Autonomous Systems for Defense Applications

Abstract

The numerous advantages of using UAV platforms, alongside with recent scientific developments in the field of autonomous vehicles in general and the lower production costs for such platforms, have increased interest in their usage in a variety of defense applications. This work investigates swarming in defense applications and provides information about the crucial modules needed for a swarm to operate and the main missions in defense applications that the swarms can be used to enhance the situational awareness.

1. Introduction

Swarming technology is a disruptive and game-changing technology that can change the way militaries conduct operations. Swarms can be employed in surveillance and reconnaissance to autonomous attack operations in defense, − and are the two potential applications of Unmanned Vehicles (UV) swarms that gaining significant interest in the last decade. UV swarms can be used to capture important information, and intelligence by monitoring certain areas for adversary positions, movements, and activities and behaviors.

Swarms, due to their collaborative nature, can cover large regions of an area, providing real-time situational awareness and decreasing the risk to human soldiers. Target Detection UV swarms can detect and track high-value targets like hostile vehicles, weapons systems, or troops, using advanced sensors, artificial intelligence, and machine learning to detect, identify, and track prospective targets, enabling more precise targeting and engagement in a faster and more accurate manner. Moreover, UV swarms can be used in offensive operations like air attacks or naval battles and overwhelm enemy defenses by coordinating their movements and assaults, improving the chances of mission success.

UV swarms are also capable of carrying out coordinated electronic warfare strikes, such as jamming enemy communications or damaging radar systems. Additionally, utilizing advanced decision-making algorithms, UV swarms may safeguard military assets by establishing a protective perimeter around high-value targets such as ships or stations. They can identify, track, and confront prospective threats, making successful strikes more difficult for opponents. UV swarms can be used to restock and sustain frontline soldiers by delivering critical resources like gasoline, ammo, and medical supplies. This reduces the need for risky convoys and human engagement in dangerous places.

One of the key advantages of swarms is their ability to operate autonomously, with minimal human supervision. This allows them to adapt to changing situations in real time and respond to threats quickly and effectively. Swarms can also operate in a coordinated manner, allowing them to perform complex tasks that would be difficult or impossible for a single vehicle.

Another advantage of swarms is their low cost and scalability. Small, low-cost unmanned vehicles can be produced in large quantities, allowing military forces to deploy swarms in large numbers. This can help to offset the numerical advantage of an enemy force and increase the effectiveness of military operations. Overall, swarming technology has the potential to be a game-changing technology on the battlefield.

This work investigates swarming in defense applications and provides information about the crucial modules needed for a swarm to operate and the main missions in defense applications that the swarms can be used to enhance the situational awareness.

2. Swarming technology

In defense applications, swarming technology refers to the coordinated behavior of multiple UVs working together to achieve a common goal. This technology provides numerous benefits by leveraging UV swarms’ collective intelligence, adaptability, and resilience.

Individual vehicles can work together as a cohesive unit thanks to effective communication and coordination algorithms, while advanced autonomy and artificial intelligence (AI) capabilities allow the swarm to make decisions and execute tasks without constant human intervention. UV swarms are also scalable and modular, making them appropriate for a wide range of defense applications.

Furthermore, they provide greater redundancy and resilience than single-platform systems, ensuring that the loss of one or more vehicles does not have a significant impact on overall mission effectiveness. UV swarms can operate in a variety of domains, including air, land, sea, and space, making them adaptable to a wide range of defense applications, including surveillance, reconnaissance, electronic warfare, logistics, and combat operations.

This section examines three key characteristics of swarms for defense applications: the most important capabilities and technologies that enable the development of efficient swarms, the primary architecture schemas used when designing a swarm, and some of the operation types that are frequently combined to describe a swarm’s mission for defense applications.

For defense applications, three major technology modules must be considered: perception, task allocation & decision-making and path planning & deconfliction [1] as illustrated in Figure 1. From perception capabilities to swarm communication protocols [2] and routing [3], each module plays an important role in swarm performance and robustness.

Figure 1.

The overview of the main characteristics.

2.1 Perception

The ability to perceive is critical for the success of swarms of unmanned vehicles. Perception allows the swarm to observe and comprehend its environment, recognize obstacles and desired targets, and retain situational awareness. The swarm’s effective perceptual abilities allow it to operate in complex and dynamic missions and complete a wide range of complex tasks with high efficiency. Effective perception ensures that the swarm can navigate through the environment, communicate, and accomplish tasks efficiently and precisely. Machine learning (ML) and AI methods, as well as sensor fusion, allow a swarm to acquire situational awareness at a previously unattainable level. UV swarms can benefit from ML and AI by processing and analyzing big data from many sources, allowing them to identify patterns, make predictions, and adapt to changing situations. These approaches can assist swarms in distinguishing desirable targets, detecting obstacles and determining the optimal course of action, and adjusting their behavior in real-time.

