**1. Introduction**

The use of autonomous vehicles, both road going and aerial, is expanding rapidly these days. Just in the case of unmanned aerial vehicles, between 2016 and 2021 Goldman Sachs [1] predicted a USD 100 billion market opportunity, with a further 16–24% annual growth predicted between 2026–2028 [2,3]. The market opportunities for autonomous road vehicles are no less significant. To serve a market of this size, a significant amount of traffic needs to be managed. In controlled airspaces, by 2050 EUROCONTROL predicts that the total flight hours of UAS (unmanned aerial systems) in Europe will account for 20% of total traffic, an estimated 7 million flight hours [4]. Higher demand will be seen at the very low level (VLL) airspace, below 120 m (400 ft) AGL (above ground level), often referred to as the U-Space in the EU. In U-Space, about 250 million commercial flight hours are predicted in urban environments, 20 million in rural settings and 80 million for hobby use. This demand is a magnitude higher than what the current manned air traffic control systems handle.

In the case of road vehicles, as there is no central control, the system relies on individual drivers to manage conflict situations. In the EU, the ACEA (European Automobile Manufacturers Association, Brussels, Belgium) estimates 268 million road vehicles in 2018, with an average growth rate of 2% compared to the previous year [5]. It is further predicted by the Victoria Transport Policy Institute (Victoria, BC, Canada) that by 2045 half of new vehicles will be autonomous, and by 2060 half of the active vehicle fleet will be autonomous [6]. Based on the defined autonomous driving levels [7] autonomous road

**Citation:** Sziroczák, D.; Rohács, D. Automated Conflict Management Framework Development for Autonomous Aerial and Ground Vehicles. *Energies* **2021**, *14*, 8344. https://doi.org/10.3390/en14248344

Academic Editor: Arno Eichberger

Received: 9 November 2021 Accepted: 8 December 2021 Published: 10 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

vehicles are designed to achieve safe and efficient driving individually, and thus control and conflict managemen<sup>t</sup> of these vehicles could also benefit road safety. Based on the high volume of traffic to be managed, it is predicted that a system similar to today's air traffic control will not be sufficient for these vehicles either. Current ATC (air traffic control) procedures are human labour intensive and offer poor automation possibilities.

Aircraft traffic managemen<sup>t</sup> for unmanned systems, or UTM in short, is an active research field today; both national and international organizations actively develop and test various concepts for the managemen<sup>t</sup> of this novel class of airspace users. ICAO has already published the third edition of their UTM harmonization guidelines [8]. The document is aimed at providing a common framework along which nations can develop harmonised solutions for UTM activities. In addition to working towards international compatibility, the initiative also helps to reduce the costs associated with the development of UTM solutions. Essentially, all nations have to find solutions for this novel challenge, and it would be beneficial if the development activities were shared in some form as well. In the USA both NASA (National Aeronautics and Space Administration, Washington, DC, USA) and the FAA (Federal Aviation Administration, Washington, DC, USA) are actively developing the nation's Next Generation Air Transportation System, planned to be in operation by 2025. While the system is aimed at improving various aspects of airspace use, integrating UAS is a significant part of the challenge. Both organizations have been actively engaged in the UAS Integration Pilot Program (IPP) (ended in 2020, now continued in the BEYOND programme) and the UTM Pilot Program (UPP) [9] since 2019, which are aimed at integrating the current estimated 350,000 UAS and future systems into the national airspace. ASTM (American Society for Testing and Materials, West Conshohocken, PA, USA) has developed standards for the remote identification of UAS, published as ASTM F3411 [10]. In the EU, the development of the U-Space system (part of the SESAR programme) is aimed at implementing UTM solutions [11]. At the national level, but also as part of the U-Space initiative, UTM systems have already been introduced, for example, in Finland (GOF, Gulf of Finland [12]), Switzerland (SUSI, Swiss U-Space Implementation, Zug, Switzerland, [13]) and the UK (CAA Innovation Hub, London, UK, [14]). There is also activity at the private level; companies such as Altitude Angel, Skyguide, AirMap and sees.ai have already developed UTM solutions, and some of them are already providing these solutions today. There are also various consortia, such as DOMUS, USIS, DIODE, EuroDRONE, SAFIR, VUTURA, GeoSAFE, PODIUM, SAFEDRONE and CORUS developing various aspects of UTM and related activities. In Hungary, where the research presented was performed, HungaroControl has founded the UTM Innovation HUB and CybAIR Cluster, which are both contributing towards UTM systems development.

