircraft performance databases for all models of UAVs that are presented in U-space traffic [9]. In Europe, a similar issue has already been solved for commercial aviation for Air Traffic Control purposes via Base of Aircraft Data (BADA) and Trajectory Computation Infrastructure software [26,27]. U-space deals with a more complicated environment since UAVs do flights at VLL with a huge diversity of potential static and dynamic obstacles. Therefore, to manage more complicated environments, it is necessary to collect various VLL-related data, which we analyze hereunder.

For flight safety matters, it is essential to control the aircraft’s angles (pitch, roll, yaw) and flight status information (altitude, speed, direction) during the flight to avoid stalling [13]. A 4D trajectory planning system must consider UAVs’ limitations with pitch and roll angles. Additional complexity comes with the fact that all models of aircraft differ concerning the aforementioned characteristics and limitations due to diversity in aircraft design, weight, and propulsion systems [28].

The shortest path does not always satisfy fuel and time efficiency. For example, the best climb rate for a fixed-wing aircraft means maximizing aerodynamic efficiency (maximum value of lift and drag coefficient ratio) [29]. In other words, a climb with maximized aerodynamic efficiency gives the best time for reaching the needed altitude.

2.1.2. Weather Conditions

S. Cambel et al. significantly contributed to the analysis of weather information needed for UAS operations [30]. Weather, for example, the wind’s vector and wind gusts, temperature, humidity, air pressure, and precipitation can heavily influence the physics of flight. T. Bonin et al. proposed weather conditions expected to affect UAM, among them: icing, temperature, reduced visibility and low ceiling, turbulence, wind gusts, urban canyons, wind shifts, updrafts and downdrafts, wind shear, precipitation, convection, and storms [31]. Generally, our literature review analysis conceptually confirmed their findings; however, we propose a slightly different terminology, and we also find it essential to recognize the lighting threat.

Temperature and air pressure directly influence the key aerodynamic forces—aircraft drag and lift [24]. The presence of humidity influences air density as well; the higher the humidity, the lower the air density [24].

With high atmospheric temperatures, the aircraft’s rate of climb capability will be significantly decreased, which must be taken into consideration to process the 4D trajectory. U-space must collect weather data in order to transmit it to an autonomous guidance system, regardless of whether this system is part of U-space or not.

In the worst scenario, weather affecting the flying conditions can temporarily close the airspace for flights. The traditional global meteorological models for manned aviation do not provide hyper-local weather information [13]. In the case of U-space ordinary and hyper-local weather information, it is essential to plan and execute flights safely; however, hyper-local ones still do not exist as a widely available service [15]. In any case, the weather susceptibility category of UAVs is essential to plan missions safely under various weather conditions. Hyper-local velocity, direction, and the gust of wind generate an additional vector of force that can influence the aircraft [24]. Strong wind can make the flight too risky, inefficient, or even impossible. For example, crosswinds can provide a significant impact on the landing for large UAVs on runways [30]. To collect hyper-local weather information, the current methods include weather microstations, mobile and portable sensors attached to transport and infrastructure, traditional weather stations, satellite data, predictive weather models, and ML and AI algorithms to predict weather changes. However, the current coverage of the hyper-local weather information collection is still at a very early stage [32].

Aircraft icing hazard threatens flight safety in case of operations near and below freezing temperatures [33]. The icing of wings and stabilizers, sensors, propellers, aircraft mechanization, aircraft control, and other moving parts is extremely dangerous and can lead to catastrophic results [34].

Visibility can be essential for robotic navigation if video cameras [35] are used for situational awareness [36]. In the case of surveillance, search and rescue (SAR) missions, inspection, etc., the level of visibility can affect the UAV 4D trajectory.

Precipitation affects landing surfaces and aircraft flight aerodynamics. For instance, heavy rain distorts the upper wing surface’s shape which can reduce total lift force by up to 30% [24].

Lightning strike threat becomes important when an aircraft is flying near the area of charge in thunderstorms or volcanic ash clouds, so UAVs can initiate a discharge [37] and be damaged. Metal protection is commonly used in the aircraft industry to protect safety-critical parts from critical damage [38]. However, light UAVs are very limited in available energy on board; thus, any additional weight is unwanted. If future U-space regulation allows the usage of UAVs without lightning protection, then information on areas with lightning strike threats would be critical for flight safety.

The overall impact of weather varies between different types of aircraft and from one UAV model to another. We expect that the issue will be addressed through a weather susceptibility classification for UAV models.

2.1.3. Turbulence

Turbulence is a well-known threat to aviation, and each year, it brings injuries, structural damage, and even deaths [39]. Significant wake vortices appear behind large airplanes, and the effect migrates in space (normally with slow descent), remaining for several minutes [24]. It can seriously affect the controllability of the aircraft, which enters a wake vortex area. To avoid the negative effects of wake vortices, it is essential to have information on the UAV wake vortex category and area of operation for all sufficiently heavy airplanes that do flights next to the operational area [13]. The heavy aircraft appear at VLL near runways. This circumstance would allow limiting the UAV operations in this area for a certain period.

Natural turbulence can appear even at VLL with high ground, mountains, high buildings, thermal effects, and convective activity [39]. Buildings and terrain generate turbulence with wind field flow. It is a subject of UTM interest since the effect’s influence on the UAV can be significant [40]. However, there is no need to install sensors on each corner of the buildings because, in conjunction with real-time observations, the computational fluid dynamics approaches [40] can help with turbulence map generation. It is essential to get access to urban buildings and surface 3D maps for the estimation of wind vectors via computational fluid dynamics techniques.

In the second phase of the U-space, the Weather Information service will be deployed to inform stakeholders about the weather conditions. The service should provide weather forecasts and corresponding warnings with hyperlocal weather information, when and where available or required [17].

Monitoring is another U-space service that will share various alerts, including the ones of high importance that come from the Weather Information service [17].

2.1.4. Identified Gaps

Nevertheless, neither service tackles the problems of turbulence, the UAV wake vortex category, or lightning strike hazards. The analysis given in Table 1 allows us to conclude that important information on angles during flight is missing. The available onboard energy is normally available to the RPIC; this could be a reason why a corresponding U-space service has not been proposed yet. In the case of centralized or hybrid systems, such data must be shared with traffic guidance autonomous systems to manage UAV trajectories safely.

