In the first of The Central Blue’s #ADFRAS2040 series, Drs Gary Waters and Patrick Bigland reflect on the wide-ranging potential of robotic and autonomous systems. Though the path to full autonomy will be a long one, Waters and Bigland draw attention to some of the initial steps that will start the ADF down the right path.
The advent of robotics and autonomous systems (RAS) has seen the convergence of technologies such as artificial intelligence (AI), swarming, alternative energy, additive manufacturing, and advanced materials. The potential implications for the use of RAS by the Australian Defence Force (ADF) are substantial as they range across autonomous systems themselves (platforms, devices, and agents); human-machine teaming; countermeasures; and autonomous behaviour. As the ADF contemplates these implications, fuelled by our allies adopting RAS, there is both an inevitability of increasing use of RAS and an advantage to be gained from the dogged pursuit of RAS as they will, increasingly, offer high levels of autonomy, stealth, and persistence; and will improve the ADF’s overall military capabilities.
As the ADF determines how best to exploit RAS in future, to gain and maintain advantage across the continuum of competition and conflict, including in countering the use of RAS by future adversaries, many aspects will need to be addressed. Recent work by the NATO Security & Technology Organisation provides some salutary observations for autonomous systems themselves, the challenge of human-machine teaming, countermeasures, and autonomous behaviour such as swarming.
Unmanned vehicles for air, sea, and land may be remotely piloted or may act at varying levels of autonomy throughout a mission. Key enabling technologies include stealth, structures and materials, propulsion, performance, stability, and control; and the processes around qualiﬁcation and certiﬁcation, and design. AI-enabled autonomous systems (especially smallsats) have emerged in recent years, extending RAS into space.
The major recent development in RAS has been the use of virtual software agents for both offensive and defensive action, thus underscoring the value of information and the importance of the cyber dimension. Achieving a level of trusted, intelligent autonomy will result in systems that work seamlessly with the warﬁghter, enhance warﬁghter trust in the systems, lead to a significant force multiplier, and substantially increase operational tempo.
Notwithstanding the recent developments in autonomy and the desire for fully autonomous systems in the long-term, for the near-term, full autonomy of unmanned systems will be practical only for more straightforward tasks. Thus, semi-autonomous systems will have more impact on operations in the short- to medium-term, with the warﬁghter retaining control and ﬁnal decision-making. These systems will enhance situational awareness, effective reach, and reduce the risk to personnel participating in the operation.
Developments in detection avoidance will be key to RAS effectiveness by reducing vulnerability to detection methods (radar, infrared, and sonar). These signature requirements are also likely to inﬂuence and drive platform design. At the same time, new miniaturised, low power sensors will be required to support embedded AI, situational awareness, and increase intelligence, surveillance, and reconnaissance (ISR) capabilities.
Commercial needs for autonomous vehicles are contributing to the development of more powerful and cheaper sensors. These developments should help improve ISR, logistics, cyber defence, and targeting in the military realm, thereby substantially increasing operational effectiveness. Maximum benefit, however, will only accrue from high-quality, easily accessible, and readily available data.
Autonomous systems will also need to be used to characterise virtual and physical environments. The associated challenges with the fusion and categorisation of information will need to be met, which will demand the use of advanced statistics, combinatorial optimisation, statistical decision theory, and mathematical game theory. Robust, systematically evolving, and adaptive fusion engines for military multiple purpose applications will need to be modular, controllable, and embedded via interoperable interfaces.
Future developments in mini, micro, and nano unmanned systems will enable the use of swarming tactics and will have a profound impact on operations. This will be especially evident in land force operations, as situational awareness will be increased, the soldier’s physical and cognitive workloads will be reduced, sustainment will be improved, movement and manoeuvre will be easier, reach and range will be increased, and force protection will be enhanced.
For maritime operations, autonomous systems will assist in mine countermeasures (MCM), denied area ISR, anti-submarine warfare (ASW), environmental characterisation, signals intelligence (SIGINT)/electronic intelligence (ELINT) collection, operational deception, and operational support.