In the autonomous ground vehicle (often referred to as self-driving car) industry primary emphasis is placed on collision avoidance systems in individual vehicles. As was previously mentioned, autonomous operation is being developed as a stepped approach, and many of these concepts also appear as driver assistance systems in conventional ground vehicles. These systems include collision warning systems (front and rear) adaptive cruise control, lane assist, road sign detection and similar systems. Approaches such as smart city or more generally smart mobility are aimed at increasing the safety, sustainability and efficiency of current traffic systems, but they do not necessarily rely on the control of vehicles. Solutions include the analysis and evaluation of traffic, variable speed limits, dynamic traffic signal timing or dynamic lanes. In the case of cooperating and connected vehicles, such systems could provide smart routing and speed control for optimum traffic efficiency, essentially reducing time spent in congestions and improving travel experience for drivers and passengers.

This paper is aimed at developing a system prototype that can be used to implement the short- and mid-term conflict managemen<sup>t</sup> of a local U-Space system containing both autonomous aerial and ground vehicles, along with other potential users. The inclusion of both aerial and ground autonomous systems in a common managemen<sup>t</sup> framework can be considered as a novel idea. The pairing of self-driving cars and UAS has been

experimented with, but only for specific applications such as food delivery [15], search and rescue [16] and infrastructure inspection [17]. There are also research works available that focus on the technological specifics of such integration [18,19]. Published studies of conflict managemen<sup>t</sup> approaches that combine autonomous ground and aerial vehicles are unknown to the author.

As the whole concept of unmanned traffic management, both for UGV and UAV, is new, there are no set standards or solutions available today, only development initiatives. The only regulation available in this field is that UGV must abide by the applicable highway code of a given country in order to be compatible with current (manned) traffic. In the case of UAV, there are regulations concerning the individual use of vehicles (such as the EASA or FAA regulations), but there are no standards regarding the urban environment, or any sort of other traffic regulation for that matter, as this type of dense traffic simply does not exist today. As a result, systems being developed today do not need to be prepared to comply with any particular standard; rather, they can form the base of developing standards and regulations for the future.

While the concept of interaction between aerial and ground can be questionable, situations do arise where they can come into conflict. The first category is related to take-off and landing type manoeuvres, in which case UAS voluntarily moves close to the ground to where it can be affected by ground traffic. A typical use of this would be collection and delivery type scenarios. Additionally, this category includes vehicle–aircraft interactions, such as when a drone takes off or lands on the top of a delivery van, which can be either static or moving. The second category includes involuntary near-ground activities, such as emergency landing or descent due to failure, weather effects or while evading other airspace users or obstacles. The third category relates to environmental effects in man-made environments. Concepts such as urban canyons [20], essentially artificial channels that have a significant local impact on wind and gus<sup>t</sup> characteristics (also on GPS and communications), are already known to researchers studying built environments. Additionally, large vehicles, such as buses, HGVs or trains can generate significant air flow in a constrained environment, which is often referred to as the piston effect. These effects can be significant enough that UAS would have to consider the presence, movement and potential path of ground vehicles even when operating at altitudes where direct physical contact is of no concern.

Developing a conflict managemen<sup>t</sup> system involving both unmanned aerial and ground vehicles is essential for the advancement of autonomous vehicle use, especially in urban or peri-urban environments. In the field of aviation, the use of vehicles in an urban environment is referred to as urban air mobility, or UAM for short. Systems being developed for autonomous ground vehicles utilise sensors that primarily detect in the ground plane, as that is where the road traffic is expected. As a result, they are not equipped to detect and deal with aerospace traffic. Based on the roadmaps and published research available today, this is not likely to change in the near future. In the case of aerospace vehicles, as it was mentioned, current ATC procedures are not applicable, mainly due to the performance constraints and poor automation opportunities of current procedures. They key capability such a combined traffic system must provide are the following:


Automation in conflict managemen<sup>t</sup> is the only possibility to provide the required capabilities, mainly due to the high number of users and conflicts envisioned. As such, a common and automated treatment for both classes of vehicles is essential. This paper

presents a prototype solution developed to provide adequate capability for the managemen<sup>t</sup> of conflicts between these users.