In this article, we use the term “needed information” which means the essential information for tackling the specific constraints in the UAV autonomous guidance deployment.

2.2. Trajectory Computation

The robotic Sense-Think-Act cycle is a classical approach to aligning the actual state of the world to robotic perception. The concept was borrowed from research on human cognition [12]. Various interconnected on-board and on-ground sensors collect data on the environment, then the software interprets it and builds a map for actions and/or navigation. Among these sensors are lidars, radars, infrared and video cameras, Global Positioning System (GPS) receivers, etc. Meanwhile, map representation is not an obvious task since various constraints, advantages, and disadvantages must be taken into consideration.

2.2.1. Time Complexity

Using hundreds of uncrewed aircraft, especially in congested airspace, puts high demands on the processing of 4D trajectories, because the safety-critical guidance decisions must be made quickly due to a separation that measures in meters [13]. This fact seriously limits borrowing practices from highly automated commercial aviation and ATM where large separation (usually 3 or 5 miles) between the aircraft gives greater freedom in time and space for making a safe decision [13].

A two-dimensional (2D) map uses XY coordinates, and such an approach aids faster computation and is well suited for guidance on the surface but is disadvantageous for UAVs because they operate in a 3D space. Different altitudes of obstacles and navigation points bring challenges for 2D mapping.

To keep the advantages of 2D map representation and not neglect the variability of altitudes, it is possible to employ a 2.5-D approach [41]. The idea is to use two-dimensional maps as layers, where each layer means a fixed altitude. Supporters of the 2.5-D approach argue that conventional air transport operates mainly on fixed altitudes and that it is reasonable to expect that this approach can be inherited (at least for the top layers of VLL) by UTM as a successor of ATM [42].

While 2-D and 2.5-D approaches exhibit better time complexity than 3D, as airspace involves three dimensions, 3D maps enable more flexible modeling with fewer assumptions. Nevertheless, the 3D approach comes with increased computation costs. To solve the pathfinding issue, there are plenty of approaches representing the UAV’s map and trajectory planning [25]. A comprehensive taxonomy for the most known techniques was proposed by S. Aggarwal et al. [43] and L. Yang et al. [44], among them Sampling-based algorithms, Node-based algorithms, Mathematic-model-based algorithms, Bioinspired algorithms, and Multifusion algorithms.

Flight time planning is vital for avoiding collisions and aiding operational efficiency. However, not all UAV trajectory planning algorithms consider time which leads to another significant issue. For example, a group of UAVs cannot move safely on the same route without being separated timely. If the time factor is ignored, then each next trajectory must be planned without intersection with all existing ones to avoid a potential collision. Neglecting the time in trajectory planning leads to inefficiency of airspace usage.

Various algorithms exhibit different advantages and limitations in the context of different missions and approaches for 4D trajectory planning. We propose the development of a set of algorithms to address the issue. Selecting among these algorithms would enhance time complexity and trajectory optimality.

It is hard to propose more quantitative demands to time complexity as it depends on the concrete mission type, system type (centralized, decentralized, hybrid), U-space traffic density, regulatory demands to separation, etc. We acknowledge the importance of further research in this direction.

2.2.2. Four-Dimensional Trajectory Optimality

Since “safety first” is a core principle of U-space [16], the shortest trajectory has a lower priority against flight safety. In the case of guidance within the U-space, the optimal trajectory for some missions means the shortest flight path after taking into consideration flight safety, security, and related obligatory regulatory demands in the U-space.

One of the most recent proposals regarding UTM notes that with services’ maturity, the Flight Planning and Authorisation service will include 4D trajectories relying on aircraft performance [9]. For example, A. Gardi et al. [9] proposed a 4D Trajectory planner/optimizer for a UAV Mission Management System using cost functions with the following parameters: maximum endurance, minimum flight time, cost, emissions, and noise. However, the minimum flight time and the maximum endurance are needed for different missions. For example, the quickest delivery puts a higher demand on flight time minimization. But surveillance often needs greater endurance—i.e., flight time maximization for flying over selected waypoints as long as possible. The various UAV users’ preferences for different missions bring additional complexity. In some cases, minimum flight time has a higher priority, or maximum endurance, or maximum distance of flight, or minimization of emissions and noise. Also, a combination of these parameters can reflect the specific demands for missions’ diversity. Potentially, a holistic U-space traffic optimality can influence the UAV’s trajectory planning. With that, we find the “optimal trajectory” planning to be a subject that must be addressed according to each specific type of mission and drone user preferences while taking into consideration a comprehensive set of the existing constraints.

2.2.3. Scalability, Adaptability, Learning Capability, Robustness

For centralized or hybrid systems, the trajectory computation scalability can have a significant impact in the event of episodic or permanent traffic (or obstacles) volume growth [45]. It is essential to address this constraint to mitigate the risk of system overload. In the context of a decentralized system, a question arises: How should multiple decentralized subsystems collaborate in the same area to provide safe 4D trajectory planning from the perspective of scalability, adaptability, learning capability, and robustness? To answer this question, additional research is suggested.

Adaptability constrains an autonomous guidance system in the event of significant changes in the environment [46], U-space traffic, regulations, and other factors. For instance, rare weather conditions can sometimes significantly alter the visual landscape, posing a challenge for autonomous surveillance systems that rely on video cameras and typical patterns for that region’s visual landscape database. Another example can be changes in regulation. Hence, the trajectory computation should be flexible for adaptations.

Robustness is essential for UAV trajectory generation [47] since it allows for accomplishing the mission despite the environmental changes. For instance, wind can change its direction and speed. To address these issues, potential changes should be properly considered at the algorithmic level. In this light, the learning capability may play a significant role in environmental change prediction, operation planning, situational awareness [48], traffic optimization, etc.

In summary, we conclude that a proper selection among software solutions is essential to achieve the optimum trajectory computation, as the expected system complexity will likely be based on a set of interconnected algorithms and ML techniques that are integrated through the software solutions.

2.2.4. Identified Gaps

Table 2 summarizes trajectory computation constraints. None of them are addressed at U-space yet [17]. This fact states an existing information provision gap concerning U-space traffic autonomous guidance. Selection among algorithms and software solutions is essential for autonomous guidance; however, an additional investigation is recommended to decide whether to include a specific service in U-space architecture or not. Alternatively, the issue can be addressed with regulation or standardization requirements.