In the air environment, autonomous systems can assist predominantly in the ISR, electronic warfare, broad area maritime surveillance, and attack roles. There is potential for the F-35 to be coupled with a swarm of unmanned combat air systems in future. The first Boeing unmanned Loyal Wingman aircraft has rolled off the production line: this aircraft uses artificial intelligence to extend the capabilities of manned and unmanned platforms.
The use of RAS in logistics has the potential to increase operational availability, weapon system effectiveness, and overall logistics system effectiveness and efﬁciency, as well as reduce casualties.
There will be signiﬁcant human-machine teaming challenges across the spectrum of autonomous operations as these systems are used to enhance traditional capabilities in the land, maritime, and air environments. While some levels of autonomy have been introduced in recent unmanned systems, autonomous systems currently lack the intelligence to reduce manning requirements, reduce warﬁghter cognitive load, or increase the pace of operations. Trusted intelligent autonomy will enable capabilities that are not currently possible, such as long-duration unmanned underwater vehicles, where the vehicle must be able to work for months without human intervention or communication.
A fundamental part of trusted, intelligent autonomy will be collaborative autonomy, where widely autonomous systems will be acting as a social-technical team (eventually with distributed tasks) under the command of a warﬁghter. Systems will need the capacity to make independent decisions and act upon these decisions rapidly, while at the same time, work as part of a human team (with attendant social, collaboration, and communication issues). This level of trusted intelligent and collaborative autonomy will result in systems that work seamlessly with the warﬁghter, can build warﬁghter trust, and act ultimately as a signiﬁcant force multiplier.
The technical limitations of human control must also be considered and mitigated as RAS are subject to problems of automation bias, lack of situational awareness context, and moral buffering (a sense of emotional and physical distance of the human operator). Finally, as the ADF moves towards the seamless integration of RAS, the socio-technical implications, especially around team behaviour and collaboration, will need to be explored and better understood.
While a swarming system will assist in this collaboration and provide a degree of robustness against the failure or destruction of individual nodes in the swarm, it will present an increase in organisational complexity. Thus, the challenge of self-synchronisation will have to be addressed. The solution lies in the use of machine learning. As learning-enabled RAS becomes more widely deployed on the battleﬁeld and in other uncertain environments, it will be necessary to develop innovative system designs, as well as new analysis, testing, and veriﬁcation and validation (V&V) methods. Success in this area will allow more rapid deployment of learning-enabled systems, as well as reducing operational risks.
Force protection will be improved through the use of RAS. Use of intercepting counter swarms, electronic countermeasures, lasers, and other-directed energy systems offer some options. However, the broad range of autonomous systems that can be arrayed against the ADF presents many technical challenges for the development of effective countermeasures. As these countermeasures are developed, counter-counter unmanned system technologies will need to be developed.
As doctrine is refined to accommodate RAS technologies, the ADF needs to note that RAS technology is not truly independent. Indeed, it is a complementary technology, as the British Prime Minister and Deputy Prime Minister stated in the United Kingdom’s 2010 Strategic Defence and Security Review: ‘by the 2020s… The fast jet fleet will be complemented by a growing fleet of Unmanned Air Vehicles in both combat and reconnaissance roles.’ This viewpoint is expressed in other Five-Eyes and NATO doctrine as well. Therefore, Australia should treat RAS as a combat enhancer to other existing manned platforms, not as an independent system. With that in mind, the ADF would be best suited to either provide an additional chapter on specific areas where unmanned systems might be employed or promulgate a joint doctrine note.
In closing, the successful development of fully autonomous systems will likely be slower than anticipated, which could lead to a sense of disillusionment and frustration. Nevertheless, successful integration of autonomous systems at lower levels of autonomy will provide a major driver of capability development and future operational success for the ADF.
Dr Gary Waters and Dr Patrick Bigland are both members of the Williams Foundation and currently work together at Jacobs Australia. This article represents their views and not necessarily those of Jacobs.
Australian Defence Force, Future Warfare