The first part of this paper deals with the methodology used to define the requirements and key components of the proposed conflict managemen<sup>t</sup> system. It begins with an introduction to conflict managemen<sup>t</sup> concepts and their applicability to the defined purpose. Then, high-level requirements are defined for the proposed conflict managemen<sup>t</sup> system. In the results section, the system architecture of the implemented conflict managemen<sup>t</sup> system is presented, based on two iterations and live testing of the system prototype. The discussion chapter presents the results and insights gained from the testing of the prototype system and proposes further development directions.

#### **2. Materials and Methods**

#### *2.1. Conflict Management Methodologies*

The proposed conflict managemen<sup>t</sup> system implements a short- and mid-term tactical conflict managemen<sup>t</sup> methodology. Conflict managemen<sup>t</sup> approaches can be generally divided into 2 categories: strategic and tactical management. The basis of the categorization relates to the time factor of conflict detection and the available means to manage the event. The discussion of conflict managemen<sup>t</sup> in this chapter is primarily based on concepts found in the field of aviation.

Strategic managemen<sup>t</sup> refers to measures taken before the actual operation has begun. The activities performed during strategic managemen<sup>t</sup> are aimed at minimizing the potential effects arising from the lack of information or faulty planning of routes. In aviation, typical planning issues include the availability of airspace (including routing through restricted and forbidden airspace), lack of air traffic control capacity (overburdened control officers) and flight plans received from other users. In all these cases, the conflict situation can be predicted in advance of the flight operation. The usual tools for strategic conflict managemen<sup>t</sup> are the rerouting of flight paths or rescheduling of the operations, or when required, even cancelling the flight altogether. As an addendum, considering from a more abstract perspective, the long-term infrastructure planning and traffic policy making can also be considered as strategic level conflict management, just at the timescale of many years. As an example, NTU (Nanyang Technological University) Singapore proposed airspace design methods for UAV traffic in an urban environment, including vertical separation, "traffic light effect", digital unidirectional lanes and other features [21]. In the future, implementing these designs could contribute towards minimizing the situations where tactical level conflict arises. The longer-term strategy also has relevance regarding the infrastructure available to detect users and conflicts during operations (the tactical level). The system proposed in this research does not implement strategic conflict management, and as such it will not be discussed further.

Tactical conflict managemen<sup>t</sup> is responsible for preventing collisions in the following situations:


As these conflict situations arise during operation, most of the time they are not predictable at the planning and path approval stages. As a result, the time available to detect and resolve these conflict situations is significantly shorter than in the case of strategic management. The managemen<sup>t</sup> of conflict situations usually requires active commands, which will result in deviation from the planned path for at least one user involved in the conflict. In the world of manned aviation these commands indicate a change in direction, altitude or speed. The command can come from 3 sources: ground-based systems (voice command from air traffic control officer through radio), on-board systems (TCAS—Traffic Collision Avoidance System—electronic indicators on flight display) or conflict analysis and decision from the pilot itself.

In the case of tactical management, the most critical factor is time. For human pilots (or drivers), processing information and coming to decisions is directly related to the time taken to do so. In the case of urban environments, the typical distances involved in conflict situations are generally small, which leaves little time for human decision making.

The tactical conflict managemen<sup>t</sup> can be divided into 2 stages: conflict detection and conflict resolution. Detection methods can be grouped into 3 categories: deterministic, probabilistic and worst-case methods. While each of the detection methods have advantages and disadvantages compared to each other, their applicability is only valid in the short term, regardless of the method. An overview and classification of published classic conflict detection methods is presented in Table 1.