2.3. Collision Avoidance

The International Civil Aviation Organization proposes three layers of conflict management: “strategic conflict management through airspace organization and management, demand and capacity balancing, and traffic synchronization; separation provision; and collision avoidance” [49]. Strategic conflict management and demand and capacity balancing alongside traffic synchronization can be perfectly achieved via a centralized system or hybrid system. However, a decentralized system significantly limits the potential benefits.

For example, achieving holistic traffic optimization with 4D trajectories is not feasible within a decentralized system. On the other hand, separation provision and collision avoidance can be effectively addressed in any type of system.

U-space ConOps defines strategic deconfliction as a process that allows for the reduction in the probability of a conflict to an appropriate level [17]. However, tactical conflict deconfliction and resolution services are becoming essential if the plan is not followed accurately enough.

Collision avoidance is a fundamental principle of flight. Trajectory planning relies on pathfinding techniques, where free space and non-free space are always distinguished [50,51]. There is a plethora of path-finding algorithms that can build a collision-free trajectory while minimizing traveling distance with varying degrees of computational complexity [44] (the rate of deviation in minimization depends on the concrete algorithm). While there are plenty of technologies to tackle the problem, typically, to avoid collision, there is a need to know the UAV and potential obstacles’ positions in a 4D environment [52]. In the case of tactical deconfliction, there is a need for information regarding motion vectors, for example, between the UAV and another aircraft [53]. To make it happen, sensing methods collect data via ground-based and air-based technologies with cooperative and non-cooperative sensing [54].

Static obstacles are buildings, bridges, cranes, trees, power lines, and other ground and on-surface infrastructure. In 3D airspace, the terrain is a basic static obstacle; to prevent controlled flight into the terrain, there is a need for the terrain map to list altitudes above sea level [55]. To address the problem of static obstacles, U-space will be supported with Geographical Information Service in phase U2 and Drone Aeronautical Information Management in phase U1 [17].

The dynamic aerial obstacles at VLL include U-space traffic, ATM traffic, airspace intruders, and wildlife [56]. Since airspace intruders and wildlife do not report their position and intentions, their presence in the area of operations is a potential hazard. To mitigate the risk of collision, it is important to detect and classify the hazard. Classification can help with 4D trajectory planning in a safe way according to the level of threat. The level of threat differs according to different types of aircraft or wildlife; the hazard directly depends on flight characteristics and the behavior in the case of animals. For example, an air balloon will follow the wind flow with a potentially slow descent or climb, meanwhile, some eagles can hunt small-size UAVs [57], reaching hundreds of kilometers per hour when diving. In practice, it means that safe separation from air balloons and hunting eagles must be very different. To allow for suitable separation, there is a need for information on the classification of airspace intruders and wildlife.

There are three services to control and advise U-space traffic in the U2 phase: Traffic Information, Strategic Conflict Resolution, and Monitoring. The U2 Strategic Conflict Prediction service relies on the probability of occupying 4D cells by the UAV, and if the probability is high, then the potential conflict is predicted [17]. In the later U3 phase, other U-space services are planned to manage 4D trajectories and temporary occupied airspace, namely Dynamic Capacity Management, Tactical Conflict Prediction, and Tactical Conflict Resolution [17].

There is a need for coordination with ATM for collision-free flight in the controlled airspace. To address the problem, the Procedural Interface with ATC was proposed in the second phase (U2) [17]. For the U3 stage, a higher level of cooperation has been planned via the Collaborative Interface with ATC. With that, the information on ATM traffic with 4D trajectories at VLL in the controlled airspace will be available for U-space stakeholders.

Operations near the ground can generate a risk of collision with dynamic on-surface obstacles, among them people, cranes, on-surface transport, machines, and animals. For example, if a transportation or parcel delivery mission includes landing outside specially allocated and protected areas (vertiports), then the classification of surrounding objects is needed to make a proper decision on the corresponding risk. For example, if dogs or children are playing nearby, then landing next to them corresponds to a higher risk. It means that the classical approach to robotic navigation space segregation between free and non-free is not enough. The objects’ classification is needed to deal with hazards during landing. In this case, the autonomous guidance system should act like a human [5]—i.e., collect data on the environment, classify surrounding objects and the related level of threat, and then make decisions on the landing options.

The traffic’s lateral, vertical, and longitudinal separation is vital for flight safety [9]. Geometrically, it means that the UAV flight path should be presented as a tube (possibly a cubic one), which is marked as temporary non-free airspace on a 3D map. Also, separation is needed to keep a safe distance from on-surface obstacles [9]. One of the recent European projects, BUBBLES [58], proposed using AI models and techniques in the U-space Separation Management Service to address the problem of separation. The service delivers computation of separation minima in an automatic regime according to the selected target level of safety.

The vertiport availability will limit flight planning since UAVs must be compatible with the vertiport. To handle this issue, there is a need for information regarding UAV size, flight characteristics, refueling/charging options, and noise [17]. Information on vertiport capacity performance is essential for organizing the vertiport traffic efficiently and safely [13,15]. The traffic planning system must therefore be informed in real time: at which vertiport, how many UAVs are landed, and how many unoccupied stands of aircraft are left at the vertiports. The Operation Plan Preparation and Optimization U-space service collects information on the vertiport capacity; however, it is not enough to address the vertiport availability. Though U-space ConOps proposes a Vertiport Dynamic Information Service, its description is at a high level: “Responsible for managing status, resources (open/closed/availability, capacity) information about the vertiport in real time” [21]. Similarly, the Vertiport Resource Allocation Management service function is explained very vaguely: “The ability to allocate resources of vertiports to accommodate UAS requests” [59]. With that, we conclude that a more detailed description of the services is essential. Without a detailed description, we recognize a gap for UAV flight characteristics, UAV size, requirements for charging and fueling, ground handling needs, and UAV noise category.

