The UK’s Plans For Drone ‘Swarm Squadrons’: What Does Gavin Williamson Mean?
by WILLIAM MERRIN
On 11th February 2019, the British Defence Secretary Gavin Williamson revealed plans to invest £7m in a swarm squadron of drones theoretically capable of jamming enemy air defences, as part of a plan to enhance the UK’s ‘lethality’ after Brexit. Williamson said he wanted the RAF to form a new ‘concept unit’ composed of hundreds of small flying craft. The UK, he said, would ‘develop swarm squadrons of network-enabled drones capable of confusing and overwhelming enemy air defences’. And he promised to have them ‘ready to be deployed by the end of this year”. Critics were quick to point out that the idea was not quite ready for the reality and that the delivery-time was unrealistic. Chris Cole, from Drone Wars UK, was quoted in The Guardianas saying the idea of swarm drones was ‘very much at the concept stage, and it’s very unlikely he can meet the deadline of the end of the year’.
In fact, what’s interesting here is how little Gavin Williamson actually said and whether he really understands the concept he’s talking about. Firstly, it’s well-known that developments in offence such as UAVs or ‘drones’ are swinging power away from defence. It’s accepted that a sufficiently large number of UAVs operating simultaneously would probably overcome modern air-defences. There are claims such a ‘swarm attack’ has already happened. On 5th January 2018, two Russian bases in Syria were attacked by what Russia described as a ‘swarm’ of drones operated by Syrian rebels. Russia claimed they were ‘aircraft-type’ drones, with ten in one attack and three in another, each armed with a 10-pound bomb. Syrian rebels and IS have been known to use commercial drones weaponized with bombs but these were custom-built and flew pre-planned routes to precise GPS-coordinates. Russia’s Pantsir-S air-defences and electronic-warfare specialist hacking brought most of the drones down, but it has to be said that a total of 13 drones in two waves doesn’t constitute much of a ‘swarm’. It’s likely that the defences would have struggled if we’d have been talking about hundreds, or thousands, of drones of the kind seen in the November 2017 Future of Life Institute ‘Slaughterbots’ video.
The Syrian drone attack also appeared to be using human remote-control operators. Williamson doesn’t say anything more about the RAF ‘swarm squadrons’ but, given the current state of A.I. autonomy and UAV robotics, we can assume that he sees these squadrons as an extension of current, humanly-operated, remote technologies. This, therefore, would be a far less-advanced conception of ‘lethality’ than we might have assumed. Human-operated drones would require considerable coordination and would depend for the final assault on each operator being able to carry out their task without impeding the actions of another. This may be possible with up to ten drones an attack but coordinating an assault with hundreds or thousands would be beyond the human mind. Also, this type of swarm attack is a numbers game. Success would come from the numbers put into the assault and the attacker would have to assume a certain loss-rate for their drones from the defences. It’s not certain that the UK is in a position to produce drone squadrons in sufficient numbers for this kind of assault and, even if they are, it’s unlikely they’re prepared to factor in a significant loss-rate as the cost of this type of warfare.
Perhaps what Williamson means is the UK is interested in weaponized, commercial off-the-shelf drones systems, of the kind recently used seen in Caracas, Venezuela, on 4thAugust 2018, where drones were used in an assassination attempt on President Maduro and in Yemen on 19thJanuary 2019 where an explosive-packed drone was deployed against a military parade, killing six. This is much more affordable and ironically is closer to the ‘swarm’ concept as they could be bought in greater numbers as they are cheaper than full aircraft-style UAVs, so losses would be less significant. Many militaries have developed their own small drones, such as the US’s successful, unit-based, hand-launched surveillance drone, the RQ-11 Raven. Perhaps what Williamson is talking about is a specially-designed and developed military micro-drone with weaponized features (e.g. explosives)?
On the other hand, The Guardian article suggests this isn’t so: ‘The MoD indicated the exact design had not yet been tendered for, although insiders said the new drone unit would be used to locate radar and missile systems from countries such as Russia and China and allow British or other aircraft to avoid or destroy them’. This sounds like a non-weaponized drone that merely prepares the way for other airstrikes. It isn’t even clear how a ‘swarm’ would operate in that situation, or why a swarm would be needed if it’s not designed to ‘confuse’ and ‘overwhelm’ defences, as Williamson claimed, but merely to locate and expose them. There’s nothing new about that tactic either: Israel was using drones as decoys for air-defences back in the 1973 Yom Kippur War.
Even if the drones are small, newly-designed, specially-made and weaponized, it still leaves us with the problem of coordination, beyond a certain level of operation. Again, how could hundreds or thousands of these be used effectively? The simple answer is they can’t. Because the real development in ‘swarm robotics’ and ‘swarm UAVs’, though it’s coming, won’t be ready for deployment by ‘the end of the year’. It’s the development of swarm artificial intelligence. As my book Digital Warexplains:
The US military began researching swarms under its Micro Autonomous Systems and technology (MAST) program, running from 2008. One example has been the ‘autonomous quadrotors’ produced by the University of Pennsylvania’s GRASP with Kemel Robotics. These are small, flying robots, able to perform a range of stunts, including flipping, changing direction, a figure-8 formation, and coordinated flight through windows and hoops. Initial versions were ‘dumb’, requiring a nearby sensor array, but by 2016 they were able to use their own sensors an on-board computation for localization, state estimation and path planning.