**Table 1.** Conflict managemen<sup>t</sup> methods overview.

In terms of conflict resolution, the earlier a conflict is detected and acted on, the more efficient and safer the conflict resolution will be. The conflict managemen<sup>t</sup> approach studied in this paper focuses on short- and mid-term tactical managemen<sup>t</sup> and as such, other methods will not be discussed further. For the sake of completeness, Table 2 presents the compiled overview of generalised approaches to conflict management, as defined by the researchers in this study. The time considered spans the complete time horizon of an impact event, including the possible very long term preceding activities and the post-impact treatment, which are not classically included in conflict management. The definitions and boundaries of the various levels can vary between researchers; the ones presented here represent the definitions used in this study and by other research activities by the researchers. Here, impact refers to the potential consequence of not managing the conflict, e.g., it can be actual physical impact between vehicles or environment or damage due to abrupt manoeuvres.



When dealing with short- and mid-term conflict managemen<sup>t</sup> in the case of autonomous vehicles, challenges arise in the resolution step. None of the current methods relying on human awareness or accepting and effecting commands are appropriate in this case, reducing the options available to digital commands either from a ground-based or on-board computational system.

#### *2.2. High-Level Conflict Management System Requirements*

The following chapter discusses the core principles and the high-level requirements that were defined to develop the conflict managemen<sup>t</sup> system.

#### 2.2.1. Performance Requirements

The proposed conflict managemen<sup>t</sup> system is aimed at serving a wide variety of users, both aerial and ground based, implementing short- and mid-term tactical conflict managemen<sup>t</sup> solutions. As such, the system needs to be able to detect conflicts and issue appropriate commands within the available time frame, which is in the order of 10 s or less. Due to the high magnitude of unmanned traffic predicted in the future, all steps of the conflict managemen<sup>t</sup> process need to work autonomously, without human interaction.

#### 2.2.2. Technology and Solution Independence

In addition to the high number of operations and users, it is also predicted that a diverse variety of users will need to be served. While there are initiatives that are (at least in theory) standardised in the automotive world (OBD—on board diagnostics—connectors for example) it is very unlikely that autonomous capabilities and communication solutions developed by individual manufacturers will be standardised in the near term. The case is similar in the UAS industry. There are initiatives, mostly in the open-source world, to standardise communication and control protocols, for example the MAVLink (micro air vehicle link) protocol. Individual manufacturers, however, still mostly rely on proprietary closed solutions, where even different models from the same manufacturer might be using different communication and control systems. This is the case for DJI for example, the most widespread civilian-use COTS (commercial off-the-shelf) drone solution. It is unknown whether any one particular technology will be adopted in the future as a standard solution, and at the moment there seem to be no initiatives either. Regarding

communication solutions, there are also no dominant approaches; however, V2X (vehicleto-anything) communication methods are rapidly changing active research field today. Potential solutions include mobile network-based (4G/LTE, 5G), LORA, WLAN, Bluetooth and other technologies in use today. While 5G solutions have been demonstrated, linked to this project [43], and it is likely that mobile network-based solutions will become more and more widespread in the future, it cannot be said with certainty that it will become the definite solution in the future. As such, the conflict managemen<sup>t</sup> system needs to be technology and solution independent.

#### 2.2.3. Wide Variety of System Users and Components

The system needs to accommodate a wide range of users. In addition to the technological aspects already discussed, the specific type of users also represents a wide range. In case of UAVs, the smaller multi-copter designs are of primary concern, but VTOL (Vertical Take-off and Landing—either small or aerial taxi sized) and potentially fixed-wing aircraft and others (balloons, airships, etc.) could be present in the airspace. In terms of groundbased users, in addition to self-driving cars, manned vehicles, cyclists and pedestrians can also be present in a traffic situation. The environment and obstacles can be regarded as a third source of "users" in a conflict managemen<sup>t</sup> system, as impact with these elements must also be avoided. As such, from the planning stage the system is aimed at integrating data from GIS (Geographic Information System) systems, including Geofence definitions.