Identified Gaps

Obstacles’ uncertainty makes 4D trajectory planning more complicated than if the position of the obstacles is already known. In such cases, Simultaneous Localization and Mapping techniques are commonly used to assist autonomous guidance via building an environmental 3D map in real time with the simultaneous interpretation of the robot’s position [60]. A set of on-board and on-surface sensors can collect data regarding the airspace participants, and alongside the airspace traffic information, they make a 3D map in real time where free and non-free spaces are distinguished. However, if the sensors cannot cover the whole area of operation, obstacle uncertainty will be experienced. In other words, the estimated time to the closest point of approach [61] can be too short. In this case, additional efforts are needed to plan the 4D trajectory safely. For example, to avoid a potential collision, it is reasonable to expect a regulatory limitation of the flight speed when the UAV approaches an unknown environment according to the level of a potential threat. Among possible solutions for making the airspace known, we can list lidars, photo/video cameras, infrared cameras, radars, and even radio [62,63,64]. A 3D map of known and unknown environments is needed to tackle the problem; however, it is not addressed among the U-space services yet [17].

With the summary of analyses in Table 3, we may conclude that information about on-surface dynamic obstacles’ position and their classification, along with the position of airspace intruders and wildlife and their classification, is missing.

2.4. Communication Navigation Surveillance

A reliable wireless communication link between the UAV and the traffic management system is essential for effective real-time U-space traffic management. The literature indicates that robust mobile network coverage should extend to over 98% of the operational area [65]. A reliable wireless communication link allows for exchanging various indispensable data including the UAV position, status of onboard systems, sensors’ data, situational awareness, commands for the correction of the 4D trajectory, mission-specific directives such as cargo drop or initiating photo/video recordings, and many more. The Quality of Service (QoS) in communication links refers to the ability to ensure a certain level of performance in terms of reliability, low latency, and high throughputs. The reliability is more associated with uninterrupted connectivity among drones and with control stations with high data rates and minimum possible packet loss. Similarly, the minimum possible time taken in a complete communication cycle refers to latency. The QoS in drone communications can be achieved with several techniques; for example, dynamic spectrum allocation, quality-aware routing, and adaptive modulation and coding schemes, which is not possible without Software Defined Networks (SDNs) [66]. The QoS is a promising approach to ensure the safety and security of drones and avoid unforeseen situations such as collision, trajectory derailing, etc., in U-space where prompt response and continuous connectivity are required. Similarly, availability, accuracy, continuity, and integrity characterize Navigation and Surveillance needs [67].

For the centralized and hybrid systems, an emergent status of the UAV, e.g., Loss-of-Control or Loss-of-Engine/Energy [14], is important information that must be transmitted to the 4D trajectory management system immediately to mitigate the risk of potential collision during uncontrolled flight or unplanned descent. To stay in an area of stable connection, there is a need for a map and the status of the communication network availability at VLL—see Table 4. In the second phase of U-space, the Communication Coverage Information service should tackle the issue with the coverage map, and the Navigation and Communication Infrastructure Monitoring service will provide control of the status information concerning communication infrastructure [17].

For the decentralized system, a stable connection is also important, for example, for real-time updates regarding restricted areas. A geographic node speed drone-based routing is proposed in [68] to minimize the communication overhead at the cohort level in order to ensure reliable communication, whereas, conventionally, an internet-protocol-based (IP-based) approach has been used so far. The trends of the presented graph in [68] show significant improvements in terms of connection loss rate, latency, data rate, and getting updates of location information. Similarly, in a decentralized system where the drone-to-everything concept is utilized in a multiple-layer cooperative architecture in combination with hybrid bioinspired grey wolf optimization, waypoint traceability has been proposed [69], resulting in significant improvements in latency and reliability.

To mitigate the risk of electromagnetic interference, there is a need for information on radio frequency availability [16,70]. To address the problem, the electromagnetic interference information service of U-space has been planned for phase U2 [17]. Other common attacks on drone communication links are denial of service, de-authentication attacks, man-in-the-middle attacks, trojans, etc., discussed in detail in [71]. The possible solution to mitigate these attacks, or at least to tackle them, could be the use of a machine learning approach for the identification of legitimate drones and, in addition, a common database [72], deep learning algorithms for the routing protocols [73], and blockchain technology for the encryption and decryption of secure information [74].

GPS and the Global Navigation Satellite System (GNSS) play a significant role in outdoor guidance and navigation by providing information on the UAV’s position [75]. The ground-based navigation network can also be used to aid navigation and guidance purposes, for example, a ground-based augmented system—a classical approach for commercial aviation [76]. Also, the beacon-based ground systems have the potential to be used for UAV navigation and guidance purposes [77]. To plan the 4D trajectories safely, it is essential to have access to maps of the GNSS coverage and the availability of supplementary navigation networks. Phase U2’s Navigation Coverage Information U-space service is responsible for that [17]. However, it is also important to control the status of the link to the navigation networks. The Navigation Infrastructure Monitoring service in phase U2 will deliver this function [17].

The real-time position of UAVs is crucial information for U-space traffic planning. There is a specific service planned in phase U2 to tackle this issue—Tracking [17]. To address the issue of the geometric height of flight, the Vertical Conversion Service was proposed in the U3 stage 17. It is also planned that the Vertical Alert Service would provide the height warnings.

Presently, the surveillance brings critical information about the situation awareness for the RPIC during BVLOS flights. During the mission, a human analyzes and interprets the video, classifying surrounding objects and subjects, their position, intentions, potential level of threat, and so on. With technological progress, such video data have the potential to be used for the same purposes by the autonomous guidance system. It is expected that the Surveillance Data Exchange service will be deployed in phase U2 to manage surveillance data [17].

2.5. Institutional Constraints

For all aircraft, including UAVs, their operation consists of Phases of Flight: planning, take-off, climb, cruise, descent, approach, and sometimes taxi [78]. In the planning phase, registration and identification are needed to execute the U-space control fundamental function. This happens via verification of the UAV’s owner, the UAV’s serial number and model, the UAV’s size, the persons responsible for continuing airworthiness, the status of airworthiness (the certification status, UAV’s system status, maintenance checklist, etc.), the legal affairs with the UAV’s owner, the legal responsibility for the UAV operation (legal recording), and incident and accident reporting [16]. Also, it is necessary to manage the capacity of the airspace [15]. Initial U-space services deliver registration and identification functions (Registration and Network Identification) [17]. In the second phase of U-space, Legal Recording, Incident/Accident Reporting, and Digital Logbook services will be responsible for incident and accident data and legal recording [17].