The technique of merging and combining data from several sensors to provide a more accurate and comprehensive picture of the environment is known as sensor fusion. Data from cameras, LiDAR, radar, sonar, and other sensors may be included and so UV swarms can gain a more detailed and reliable situational awareness by fusing input from several sensors, which is critical for navigation, communication, and task execution.

Several works address the topic of efficiently merging and combining data from different types of sensors [4]. In the work presented in [5], the authors introduce a system including static sensor network design, mobile sensor tasking, and information and algorithms for use in commercially available unmanned vehicles with low computational power requirements, flexibility, and reconfigurability. A practical demonstration of swarm technologies in a scaled outdoor environment aimed to test and validate the effectiveness of various algorithms and techniques for persistent monitoring in military and law enforcement applications. The swarm technologies addressed several critical aspects, including task allocation, situational awareness, target detection and tracking, and cooperative guidance. The greedily excluding technique was used for task allocation, which ensures near-optimal allocation in real-time and allowed the swarm to adapt and respond effectively to rapidly changing environments. Enhanced situational awareness was achieved through Unmanned Aerial Vehicle (UAV) aerial sensing, a general framework for autonomous behavior monitoring, and trajectory analysis tools. Sensor fusion technology and decentralized tracking algorithms supported automatic target detection and tracking. A reactive and distributed cooperative guidance law was designed for mobile vehicles, addressing mission and safety objectives, as well as interactions between the vehicles and the static sensor network.

2.2 Task allocation and decision-making

Task allocation aims to analyze existing tasks that need to be performed and distribute them to the available agents. More specifically, it assigns tasks to an agent or a group of agents intending to find an optimal or near-optimal mapping between agents and tasks. Effective task allocation guarantees that each agent is allocated a task that matches its capabilities, maximizing the utilization of available resources. Swarms can compensate for the failure of one or more agents, and task allocation aids in the distribution of decision-making, allowing the swarm to quickly adjust to changing conditions. In general task allocation also improves adaptability, scalability, and decision-making speed, allowing swarms to be more successful in dynamic and uncertain environments and providing efficient resource utilization, increased resilience, and fault tolerance.

Under multi-constrained conditions, multiple UAV task allocation models were established [6]. This work used the Multi-Agent Systems (MAS’s) abilities of environmental perception, collaboration, and self-learning to establish a distributed immune multi-agent algorithm (DIMAA) by using a high-performance artificial immune system for solving complex problems. The following operators are proposed: immune memory, neighborhood clonal, neighborhood suppression, neighborhood crossover, and self-learning. Simulation experiments validate the proposed algorithm’s performance under three dynamic conditions task allocation, new targets, damaged UAVs, and pop-up weather threats. The simulation results revealed that the proposed algorithm has a fast solution speed, a high optimization capability, and the ability to balance the missions of the UAVs and reduce communication loss. Furthermore, in a changing task environment, it can still obtain good task allocation results. All these results mean that the proposed algorithm has better global optimization capability, dynamism quality, and robustness.

In [7], the paper analyzes a described scenario to answer the question “Should the robots compete, or should they cooperate?”. There are two game-theoretical algorithms developed. The competitive algorithm plays games with each drone and its neighbors while looking for the Nash Equilibrium. The cooperative one defines electoral systems that allow drones to vote on the task allocations they prefer for their neighbors. Both algorithms are extensively tested in multiple scenarios with different features. After the experiments, the question can be answered: “The robots should cooperate!”.

The work presented in [8] describes a probabilistic strategy for assigning specialized individual agents within a robotic swarm to match limited tasks. The proposed approach evaluates the probabilistic fitting of the available robot individuals based on the requirements imposed by the current task, which takes the form of a recognized target object in a specific environment, on the assumption that each agent possesses specialized capabilities. To evaluate a task-agent fitting score among all accessible agents, a formal matching scheme is devised. As the best responder, it assigns the most qualified and available specialized robotic agent to do the recognized task. A simulation study is presented to validate the efficiency and robustness of the proposed approach.

The work presented in [9] illustrates the Bird Swarm Algorithm (BSA), a new bio-inspired method for tackling optimization problems. BSA is based on swarm intelligence collected from bird swarm social behaviors and interactions. Birds primarily engage in three types of behaviors: foraging, alertness, and flight. Birds can seek food and avoid predators through social interactions, giving them a high probability of survival. BSA develops four search techniques connected with five simplified rules by modeling these social behaviors, social interactions, and related swarm intelligence. Simulations and comparisons based on eighteen benchmark problems demonstrate the effectiveness, superiority, and stability of BSA.

