Swarming and Counterswarming
Report on Applied Research Directions and Future Opportunities for Swarm Systems in Defence
The Australian Government’s 2020 Force Structure Plan outlined a total package of capability investment of approximately $200 billion over the next decade (2020). This expenditure will equip Defence to meet challenges in the future with new investments in strike platforms, littoral assets, helicopters, information effects, logistics resilience, and emerging robotics and autonomous systems. The 2018 Army Robotics and Autonomous Systems (RAS) Strategy (Australian Army, 2018) identified swarming technologies as a force multiplier for Defence, generating mass that would enable fewer humans to achieve greater output capacity than they can today. These technologies offer novel opportunities for Defence to develop cross-domain effects, delivering persistent and scalable capabilities not previously possible.
This paper outlines the current state of swarm and counter-swarm research and technologies through a high-level review of academic, industry, and coalition partner efforts. The analysis is a systematic survey of academic literature in Part One and a survey of the publicly available swarm and related programs in Part Two. The purpose is to inform the current capability state, illuminate current efforts and challenges, and identify possible options to prioritise the generation of future swarm and counter-swarm capabilities for Defence. The advancement of technology in Australia’s near region is seeing the development of disruptive engagements at an accelerated rate, with asymmetric capabilities fielded across multiple domains.
First, we discuss swarming, swarm intelligence, and swarm system. We present the term swarming as a tactic in that swarming may be a plan utilised by a force to achieve its desired end state. The Australian Defence Glossary (ADG) states that swarming is ‘[t]he large mass of autonomous systems interoperating collectively to act and respond in a coordinated effort to provide an overwhelming effect’. The ADG definition of swarming implies that the tactic may only be implemented by large masses of agents (akin to a plague) to provide an overwhelming effect. However, not all systems that use the tactic of swarming may constitute a plague. For example, a flock of sheep may employ the tactic of swarming in support of a survival strategy.
Considering swarming as a tactic applied by a system, the property of a system to realise the swarming tactic is swarm intelligence. Swarm intelligence is the collective behaviour exhibited by agents to self-organise (Bayındır, 2016), such that rules specifying the interactions between the agents are executed based on purely local information, without reference to the global pattern, and is an emergent property of the system rather than a property imposed by an external ordering influence (Bonabeau, Theraulaz, & Dorigo, 1999).
Given the tactic of swarming and the property of swarm intelligence, the definition of a swarm system offered by Abbass and Hunjet (2020) is appropriate: a team with actions of individuals aligned spatially and temporally using a synchronisation strategy. Similarly, Farina, Chisci, and Fedi (2017) define a swarm system employing the tactic of swarming as emergent behaviour arising from simple rules that are followed by individuals that does not involve any central coordination. A deduction of the tactic of swarming, realised using the property of swarm intelligence and employed by a swarm system capability, is that swarms do not require complex engineering of autonomy inside each agent. Instead, autonomy is distributed across the swarm mass, which simplifies the single-agent design and reduces costs, enabling an effect to be generated.
We define the tactic of swarming as the synchronised actions of a team of autonomous agents to provide a coordinated effect. We define the property of swarm intelligence as the ability of a team of autonomous agents to collectively self-organise. We define swarm system capability as a team of robust, flexible, and scalable agents who act collectively to achieve an effect. The ADG defines countermeasure as the reactive methods used to prevent an exploit from successfully occurring once a threat has been detected. Through this lens, and given that a swarm system will display robustness, flexibility, and scalability (emergence), we define the term counter-swarm as the offensive and defensive measures employed to deny a capability from achieving a swarming effect.
Various militaries are exploring swarm-related capabilities across several lines of effort, spanning from concepts to research and physical trials. Current naval efforts focus on autonomous underwater vehicles (AUVs) research for subsurface swarm surveying and directed energy weapons for future surface-platform swarm defence. Land efforts focus on the guidance and control of multiple uninhabited ground vehicles (UGVs), while research efforts in the air domain heavily favour teamed autonomous collaborative platforms with small-scale physical tests. Swarm demonstrations have successfully integrated and operationally tested heterogeneous swarms, ranging in numbers from 20 to 250, in military settings.
 The Australian Defence Force Concept for Multi-Domain Strike defines the five domains as land, air, maritime, space, and cyber.
 An agent is a computer system capable of autonomous action—of deciding for itself what behaviours are needed to satisfy its design objects, and capable of interacting with other agents. The ADG definition of an agent only considers a computer system, whereas Sahin (2005) defines an autonomous robot as a physical embodiment in the world, situated, can physically interact with the world and be autonomous. Both an agent and an autonomous robot, by the presented definition, can interact with other agents. We define agent in this work as an autonomous component of a team or system, capable of interacting with other agents or systems. An agent may have a physical form or be a computer system capable of autonomous control of physical actuators.
 This remains a partially incomplete definition. Traditional swarming is defined by decentralised coordination efforts; however, some modern swarming systems also incorporate elements of centralised coordination. This is reflective of the rapidly evolving state of swarm system characteristics and may require further investigation in future. As such, it is important to recognise and differentiate the delivery of a swarming effect from swarming behaviour. To deliver a swarming effect, the system requires its agents to possess not only swarming behaviour but also a shared goal and a synchronisation mechanism.
ISSN (Online) 2653-0406
ISSN (Print) 2653-0414