Presently, UAS operators must comply with airworthiness and operational directives issued by the European Union Aviation Safety Agency [79]. However, in the future, it is possible to expect an emergence of AI-based services for controlling the airworthiness status of UAVs. In any case, the information on the UAV’s airworthiness status is critical for flight safety, which is needed for an autonomous traffic planner. Without such information, it can be too risky to permit take-off.

According to [80], the Network Identification service should collect information on the emergency status of the UAS. In the second phase, U2, the Emergency Management service will assist RPIC in case of emergency. However, we expect that at some level of automation integration into the U-space, AI will be analyzing the problem by delivering aircraft guidance to tackle the appearing risk efficiently.

To mitigate the risks, it is essential to make a risk evaluation before the flight [81]. Joint Authorities for Rulemaking on Unmanned Systems proposed a Specific Operations Risk Assessment that can be used for the estimation of the level of risk for specific missions [82]. P. Hullah et al. proposed U-space Airspace Risk Assessment that addresses four core sources of risk in U-space operations: safety risks, security risks, privacy risks, and environmental risks [83]. The U-space ConOps mentioned three U2-phase U-space services to address flight risk: Operation Plan Preparation/Optimization, Risk Analysis Assistance, and Flight Authorisation service [17].

To mitigate the risk of collision with people and on-surface objects, it is essential to take into consideration the potential emergency landing options [17,84] in UAV trajectory planning. For example, the autonomous trajectory planner should be informed in advance of what places will be potentially safe for landing (uninhabited roofs, wastelands, etc.) and what must be avoided (highways, railways, crowd-gathering areas, etc.). To decrease the risk of collision with pedestrians during an emergency landing or loss of control, it is essential to have a population density map [16,85]. For example, E. Arcel et al. describe a casualty estimation model that relies on population density data to estimate the nonparticipant casualty risk [86]. The Population Density Information service was proposed for the second phase of U-space deployment for this purpose [17].

Runway and Vertiport Surface Contamination can be critical for taxing, take-off, and landing safety, especially with a combination of water, ice, and snow [87]. Surface contamination is a hazard not only for fixed-wing UAVs, but can also be dangerous for small rotorcraft in cases of heavy contamination of water due to a sink risk. Runway and Vertiport Surface Condition includes contamination and other issues [88]. Though the authors of [59] discuss Surface Condition Awareness as a part of U-space capabilities, it is not clearly stated what U-space service should be responsible for that.

Restricted areas [17] can be deployed to mitigate the safety and security risks. For instance, there can be a permanent restricted area over a nuclear plant, or a temporary one during an airshow. To plan the UAV 4D trajectory, there is a clear need for real-time updates on the position of restricted areas and their duration. For autonomous trajectory planners, those areas mean non-free airspace. Automated reading of Notice to Airmen or Notice to Air Missions (NOTAM) messages could interpret and remove irrelevant information [89]. To make it readable for the autonomous guidance system, there is a need for an interpretation of these data into free/non-free airspace coordinates within a time frame. In phase U2, it has been planned that the Geo-awareness service will start providing geofencing information with 4D coordinates [17].

Fair and equal access to airspace is one of the U-space principles [1]. However, ordinarily, the police and emergency services have a privilege of priority; thus, UAV 4D trajectory planning should be able to work within the metrics of priorities [90]. Hence, it is possible to say that for planning 4D trajectories, a regulation priority policy is needed. The European Union regulation 2021/664 article 10, paragraph 8 describes two levels of priorities for flights: normal and prioritized [80]. Meanwhile, Airbus UTM researchers argue that complex implementations of fairness (related to airspace access) need further study [90]. We expect a significant development of priority regulation as U-space is further developed. One of the potential improvements can be related to holistic traffic optimization versus a single flight.

The spatial limits of the U-space airspace are essential for conducting operations in the allowed area. The U3 phase, Common Information Service, will be responsible for that [1].

UTM secure operations reflect U-space fundamental principles [1]. It is essential to secure the communication channels, provide reliable identity, organize proper access management, and respond timely to incidents and breaches. To mitigate the risk of security breaches, it is vital to be informed about the security system status and level of threat [91,92].

Noise is another major public concern [93] regarding the mass integration of UAVs into airspace, so it is reasonable to expect the development of regulations related to this issue. Restricted areas may be marked to protect some areas from acoustic disturbance. Noise reduction is a complicated multifactor task, and it is reasonable to predict that flights over noise-sensitive areas [94] will be forbidden for some types of UAVs. Exceptions can be established for flights in the gliding regime or noiseless air balloons. To protect the population from unacceptable noise, there is a need for a map with areas of limited operation time windows and unacceptable levels of noise. We expect a performance-based approach for deploying such limitations.

Rule awareness regarding priority, privacy, noise reduction, etc., is essential for executing regulation demands. Presently, the regulation is written by humans for humans. We find it possible that in the later stages of U-space development, the regulation demands will be written by humans (regulators) for machines (the autonomous traffic management systems) in terms of algorithms (computer code) that will be readable for the U-space AI. Such an approach has the potential to address and implement regulation updates immediately to the autonomous traffic management systems on a large or even global scale.

Identified Gaps

The analysis provided in Table 5 identifies the next essential information, which is not addressed within U-space services yet: status of airworthiness, runway and vertiport surface conditions, location of suitable landing areas in case of emergency, UTM security breakthrough status and level of threat, map of noise-sensitive areas, and rule awareness in terms of robotic algorithms. To solve the problem of information provision for the U-space traffic autonomous guidance, we expect that new U-space services will be appearing respectively.

2.6. Mission Type

Missions’ diversity brings additional demands to UAV trajectory planning—see Table 6. For example, the delivery mission can be carried out in various ways. Some approaches propose dropping cargo on a parachute [95]. For heavy cargo delivery via fixed-wing UAVs, it is also possible to expect a strict demand for landing on a runway only. Another approach has been tested by the American company Wing for parcel delivery by hovering where the parcel is roped down with a winch [96].

Human transportation entails a risk for human lives [97]; thus, UAV 4D trajectory planning with passengers on board can have additional safety demands; for example, a demand for a greater value of separation. Maneuvering with passengers onboard must be done in a slow and non-aggressive way. To aid passengers’ comfort and safety, it is reasonable to expect stricter (compared to the other types of missions) limitations on the descent speed, load factor, aircraft angles, or planning trajectory in favor of a better view.