2.3 Path planning and deconfliction

Swarms are expected to contain a high number of robots. Path planning and deconfliction procedures are therefore critical for efficient and safe swarm operations. Path planning in swarms seeks to discover the best way for each agent to take to its destination while avoiding obstacles and consuming the least amount of time and energy. Deconfliction prevents robots from colliding, enabling each agent to fulfill its duty properly. Path planning, for example, can optimize the trajectories of the agents in a surveillance mission to reduce overlap and enhance coverage area. A range of methodologies, including centralized and decentralized ones, can be used to achieve path planning and deconfliction. Centralized techniques entail a single entity planning and coordinating the activities. Decentralized approaches, on the other hand, involve each robot making its own path-planning decisions based on available information.

This paper [10] investigates collision avoidance algorithms for many UAVs based on geometry. By extending the collision-cone technique to UAV formation, the suggested strategies allow a group of UAVs to avoid obstacles and split if required, using a simple algorithm with cheap processing. The geometric technique incorporates dynamic limitations into the construction by using line-of-sight vectors and relative velocity vectors. Each UAV may choose the plane and direction to use for collision avoidance. An analysis is undertaken to design an envelope for collision avoidance, considering angular rate constraints and object detection range limits. Each UAV in a formation decides if the formation can be maintained while avoiding obstacles based on the collision avoidance envelope. Numerical simulations are performed to demonstrate the performance of the proposed strategies and the results indicate good performance for coordinated collision avoidance by multiple UAVs. The work presented in [11], describes a method for collision-free trajectory planning with several UAVs that detect conflicts automatically. When problems between UAVs are detected, the system collaboratively resolves them using a collision-free trajectory planning method based on a stochastic optimization approach known as Particle Swarm Optimization (PSO). The novel Maneuver Selection Particle Swarm Optimization (MS-PSO) implementation of the PSO method outperforms prior implementations. Because the scale of the problem is decreased, the execution time is lowered, and other types of maneuvers can be used to resolve discovered conflicts: course/heading, speed, or altitude modifications. The MS-PSO has been validated with simulations in scenarios with multiple UAVs in a common air space.

3. Applications

Unmanned vehicle swarms communicate information through communication protocols, which are established by the mission’s unique requirements as well as the swarm’s features [12]. An efficient architecture is required for exchanging information among a swarm of unmanned vehicles, allowing them to operate safely and accomplish their missions.

To ensure efficient information exchange in a swarm, it is crucial to choose the appropriate architecture and communication protocols that are scalable, reliable, adaptable, and interoperable. The architecture and protocols should be scalable to handle the increasing number of vehicles in the swarm without affecting performance. The communication system should be reliable, ensuring accurate, timely, and consistent information exchange, while also being resilient to adversarial forces.

Furthermore, the architecture and protocols must be adaptable to changing mission requirements and swarm compositions, allowing the swarm to respond to unexpected events or challenges effectively. Interoperability between various types of vehicles, sensors, and systems is critical for seamless cooperation among heterogeneous swarm members and the incorporation of new technologies. Finally, low latency and scalable bandwidth are critical for time-critical missions that require quick decision-making and response times.

The three primary architectural strategies for developing swarms of unmanned vehicles for defense purposes include Centralized, Decentralized, and Hybrid architectures.

Hybrid architecture is appropriate for larger swarms and complex missions requiring a high level of coordination and autonomy between vehicles. The centralized component can provide higher-level decision-making and overall mission objectives, whereas the decentralized component can provide local decision-making based on real-time data. This approach ensures the swarm’s ability to adapt to changing mission requirements, maintain a robust and resilient communication network, and remain effective in dynamic and uncertain environments. Finally, the architecture chosen will be determined by the specific mission requirements, the number of vehicles in the swarm, and the level of coordination and autonomy required for mission success.

Swarms of unmanned vehicles present a plethora of military applications and can conduct various missions. Some key examples of operation types of autonomous swarms are identified here. The presented operation can be combined to create a series of missions. In Figure 2, the main applications of the swarms of UVs are illustrated.

Figure 2.

The overview of the main applications.

3.1 Full and persistent area coverage

For low-level automated coordination of a swarm of robotic agents, the authors in [13] suggest using a straightforward force law that draws inspiration from both swarm intelligence and classical physics. Authors show how the coverage issue can be solved by using this control law to build a lattice of sensors for an airborne surveillance application.

