IN DETAIL: Anduril – a software company enabled by hardware

Anduril Industries is one of a new crop of cashed-up defence technology companies – most of them in the United States – that is challenging the business models of long-established European and US prime contractors. The company is known for its hardware products but underlying them all is a software platform called the Lattice Operating System (OS). It is this that makes Anduril describe itself as a software company that is enabled by its hardware.

Gregor Ferguson

“A common misconception is that Lattice is just for our stuff,” says Anduril’s co-founder Palmer Luckey. “It’s integrated with over 100 existing US Department of Defense systems. So it’s actually, primarily, a command and control tool and matchmaking tool for things that we don’t even manufacture.”

He founded Anduril Industries in 2017 to radically transform the US’s defence capabilities by fusing Artificial Intelligence – software – and advanced hardware. He has integrated a consumer high-technology business model with Defence mission-driven objectives. The result has been rapid product development and deployment at scale, overturning established business models in the traditional defence industry.

The focus on AI and Lattice-based software is essential in a fast-moving technology battle, says Anduril: exquisite hardware can take a long time and a lot of money to get right and then remains in service for years, or decades. Upgrading that hardware can take just as long a time and just as much money. But software-based systems can be upgraded or changed radically within weeks, even days or hours – and those changes can be applied to every affected system in service, quickly and virtually simultaneously.

So although its web site describes some 19 different products or families of hardware products, Anduril describes itself as a software company that is enabled by hardware: “We have more people working on Lattice than all our hardware systems combined,” comments Palmer Luckey.

Dr Shane Arnott, who started Anduril’s maritime business after 20 years with Boeing, states, “All of our assets talk Lattice, so they can talk to each other.” Arnott’s biggest claim to fame before joining Anduril was leadership of the Boeing MQ-28A Ghost Bat program. This autonomous Collaborative Combat Aircraft (CCA) was launched publicly at the Avalon Airshow in 2019 and is reported to be close to some sort of final decision – probably a production decision. But in the four years since he joined Anduril he has won a $1.7 billion contract from the Australian Department of Defence to manufacture an undisclosed number (‘dozens’ according to Deputy Prime Minister and Defence Minister Richard Marles) of Ghost Shark Extra Large Autonomous Undersea  Vehicles (XL-AUVs) in Sydney.

His particular responsibility now is Anduril’s growing range of relatively cheap maritime systems: Ghost Shark XL-AUVs and Dive-LD submarines, torpedo-like weapons such as Copperhead, and sensor carriers such as Seabed Sentry. And they are all designed to talk and listen to each other and to act autonomously, using Lattice.

Why? Because undersea communication (and therefore real-time Command and Control, or C2) is terrible, he says: “Autonomy, therefore needs to be at a level that it is in no other domain, because you’re losing your primary sensors, your ability to know where you are, your ability to talk to other people, and to have that human supervision that the vast majority of uncrewed systems today have.”

At depths of 5,000 metres and more that isolation is total. But from Anduril’s point of view, the software remains much the same, says Dr Arnott: “It doesn’t matter how big the thing is, whether it’s the Dive-XL, whether it’s the Seabed Sentry, it’s all running Lattice: they all have the same level of smarts, regardless of their size. They’ll have different size batteries, they’ll have different communication mechanisms, but their ability to reason, their ability to get a mission plan, play their role, do it in concert with each other. That’s an incredible advantage for us to be able to do that.

“The maritime environment is where the innovation is happening,” he adds. “We’re solving the networking problem undersea. No one’s really done that (before), and we’ve had success in that.”

Lattice is Anduril’s core product: a sensor-, network- and system-agnostic platform designed to be used by both civilian and military users. It uses open architecture so it can integrate both Anduril’s own software and products and those of third parties. It integrates data from sensors, data feeds and other combat systems and moves everything into a single integration layer.

This is where the magic happens: Artificial Intelligence and Machine Learning (AI/ML) and a variety of processing techniques filter the high-value information and put it out to users in a mesh network.

However, Lattice doesn’t exist as a one-size-fits-all platform, says Allen Ringel, Anduril’s Vice President of Software Platforms.

“You may have different things within your network,” he points out. “It could be a node. It could be a robot. How do they interact with each other, and how do they connect with each other, and how do they use that information?”

Lattice could end up taking data from all types of sensor, combat system or effector, Ringel explains. And it could take data simultaneously from hundreds of them, and that takes a very resilient network. To compound the difficulties, the system might be used where communications, or the way that users create networks, are limited for some reason, and those limitations might be quite different across the same network: part of it may be on a high-bandwidth compute node, for example, while part of it may be out in the field using a low-bandwidth RF combat net radio link.