Missions for data collection and dissemination include surveillance [98], photography [99], inspection [100], mapping [101], and communication network deployment [102,103]. Such missions can have a wide range of degrees of trajectory planning complexity. For instance, surveillance or communication network deployment missions with dozens of meters of altitude can put less demand on the UAV trajectory planner than an inspection of a bridge involving taking high-resolution photos of the specifically selected areas while operating in a congested urban environment.

SAR missions such as assisting humans in danger via UAV [104] can be very complicated due to a high diversity of circumstances. Sometimes SAR must relate to very specific conditions of operation; for instance, fires or natural disasters. Fire itself is an imminent danger for a drone because the combustion is accompanied by a high temperature, vortices [105], and the degradation of visibility. This means that the autonomous guidance system would need a model to estimate the area and level of hazard to operate safely.

Watering, sowing, and spraying as well as inspecting are potential operations in agriculture [106,107]. For such missions, automation of the 4D trajectory planner has its own specialty. Since UAVs can fly at the lowest altitude, they can share airspace with manned agricultural aircraft which can potentially operate without ADS-B [108]. Finally, such operations can also happen in fields with lower human presence.

Entertainment missions provide filming, UAV shows, leisure flights, etc. [109]. Following and recording a cyclist in the mountains or filming a wedding day put specific demands on the UAV trajectory planner.

Surveillance and photography can be performed by recording an allocated area if needed. However, additional demands for such missions are also possible. For instance, it can be following a selected target during a surveillance mission or taking photos of specific objects (or subjects) [110].

Potentially, an even greater variety of missions lies ahead. It could be firefighting [111], construction, etc. [112]. With technological progress, more and more UAS missions will be delegated to robotic and autonomous guidance; it is just a matter of time [17]. Nevertheless, there is a commonality for all missions, namely information concerning take-off and landing locations. The AI of autonomous guidance systems needs to be informed of where the mission starts and where the UAS should finish it.

U-space ConOps notes that the drone operator will be responsible for the mission planning [21]. We expect an autonomous mission planner deployment in the later stages, initially for simpler missions and subsequently for more complex ones. The level of centralization in autonomous mission planning is an open issue, and we acknowledge a research gap here.

We distinguish between the terms “dynamic mission” and “pre-defined mission” to describe a UAS mission where the UAV does more than fly from one location to another with known waypoints before take-off [113]. For example, a routing mission, such as delivering medical samples from a remote hospital to a laboratory, is classified as a predefined mission. Conversely, surveillance or inspection of a new object may require decision-making during the flight; thus, we classify it as a dynamic mission.

The autonomous guidance system needs to be informed where the UAV’s service is needed. It means that the coordinates of the object/subject/area of interest are a common information demand for any dynamic mission execution.

Classification of the object/subject of interest can be necessary to execute some dynamic missions. For example, it can be surveillance operations involving following a car to aid the police, searching for an injured person in a SAR mission, plant classification for watering and spraying, etc.

The mission’s purpose and specialty information are needed to explain to the autonomous guidance system what the UAV should or should not do during the dynamic mission. For that, the UAS’s user inputs the specific demands to the autonomous guidance system, which explains to AI the type of service and where it is needed. For example, for filming, the user can expect a flight over the potential area/object/subject of interest with taking photos and videos during a selected period. The mission’s purpose, complexity, and special character should be compared with the capabilities of available specific algorithms. If the mission entails complex or unique decisions and there is no suitable algorithm, then human intervention is needed, and this statement is fair for all types of missions. The issue is one of a trade-off between the mission complexity and the algorithm’s ability to complete the operation effectively and safely.

The U-space deploys Operation Plan Preparation and Optimization service in the second phase U2 [21]. The service will collect information on take-off and landing locations and the area of operation. This approach can be sufficient for strategic deconfliction, but as was argued, it is not enough for dynamic missions.

Identified Gaps

With that, we may conclude that there are gaps related to information needs, namely the classification of object/subject/area of interest, mission’s purpose and specialty inputs, and matching with available algorithms. For pre-defined missions, such information may not be relevant. However, for some dynamic missions, it can be essential.

3. Identified Gap Prioritization

3.1. Gaps with a High Priority

Not all the gaps constrain U-space traffic autonomous guidance equally. For example, the gaps related to flight physics are critical for flight safety [23]. Specifically, the monitoring and control of aircraft angles, available energy on board, natural turbulence map, area and time of operation for heavy aircraft generating wake vortices at VLL, UAV wake vortex category, and lightning strike threat areas are essential to guarantee flight safety.

Adaptability and robustness of the 4D trajectory computation can be safety-critical in case the airspace situation requires timely and proper trajectory re-computation. The autonomous guidance system must be able to respond correctly to each unique combination of the new circumstances that UAVs could encounter in the U-space airspace.

Time complexity can pose a threat to flight safety in scenarios like tactical deconfliction when the speed of calculations can have a significant impact on a new safe trajectory recomputation [43] or collision-avoidance maneuver. However, it should be less critical at the strategical deconfliction stage. Four-dimensional trajectory optimality can be essential for cost efficiency; however, safe flight remains possible even if the aircraft’s trajectory is not optimal.

On-surface dynamic obstacles’ position and classification, alongside information regarding UAV flight characteristics, and UAV size are safety-critical for avoiding collisions during taxi, take-off, landing, and flight at altitudes near on-surface objects. Additionally, information on UAV flight characteristics can be essential to avoid runway excursion.

The position and classification of airspace intruders and wildlife are safety-critical in any case, as a midair collision could result in aircraft damage, injuries, loss of life, or pose a threat to public safety and property [114].

Three-dimensional maps of known and unknown environments can have a significant influence on flight safety if UAVs enter unknown areas at a high speed. However, more research is needed to precisely define the level of threat. For example, a new wind turbine construction raises a new obstacle that must be considered by the autonomous uncrewed traffic planner to mitigate the risk of a potential collision.

The status of airworthiness, location of suitable landing areas, runway surface conditions, and UTM security breakthrough status can be critical for flight safety.

Rule awareness in terms of robotic algorithms is essential in order to follow official demands, where some rules can be critical for flight safety and security. For instance, changes in regulation on separation requirements must be reflected in the 4D trajectory planning algorithm. In general terms, U-space autonomous systems must always be updated according to the latest regulations.