In March 2013, BAE won a $43m contract to lead the extended MAST program which aims to provide a soldier with a swarm of insect-sized robots that can operate without supervision and provide intelligence about what is around corners or in buildings. The program envisaged a range of robots, from spider-like robots providing initial reconnaissance to dragon-fly type, camera-equipped robots for more information. MAST concluded in August 2017 with 17 live demonstrations by researchers at the Army’s Aberdeen Proving Ground. As Allison Mathis, the program’s deputy manager commented, ‘We have created advances in everything. There are new platforms, new algorithms, new sensors. Not all of this will be ready next year, or even the next five years, but we have absolutely advanced technology. We are making an impact right now’.
When MAST began in 2008 the idea of hand-held micro-drones was science-fiction. By the time it ended, the technology had advanced so far that they had become a commercial reality as a child and hobbyist’s toy flown by millions. MAST had numerous successes, including developing insect-like cyclocopters under 30 grams which could outperform polycopters and were quieter; another was Salto, a 98-gram hopping monopod that can move at 2m a second, can operate in spaces where flying robots couldn’t and travel over terrain a tracked robot couldn’t. It was developed by University of California, Berkeley’s Biomimetic Millisystems Laboratory who are also working on cockroach-like robots able to crawl into small spaces and right themselves if they are tipped over. DARPA are continuing MAST’s work on navigation with the Fast Lightweight Autonomy (FLA) program, which aims to develop small drones that can ‘ingress and egress into buildings and navigate within those buildings at high-speeds’.
MAST was replaced in 2017 by the Distributed and Collaborative Intelligent Systems and Technology (DCIST) program (running till 2022) which is interested in distributed intelligence, heterogeneous group control and adaptive and resilient behaviours. In the final MAST demonstration, GRASP was able to show three autonomous quadrotors coordinating their activities using their own sensors and computers, and the next phase is coordinating heterogeneous elements, including robots of different sizes and other military systems including soldiers. By December 2016 the US Office of Naval Research held a month-long swarm robotics demonstration in the lower Chesapeake Bay, Virginia. Four drone boats with control and navigation software, patrolling an area of 4x4 nautical miles, showed they could cooperate collectively to defend a harbour, collectively deciding which boat would track and trail an intruder vessel. It is believed swarms of drone boats could help prevent attacks such as the suicide bombing of the U.S.S. Cole on 12th October 2000.
To this, we need to add Chinese research. On 20th July 2017, China published its ‘Next Generation Artificial Intelligence Development Plan’ which designated AI as the transformative technology that will secure its future economic and military power. China sees AI as a disruptive technology in war that will enable them to gain an advantage over their rivals, hence their stated aim to have overtaken the US in AI by 2025, whilst ‘by 2030, China’s AI theories, technologies, and applications should achieve world-leading levels, making China the world’s primary AI innovation center’. Simply put, this is a plan for global AI dominance. China are expected to devote huge sums to this AI project, comparable at least to the $150bn they allocated in 2014 to their 10-year national semiconductor plan.
Chinese authorities are prioritising research into swarm robotics, as a key future military technology. On 11th June 2017, the state-owned China Electronics Technology Group demonstrated the largest coordinated micro-drone swarm to date, with 119 drones flying in formation. China sees swarm robotics as a cheaper option, and, in sufficient numbers, as potentially able to defeat more expensive military systems such as those deployed by the US. Swarm drones would also allow China to project force without risking a major confrontation, exploiting the ‘grey zone’ of limited military operations.
It is, of course, possible that Gavin Williamson sees this as the model for the UK to emulate. It’s possible that his description of RAF ‘swarm squadrons’ with hundreds of small network-enabled drones does actually mean hundreds of AI-coordinated and piloted UAVs, rather than individual units remaining under human control. If so then he is indeed at the forefront of military thinking on this issue. But if this is the case, then where is the related discussion of AI and commitment to its funding? And how does he think an advanced swarm-based AI system could be deployed within the year when the AI superpowers, the US and China, don’t seem able yet to militarily deploy such a futuristic system? In a time of continued cuts to a UK military force that appears to be more stretched than ever before, do we really believe an effective AI-drone force will be running by 2020? Given the huge capital investment by the US and Chinese governments in AI, do we really think a £7m injection is sufficient to match them? None of this adds up, making it likely that we really are talking about the quicker, easier, less-advanced option. So, whilst there is a future of ‘clouds’ of weaponized micro and nano-drones, operating with advanced swarm software, confusing and overwhelming any defence system, Gavin Williamson’s statement doesn’t really seem to point in that direction. It seems to have more in common with the failed humanly-operated, numerically-limited drone ‘swarm’ attack Russia fended off in January 2018. If that’s the case, it doesn’t bode well for the UK’s ‘lethality’.