#### 2.2.4. No Controller Development

A very important criterium for the system is that the individual users must be able to operate using their own control algorithms. That practically means all hardware must work in a plug and play manner. It is infeasible to rely on the conflict managemen<sup>t</sup> system for low-level control for 2 reasons. For one, it would involve extensive development for each vehicle as a dynamic system, not even considering payload, centre of gravity and other configuration possibilities. This is an activity that the provider of a conflict managemen<sup>t</sup> system cannot afford to do. Another aspect is responsibility. The conflict managemen<sup>t</sup> system must provide a command to follow to avoid the conflict, but it should not decide how to follow the command. Individual control tuning, dynamics, limitations and assistance systems (collision prevention, obstacle detection, lane assist, etc.) need to function individually and in addition to the conflict managemen<sup>t</sup> system. Some, such as the assistance systems or geofencing, essentially act as additional safety barriers for the managemen<sup>t</sup> of conflicts. The same logic applies to mission planning software. For UAV these tools are referred to as ground control stations (GCS or GC). All mission (autonomous flight)-capable UAVs have some form of GCS for the operator to setup the desired mission. The concept is very similar for UGV, but the mission planning software will likely appear as a map application integrated into the dashboard or similar. For UAV the open-source world (PX4 and ArduPilot are the most common control software) has well-established solutions (QGroundControl, MissionPlanner or APM), and other manufacturers use their own solutions (DJI Pilot, for example). There are also dedicated commercial solutions aimed mainly at commercial and enterprise level users, such as UGCS or Auterion Mission Control. These GCS software must be used by the operator, as the responsibility of planning and executing their operations cannot be taken up by the conflict managemen<sup>t</sup> system provider.

#### 2.2.5. Modular Software Solution

The software solution needs to have modular architecture. This enables the integration of the various types of users by developing type-specific communication interfaces for a given solution (MAVLink, DJI SDK, etc.). This also futureproofs the system as upcoming new types can be included by adding new modules to connect them. The conflict management algorithms also benefit from the modular architecture, as different types of detection and resolution methods can be developed as separate modules and their performance tested. The development process follows LEAN software development principles, aiming

to minimise the time of each plan–build–test–evaluate cycle when new modules and functionality are implemented. To achieve this, it is important to start the development with the most widespread and available technologies, solutions and software stacks.

#### 2.2.6. XITL Tools Integration

XITL [44] (X = anything in the loop) integration is a generalisation of using testing tools such as HITL (hardware), SITL (software), VITL (vehicle) and so on. This enables the development, testing and even operation of systems using a wide variety of sources, real world or simulated. Advanced concepts such as digital twins [45] or scenarios in the loop [43] have already been demonstrated for UGV as part of the Hungarian Autonomous Systems National Laboratory, in cooperation with the ZalaZone Automotive Proving Ground facility [46,47]. Simulation tools are especially beneficial in the case of conflict management, primarily because the consequences of mismanaging the conflict are minimal as opposed to physical impact between real-world vehicles. There are also additional benefits in terms of development time, cost, flexibility, safety, security and also convenience. While UGV operations can also be affected, especially in the case of UAVs, simulation tools also remove constraints arising from weather. Another major benefit is that significantly more data can be collected or generated from simulations, with more accuracy; measurement does not disrupt the phenomenon to be measured, and there is no measurement of uncertainty and noise. As an added benefit it is very easy to generate "fake" data programmatically, which can be used to test situations that would be difficult to orchestrate using real hardware or even high-fidelity simulators. The proposed conflict managemen<sup>t</sup> tool needs to be designed so that it can integrate with XITL tools seamlessly.

#### 2.2.7. Summary of Requirements



Figure 1 shows the architecture of the proposed framework concept. The key development areas are coloured in blue; these are the components that need to be built from scratch to achieve the desired functionality. The communication and user layers are items that are intentionally left to use only COTS solutions. The detection and resolution methods are coloured differently. The framework can work by integrating these algorithms, even if they are already developed or available only as a black box tool. The tools used can be changed in a modular way in the system. The final piece of the system is the GUI (graphical user interface), which enables a human supervisor to monitor the system and interact with it if necessary. It is not coloured blue, because there is no strict need to develop an independent GUI; the proposed system could integrate with other tools with existing GUIs, as long as appropriate interfaces are developed for data sharing.

**Figure 1.** Conflict managemen<sup>t</sup> system high level architecture.