Classification/identification and coordinates of the object/subject/area of interest alongside the mission’s purpose and specialty inputs are essential only in the case of dynamic missions. It is highly possible that autonomous guidance at U-space will be started with the most simple missions like flying from one point to another. Nevertheless, for dynamic missions, the gap has a high priority.

3.2. Gaps with a Moderate Priority

In this subsection, we discuss the gaps for which the deployment priority can be classified as moderate.

Learning capability and 4D trajectory computation should be considered as essential fundamental principles for the development of the U-space autonomous guidance system. It must mitigate the risk of the system’s inability to improve itself, based on the registered failures. Such demands may require significant technological advancement from the current state of the art of artificial intelligence [115]. We expect that the system learning function will be deployed under human supervision in the early stages and potentially self-learning characteristics can be incorporated in the later stages. We classify this as a long-term task with a moderate initial priority.

The scalability issue is essential to address for centralized or hybrid system deployment. However, if the system is at full capacity, it means no more orders for a certain period of time. It may have a significant negative economic impact, but it is not safety-critical. In this light, we consider the priority to be moderate.

Requirements for charging and fueling, ground handling needs, and UAV noise category are essential. We suggest collecting this information at the time of UAV certification or registration before flight. The lack of this information can result in significant inefficiency in vertiport usage. For example, in this case, a situation may arise when a hybrid-engine UAV is landing, but a vertiport does not have suitable fuel to fill the drones’ tanks as only electrical charging is available. However, it is possible to state that this information is not safety-critical.

Maps of noise-sensitive areas relate to public comfort and have a negligible impact on flight safety.

4. A methodology to Identify, Measure, and Close the Gaps

This paper sheds light on the information needs and how they match with the U-space services. However, the word “match” does not mean that the services are ready to deliver their functions fully. As was shown in the previous sections, some of the U-space services do not exist yet. A real match will be achieved and tested with a step-by-step system development with significant contributions from daily practice and further research.

In a scientific approach, the fundamental component is the measurement of the topic of study. In this light, we propose a high-level methodology to identify, measure, and close the gaps (see Figure 2). In the initial stage, conducting a literature review is essential for investigating the flow of safety-critical information required for autonomous guidance. This is precisely what our article delivers.

In the second stage, we suggest developing a scale to measure the information needs for each service specifically. Information required on angles during flight may significantly vary from one approach to 4D trajectory planning to another. Some approaches could be based on simplified flight models, while others on more advanced, leaving the responsibility for angle control to the onboard systems. The less precise flight models, the greater separation reserve will be needed.

For example, information regarding the angle of attack for a fixed-wing UAV is crucial for flight safety within 4D trajectory planning and replanning during flight. If this information is not of sufficient quality and comes to the remote autonomous guidance system with delays, it could lead to stalling and even spin of the aircraft posing a significant threat to flight safety. In case of approaching a high angle of attack, the autonomous aircraft should normally decrease its flight path angle to get acceleration from gravity and normalize the angle of attack. However, if a remote trajectory planner does not get angle information in time and of proper quality, it could lead to a situation where the autonomous trajectory planner provides a flight path that leads to an angle of attack greater than critical. It is a multi-factor issue that requires additional research on the quality and acceptable deviations of the information required.

Information on natural turbulence map, area, and time of operation for heavy aircraft generating wake vortices at VLL, UAV wake vortex category, and lightning strike threat areas can be estimated via modeling the areas and the identification of the acceptable level of the weather phenomena hazard and probability of the appearance.

To investigate a suitable level of information provision for trajectory computation, we suggest a mission-centric approach under a set of existing conditions. Each type of mission should be modeled and tested in simulations and experimental flights, and based on that, the acceptable level of the quality of the information can be found.

Position and classification of on-surface dynamic obstacles, airspace intruders, and wildlife are directly related to the separation required to guarantee the European level of flight safety. We recommend an additional study to investigate the existing aircraft performance range for the existing drone models and the wildlife behavior that is typical for European aerospace. Based on that data, the quality of information required can be identified.

A study on 3D maps of known and unknown environmental information should investigate what level of precision and online update is essential to guarantee a collision-free flight. Is it important to have a detailed 3D map of the area of operation; alternatively, simplified “generic” models can fulfill this role.

We suppose that UAV flight characteristics, UAV size, requirements for charging and fueling, ground handling needs, and UAV noise category must be collected during UAV certification processes, and the quality of this information should be similar to the general aviation requirements. However, it is also reasonable to expect a more generic approach and simplified models for the UAVs without people onboard. The status of airworthiness identification and corresponding procedures will likely vary for the different classes of UAVs. The expected range varies from a quick simple check for the smaller UAVs on the ground to a comprehensive pre-flight check for human transportation that is similar to the general aviation procedures.

The provided analysis identified that risk assessment should be extended with information on the location of suitable landing areas, in case of emergency, and runway and vertiport surface conditions. To identify the quality of the information, we suggest simulating flights with a need for an emergency landing. Analysis of such data could give evidence-based recommendations for measuring the quality of the information.

In a similar manner, UTM security breakthrough (status and level of threat) and noise reduction can be analyzed via modeling and simulations.

We suggest classifying rule awareness in terms of robotic algorithms on safety- and security-critical aspects and others. Safety- and security-critical updates should have higher priority for incorporation into the autonomous system.

Classification/identification and coordinates of the object, subject, or area of interest, along with information on the mission’s purpose and specialized inputs that match available algorithms, require extensive simulations and flight experiments. These are necessary to collect reliable evidence that the autonomous system can safely guide the UAV under certain conditions. As we expect that objects’ classification and identification will rely on machine learning techniques, comprehensive statistics are essential to estimate what level of deviation for information provision is acceptable.

In the third stage, it is essential to identify dependencies between flight safety, U-space efficiency with data accuracy, and time of delay for data transfer. It can be done via U-space simulation to get rough results without significant investments.

In the fourth stage, the optimal balance between dependencies should be investigated. This knowledge would allow us to recommend how to optimally close the gaps.

In the fifth stage, we suggest result correction by obtaining more data via flight experiments in U-space and test zones.

Finally, it is essential to continue improvement by collecting data through U-space usage.

