By Robbin Laird
The development and incorporation of autonomous systems into the Australian fifth-generation force requires a close working relationship between defense industry and the Department of Defence and the services. But also requires meeting the challenge already seen with regard to the introduction of software upgradeable systems such as F-35, P-8 and Triton. That challenge is for the warfighters to be able to drive advantages through rapid code rewrites and being able to develop and generate with the autonomous systems advances, disposable systems as well.
This was well put in the discussion with Jason Scholz. As Scholz put it in our interview, with the Centre’s focus on the “smart, the small and the many”, compared with traditional “complex, large and few” manned systems, code rewriting can be much faster.
It is also the case that digital engineering and digital twins is changing how all platforms are designed and supported. But in the case of next generation autonomous systems, the entire life cycle of these “smart, small and many” systems is very different. “They will be attritable; there will be no need to develop and maintain 30 years of systems engineering documentation – some of these might be used only once or a few times before disposal. When you need to adapt to the threat, digital engineering supports fast redesign and T&E in the virtual, and to add a new capability you just download it as software.” Scholz says.
At the seminar, there were presentations by six industrial partners.
Boeing Australia focused on the development of the Loyal Wingman.
With the space which Australia has along with the technical capabilities in the country, Australia is in a good position to be a leader in autonomous systems. According to Andrew Glynn, Boeing Australia, the Airpower Teaming System, which has been established to develop loyal wingman is an important approach to shaping a way ahead. He noted that “the loyal wingman program is an experimentation program with RAAF to provide insights in the use for this type of capability.” Clearly, from Boeing’s perspective, the standup of the program in Australia is not simply about meeting the ADF needs, but provides a launch pad for broader global development.
Glynn underscored that the teaming approach is designed to develop capabilities which can work with existing fleets. “Once ATS is operational, it will assist the RAAF to protect and project force by addressing quantity at the right quality.”
He noted that the program went from concept to first flight in under three years and that was enabled by the fact that the platform is built on a completely digital foundation.
A second presentation was given by Dr. John Best of Thales, Australia.
In his presentation, he focused on the opportunity to build or enhance sovereign defence capability built around autonomous systems. But to do is not just an industry challenge, it is in his words “an enterprise challenge.” What is required is defence and industry to shape an ecosystem where they can work together to deliver the desired outcome. He also underscored that Australia has the building blocks in place to deliver such a solution.
A particularly important aspect of shaping a way ahead is to shape more effective ways to manage complex systems, and within that context the autonomous piece of complexity. And the continuous learning piece which comes from use in the real world of combat and gray zone operations needs to be fed into the evolution of systems, This means that the “learning piece” needs to be worked into the contractual relationships which industry has with Defence.
A third presentation was by Dr. Andrew Lucas of Agent Oriented Software.
He provided a very helpful clarification of how autonomous systems are different from automatic or automated systems. For Lucas, an autonomous system is goal directed and capable of rational reasoning with regard to those goals. An autonomous system can perceive its environment, determine if the environment affects its goals and then takes actions in alignment with its goals up against the operational environment. It is able to balance proactive and reactive behaviors. This is in contrast to automated system which simply follows a task list or script.
He then provided a way to look at the evolution of autonomy within the context of man-machine operations. His schema is built around the axis of increasing levels of delegation to the system. It moves from the machine simply providing timely advice with the humans making all the decisions; to a semi-autonomous relationship in which the machine is subordinate to a certain level of human authority; and then operating fully autonomous in which the machine operates without huma control or direct oversight.
Dr. Lucas then discussed the relationship between AI and autonomous systems. AI is defined by him as the “theory and development of computer systems able to perform tasks normally requiring human intelligence.” AI enables machine learning which is to be understand in terms of “a computer program can learn and adapt to new data without human intervention.” And this then enables to creation of intelligent agents which can be defined as “a computer program that is capable of perceiving and interpretating data sensed from its environment, reflecting events in its environment, and taking actions to achieve given goals without permanent guidance from its user.”
He underscored that to achieve the capability for autonomous systems, several trend lines in the evolution of AI and machine learning need to coalesce or synthesize. From this point of view, he noted that there were several AI trends, thrusts and directions which are shaping a way ahead. These included image analysis and recognition, or sensor identification and assessment, speech recognition and the ability of the human and machine enablement of dialogue, intelligent search and assessment, and intelligent software agents able to work through big data. With such capabilities evolving then situate AI and robotics are enabled with the possibility of machine learning.