“We build this mesh network to be intelligent, so that it’s dynamic, configurable,” he says. “We also have things that are built into it so you can prioritize the right things at the right time. So if you are on a low-grade network, it’ll say, Hey, what is the most important thing that we should transport in that scenario? So there is some very good logic in there that will help to make things better.”

Interoperability is an issue as well, he adds: “This is when we have so many systems, so many networks (from different vendors): how do we normalize that, and how do we put it into a translatable layer that says, Hey, we know how to communicate with all these things?”

The traditional data (and even management) system had a centre communicating out to dispersed nodes. If a link broke that node would be completely out of touch because it was only in contact (formally, at least) with the centre. “But what we have with the Lattice data mesh network is that everything is connected to everything else via many different pathways,” says Ringel.

So if one part of the mesh becomes degraded or denied, another part of it can transmit the data that’s required by an edge node – and only the data that it needs. “And that might be targeting information, it might be threat information, it might be the depth of the magazine or whatever. But the intelligent piece knows what every single node needs, and that is the magic, in my humble opinion, of Lattice: that it’s not a dumb system that’s trying to pump everything out to every node,” comments Ringel. “It knows exactly what a particular shooter or sensor or actor needs to have in order to be useful.

“And  it’s happening in real time, because this mesh is being degraded and denied in real time. And then it makes sure that every single node, at some point in time, gets refreshed as to what another part of that mesh has become aware of. So it’s self-healing and it’s self-propagating, and that is the magic of Lattice.”

How does it do that? Ringel won’t say because Lattice is Anduril’s ‘Crown Jewels’ and he doesn’t give anything away.

But that’s what Anduril is becoming more and more known for: enabling warfighters to operate effectively at the edge, as opposed to a traditional, central C2 model. The company’s software and AI/ML is becoming recognised as important as its hardware products.

“We can’t have the requirement for all data to get back to a central node where we work on processing the data and then push it back out again,” says Ringel. “By that time, if we still operate on that model, our assets will be killed. So that’s why this mesh network is absolutely critical to the future of warfare.”

It all makes sense, but every dependency carries with it the seeds of a future vulnerability: what happens if the Lattice system gets disrupted or hacked?

Disruption is relatively easy to handle: a node could be damaged or a link severed, Ringel acknowledges; but there are other links and thanks to the Lattice mesh network architecture that data will find a way around a damaged link. “Every single node is, at some stage in time, updated to reflect the reality of the current circumstances,” Ringel says. “It might not be in real time, but at some point, this will heal itself, and hence the mesh is so important to winning the next war.”

What about a node being hacked or compromised in some way? “One of the things that’s really important is that it’s redundant, secure and decentralized,” explains Ringel. “We want to make sure that these networks are secure.”

It’s a given that that cyber security and encryption has to be at the very core of everything the company does, he adds. It is also a given that Anduril’s systems are extraordinarily difficult to hack.

“If we lose a node – if we lost a Ghost Shark, say – Ghost Shark has systems which close it down, which destroy the data,” says Ringel. “It’s attritable right? It doesn’t matter if it washes up on a beach somewhere. And the Chinese will not gain anything from stealing one of them, because of our very, very smart software capabilities which will destroy that data and make it impossible to unravel.”

Also, most nodes have only what they need to do their jobs: they may not need high definition video, but only the coordinates of a particular target. That would help both the security and the self-healing process because the network isn’t passing unnecessary, and possibly high-bandwidth, information. And that’s the AI piece of Lattice, Ringel explains: giving the various nodes the information they need to do their job and make decisions, but no more than that.

An important point about Lattice, says David Goodrich OAM, Anduril Australia’s CEO, is that it is an open architecture system, so it also avoids the dangers of vendor lock. “Historically, many primes have locked a locked a customer into their ecosystem because they would not share data or provide an ability for a third-party system to integrate into their software platforms, and that’s been a massive frustration to Defence customers,” he points out.

“From the very beginning, we always were open architecture, so we have integrated dozens and dozens and dozens of third-party sensors and systems software platforms.”

“Every dollar that I put into Lattice, making that brain smarter, making it better at interpreting sensors, allowing existing sensors to see further, be more useful, gets multiplied 30 times across all of our products,” says Palmer Luckey. “That work generally applies to all my systems.”

Putting $1 into something like a jet engine would make that engine a better propulsion system in some way, but the benefit would be confined to that engine and possibly the aircraft that use it. Updating the entire engine family could take a long time and won’t be cheap. It’s different with software: everything that uses that particular software benefits, and almost everything can be updated quickly, simultaneously and cheaply.

“Of course, we still build good hardware,” says Luckey, “but it wouldn’t be able to do nearly what it does if it wasn’t for Lattice.”