5. Promising Approaches and Solutions

In the previous sections, we analyzed the provision of missing information that must be addressed to allow U-space autonomous traffic guidance. In the current section, we propose our view on the approaches and technologies that have the potential to fill the existing gaps.

UAV performance, maneuverability, and UAV wake vortex category data can be collected as part of an obligatory UAS certification process. We suggest classical approaches such as wind turbine tests, ordinary flight tests, or numerical computation methods. By collecting more data on UAV characteristics, it will be possible to use machine learning techniques for the quick prediction of the tested parameters.

Available onboard energy data can be collected by the U-space via cellular networks (4G/5G) [116], Wi-Fi, very high radio frequency, ultra-high frequency bands, or even microwave frequencies. Optical or laser communication has the potential to transfer data via laser beams with an advanced level of security, as it is hard to intercept the signal. Finally, a recent light fidelity (Li-Fi) technology [117] can be added to the list, as it promises high-speed communication.

The natural turbulence map can be built and updated with computational fluid dynamics, a set of sensors, and data on the weather [40]. For example, weather satellites are essential for the continuous observation of cloud movements, storms, etc. Such data provide a significant basis for working with natural turbulence phenomena. Lidars can collect data on the wind speed, its changes, and atmospheric structure. Radars are useful for detailed data collection on the atmosphere, specifically essential near weather fronts or storms. The weather stations can be placed on the ground, aircraft, weather balloons, sew buoy networks, sea vehicles, and ground vehicles.

Data on the area and time of operation for heavy aircraft generating wake vortices at VLL can be obtained directly from ATM services because heavy aircraft normally fly at VLL in controlled airspace.

Weather forecast analysis can help with the prediction of lightning strike threat areas. There is a set of technologies that can collect the data required. Weather Radar Systems tracks data on thunderstorms—their development, potential for lightning, and intensity. Satellites can monitor temperature, moisture, and cloud formations which is essential for predictions. Ground-based lightning detection networks, lightning detection and ranging systems [118], and atmospheric electric field meters can provide real-time electromagnetic pulses. Numerical weather prediction models can be beneficial as well.

Time complexity estimation can be achieved with mathematical modeling, graph theory [119] usage, and other trajectory planning approaches [43]. However, finding the appropriate level of time complexity is a non-trivial problem. For example, the following factors have a direct impact on it: regulatory safety demands, minimum separation requirements, type of mission, area of operation congestion, airspace availability, quality of communication, and aircraft performance.

Drone user preferences in 4D trajectory planning can be collected via the online software interface Drone User—U-space.

On-surface dynamic obstacles, airspace intruders, and wildlife data can be collected with various on-board and on-surface sensors and then classified with ML techniques [8]. Among the potential solutions are GPS and GNSS systems, lidar, radar, infrared and thermal cameras, and optical and video cameras. Finally, sonar can be used for special cases like the detection of an object in a forest.

Information about known and unknown environments can be collected and updated with on-board and on-surface sensors. However, a specific study on how to fuse multiple-source information in a constantly updated map will be needed.

The status of airworthiness and runway surface conditions should be the area of responsibility of the vertiports. The vertiports must inform the unmanned traffic management system of the corresponding issues via a software interface.

The location of suitable landing areas in case of emergency can be collected with a specific study and updated regularly. Potentially, ML [8] can aid this task by recording and analyzing the ground surface.

UTM security breakthrough status and level of threat can be analyzed with software solutions, where AI can play a significant role in identifying atypical activities that correspond with security breakthroughs.

A map of noise-sensitive areas and regulation demands can be shared by the regulators via a software interface.

In this section, we described the potential approaches that can aid in collecting essential information and filling the gaps. However, this is a high-level vision. There is a need for further detailed research and experiments to make a reliable U-space autonomous traffic guidance system.

This paper did not study how to determine what aspects must be incorporated into the U-space design regarding the U4 phase. To answer this question, we suggest a comprehensive study due to the essentiality of the issue. We admit that it is a multidimensional problem, where experience from manned aviation cannot be completely relied upon due to the different conditions associated with operations at VLL. Flight safety demands, technical limitations, business efficiency, stakeholders’ expectations, and the principles of equality and open competition should be taken into consideration, and the optimal combination of U-space services and delegated functions could then be proposed.

6. Conclusions

This paper presented an identification and classification of the existing constraints, and essential information needs to allow for U-space traffic autonomous guidance. Further, it analyzed how information needs match U-space services, and gaps in information provision were identified. Also, we proposed a high-level methodology to identify, measure, and close the gaps. Finally, suggestions on promising approaches and solutions were proposed.

Based on the article’s findings, the following conclusions can be summarized:

(1)

The present concept of U-space does not satisfy essential information needs for the U-space traffic autonomous guidance.

(2)

The identified gaps in information provision must be closed to allow U-space traffic autonomous guidance.

(3)

Filling existing gaps has different urgencies.

The identification of the information needs does not mean that the U-space traffic autonomous guidance is doable for all types of missions in a safe and efficient way with the modern progress of technologies. It is essential to continue the research to find what technologies can address the identified information needs. Therefore, a large-scale experiment, including dozens/hundreds of UASs, is recommended for testing promising technologies and systems via simulations, mathematical models, and flight experiments. We expect a significant impact of AI implementation for a varied set of tasks—from data capture and classification, including interpretation, pathfinding techniques, multi-factor optimization, and decision-making in heterogeneous scenarios.

We encourage the scientific community to continue researching the measurement and addressing of the identified gaps. The measurement will play an essential role in the U4 deployment by creating quantitative demands on the information and its quality. We suggest our high-level methodology as strategic guidance for future research in this area. The identified constraints and gaps significantly impact flight safety and/or the efficiency of U-space as a set of services. In this light, the paper’s findings provide a solid foundation for information provision in the final stage of U-space design and deployment.

We also recognize that the literature review consolidates existing experiences with systems, theories, and solutions. However, we expect that the practical implementation of U-space U4 could pose new and non-trivial challenges. It is also reasonable to expect that potential AI advancements may radically impact many aspects of uncrewed traffic, including its operations and autonomous guidance.

To conclude, we acknowledge that autonomous guidance for uncrewed traffic at VLL is an immature concept that requires extensive research to measure and address the essential information provision gaps discussed in the paper.