In other words, there is clearly a pathway already being shaped to autonomous systems, but investments, experimentation and development remains. And he argued that Australia needs to invest in this development to deliver on the promise of autonomous systems both for the civil and defence sectors.
A fourth presentation was by Northrop Grumman, but done by remote participation from the United States.
Northrop is a key player in Australian defence, and in the area of man-machine teaming, clearly the arrive of Triton is part of the way ahead. But the key focus which the company has upon expanding the types and nature of sensor networks and finding ways to leverage those networks to deliver timely decision-making data is a key part of the way ahead.
Clearly one way already that Northrop is contributing to such a way ahead is with regard to how the F-35 functions as a C2 and sensor “flying combat system” and enables it to fight in an 8-ship formation as a wolfpack. This was not mentioned at the seminar but clearly is a key enabler of the way ahead for the ADF which also provides a foundation for shaping the broader efforts to shape an integrated distributed force able to work a diversity of sensor and communication networks to deliver the desired combat effects.
In the Northrop Grumman presentation, a key point is to shape a way ahead for battleship control where sensors can track operations within that battlespace. Clearly, remote sensors are proliferating and the importance of AI and autonomy for helping the tactical and strategic decision makers to que systems to perform key and dynamic tasks will become increasingly important in the extended battlespace.
A fifth presentation by defense industry was given by Air Vice Marshal Bill Henman (Retd), who is currently a strategic advisor Air and Space for Raytheon Australia.
His presentation focused on a key issue, namely, the countering of autonomous systems. This subject is a key one both in terms of understanding how adversaries are using their own automated systems, which needs to include not simply the question of technology but concepts of operations, which in turn allows you to examine the vulnerabilities of your own systems.
Autonomous systems have vulnerabilities which can be built around spoofing, data corruption as well as braking effective sensor network to C2 communication flows. At the heart of the autonomous system impact is the control challenge. Henman defined control as “enabling friendly manoeuvre while denying adversary manoeuvre. He argued as well that control also involves deconfliction of friendly manoeuvre in terms both of kinetic and non-kinetic fratricide.
In his presentation, he highlighted how autonomous systems can be disabled or eliminated in combat. This can happen by disrupting or spoofing the detection and tracking systems. Or this can happen through various softkill options, such as jamming and spoofing, electronic attack or cyber intrusions. He then highlighted a number of hardkill options both kinetic (close in weapons systems) and non-kinetic, notably high energy lasers and directed energy systems.
These vulnerabilities clearly provide a realistic cautionary note to how autonomous systems can be introduced and used. And this is why their introduction into infrastructure defense such as extended port security or into logistical support, up to and included logistical management systems, are good places to start and to sort through how to do so enhancing rather than compromising national defense and security. `
He ended his presentation by proposing a number of key actions as the way forward with regard to autonomous systems is worked.
First, it is important to create the optimum environment to allow industry, Academia, Research and Development communities to be able to create and field innovative control soltuions. Second, it is crucial to be able to understand the operational implications of an adversary who has greater freedom of manoeuvre with respect to all dimensions of governance of autonomous systems than do the liberal democracies. Third, there needs to be as much focus on the development of superior autonomous decision cycles as there is upon the development of the sense and respond control systems enabling autonomous systems. And finally, he argued for shaping autonomous decision making ‘tutorship’ in roder to build the future confidence needed to send autonomous systems into combat as the force evolves with more agile, multi-domain C2 capabilities.
A final defense industry presentation was by Lockheed Martin Australia. Dr. Tony Lindsay, formerly of the Australian Department of Defence’s Department of Science and Technology or DST, provided insights as well.
Lindsay provided an overview on the decision cycles for the force and how autonomy might provide both enhancements to such decision making as well as more capability to leverage the proliferating sensor networks.
He argued that AI and autonomous systems have the possibility for impacting on all parts of the decision cycle, such as persistent ISR, sensor cuing, scene understanding, and cognitive battle management. They could just as well undercut decision effectiveness if not worked into an effective decision-making cycle. This requires allowing the operator to act using relevant information, be able to align with commander’s intent and provide for system adaptation.
The broader shift associated with the fifth-generation force transformation, namely distributed force which can be scalable and integratable provides a solid foundation for managing proliferating sensor networks as well as finding ways to use autonomous systems within distributed decision-making systems.
In short, shaping a new enterprise approach already underway driven by software-upgradeable systems, and reworking how platform providers work with the forces and defence, presages the significant changes underway into which autonomous systems will be introduced and integrated within the force.
The featured graphic is taken from Dr. Tony Lindsay’s presentation to the Seminar.