They aim to find the correct separation distance R to maximize overall coverage and the constant G that affects the rate at which the swarm converges to a stable configuration using a genetic algorithm for optimizing the control law for a relatively simple scenario with a limited search space having a simulation of 7 robot agents in a hexagonal lattice formation which can control their velocity and sense the range and bearing of neighboring agents. They found that this approach is a viable option for learning parameter settings within the control-law framework and application/problem setting considered in this paper.

3.2 Area search

Swarm capabilities present a great benefit to area search operations due to the easiness of distributing the work. The task of area search differs from the previously mentioned operations since target specificity can be applied. Moreover, optimizing swarm performance means identifying the target in as little time as possible, hence it is unnecessary to search the whole area of interest. Area search algorithms play an important role in the effectiveness of the swarm, the authors in [14] studied data fusion to enhance swarm control and decision-making. To do so the search region was discretized by turning it into cells and a probability map is built on each UAV. A distributed fusion scheme was developed which showed that all probability maps converged to the same one that reflected the true environment. In [15], the authors propose three cooperative search algorithms for military applications using a heterogeneous swarm. A decentralized task allocation has been developed, and the tradeoff between prediction-based and search-based approaches is examined. The results indicate an intelligent use of prediction can be helpful. Lastly, the authors in [16] have developed an efficient cooperative search that takes into account potential intelligence being present on the target of interest in the form of evasion tactics. The two search algorithms proposed to enable the swarm to perform decentralized area search for one or more targets. Both algorithms provided exceptional results and it is worth noting that one of them also can search in areas of undefined shape and size.

3.3 Area surveillance

In [17] the authors present a project called “ASIMUT”, a distributed decentralized surveillance system that utilizes UAV swarms to collect information and generate data of higher quality. The system is composed of a Command and Control (C2) system and a set of heterogeneous UAVs that constitute swarms. The UAVs are equipped with various onboard sensors including long-range radar and electro-optical/infrared cameras to monitor the movements of ground vehicles. The system has a basic layered model structure consisting of three layers, each serving specific purposes: Collaboration of UAVs, Detection by UAVs, and Exploitation of Data. The ASIMUT system is designed to perform automatic detection of targets in a full motion video, and the process has been divided into several steps, including work on optical flow, detection of points of interest, projective image transformation, compensation for camera movement, and motion detection.

Work in [18] proposes an adaptive recommender-based system for swarm guidance and control for aerial surveillance tasks that can adapt the level of swarm autonomy in real time according to the needs of human operators. The system uses a sheepdog shepherding control method for the swarm, which provides a single point of control for a large swarm of autonomous UAVs. A recommender system powers the suggested system, giving the human operator optimized recommendations in real time for maximizing work performance. The frequency with which the system provides those recommendations to the operator is controlled by variable adaptation, which adjusts according to their cognitive load in real time. A particle swarm optimization algorithm is used to find the best strategies for shepherding the swarm of UAVs, and its effectiveness and efficiency were evaluated through simulation experiments. However, the study was limited to simulation-based experiments, and the authors plan to investigate the performance of the proposed approach in an adaptive human-swarm interaction context in the future.

3.4 Target tracking

Target tracking: Commonly a target tracking operation involves one target and one vehicle. The scope of the vehicle is to online plan its path based on its sensory data and its estimation of the target’s location and in some cases the predicted behavior or future location of the target [19]. The vehicle must guide itself to constantly follow the target. With the introduction of swarming capabilities, the target tracking problem can be augmented into a multi-vehicle problem, tracking a single [20] or multiple targets [21]. Moreover, the concept of target tracking has been extended from the ground to aerial targets where a swarm of autonomous UAVs are tasked with tracking another malicious UAV [22]. In this work, the authors explore different methods of UAV formation to effectively tackle the tracking of a UAV with superior flight capabilities. Lastly, multi-agent target tracking using Fixed-Wing UAVs has been researched extensively along with formation control and performance outcomes in [23] alongside a proposed architecture for the above in [24]. Overall, the design of swarms for defense applications requires careful consideration of several factors, including mission requirements, swarm size, communication capabilities, and computational resources.

4. Conclusions

In conclusion, unmanned vehicle swarms have the potential to transform military operations. However, in order to ensure efficient information exchange in a swarm, it is critical to select scalable, reliable, adaptable, and interoperable architecture and communication protocols. Centralized, decentralized, and hybrid architectures are the three primary architectural strategies for developing swarms of unmanned vehicles for defense purposes. The architecture selected will be determined by the mission requirements, the number of vehicles in the swarm, and the level of coordination and autonomy required for mission success. The main modules needed for a swarm to operate efficiently are task allocation, path planning, object detection and tracking and deconfliction. The combination of these modules can provide the necessary swarm intelligence for military applications for missions like full and persistent area coverage, area search, area surveillance, and target tracking.

Conflict of interest

The authors declare no conflict of interest.