Autonomous Warships - how technology will make science fiction a reality

It wasn't long ago that we thought autonomous flight was just for science fiction films. But look at us now, my children have at least three drones between the two of them and my son is already saving for the next generation of Ehang184 (pictured right). We might be some way off allowing self-flying aircraft replace commercial airliners however the fact is autonomous flight for humans is here.

Could we really allow a 10,000+ tonne chunk of metal armed with guns, missiles and other modern combat systems simply roam the open ocean by itself?

But what about Warships? Is the notion that the US Navy wants to start deploying unmanned Warships real? Could we really allow a 10,000+ tonne chunk of metal armed with guns, missiles and other modern combat systems simply roam the open ocean by itself? It's hard to fathom, especially when less complex platforms like the Patriot missile system has failed and killed civilians and friendlies in the past. One thing is for sure, you wouldn't want to let it sail off into the sunset without a few assurances, like bug-free software, loads of testing and a manual override mode.

There are however some obvious advantages of deploying autonomous ships:

1. You remove humans from harm's way
2. You remove the requirement for the ship to have to refill food and water and empty waste. In fact if an autonomous nuclear powered warship was ever built it would only ever need to return to port for non standard maintenance, fluid replacement or decommissioning
3. Platform availability is significantly increased
4. They are cheaper to build and maintain. Without the requirement to accommodate, feed and entertain humans, cost is significantly lower
5. Conventionally powered variants can travel further as you can fit a lot more fuel on-board
6. Mutinies are unlikely to occur

However before you get too excited (or worried), the US Navy's first autonomous ship only weighs 140 tonne (127,000 kg) and is 132 feet (40 metres) long. So we are not talking destroyer size. However the experimental "Sea Hunter" is still incredibly advanced despite its size and technically it is possible to drag and drop its technology into larger vessels. In fact theoretically you could build an autonomous aircraft carrier carrying autonomous military aircraft. You would need to write a little bit more code to manage something of that complexity but its possible and that's all you need to make it real.

Above: US Navy's Sea Hunter experimental autonomous ship

So what are the key technologies making this capability possible? I am going to start answering that question by suggesting there are two ways to achieve the six benefits above:

1. Build a full autonomous warship. Pros: Completely manages itself Cons: Completely manages itself
2. Build a semi-autonomous warship. Pros: Combines human and artificial intelligence Cons: Still requires humans to operate it

Lets break each of those options down to get a better picture of how each would play out.

Fully Autonomous Warships

By definition, a fully autonomous warship would be able to undertake every activity a manned warship can minus anything relating to the management of humans. So let's list ten such activities and then identify what technologies would be required to enable them with no humans on-board.

As you can see the majority of technologies that would enable the operation of an autonomous warship have actually been around for some time. There is however one technology that has been around for decades but has undergone considerable evolution in the last five years, artificial intelligence (AI). There are two reasons for this evolution.

Firstly, as a result of advances in algorithms, AI is in the middle of its progression from Artificial Narrow Intelligence (ANI) to Artificial General Intelligence (AGI). What does this mean? Well in a nutshell it makes the leap from a capability that can only focus on one functional area to one that can focus across multiple areas of the cognitive spectrum. Basically it can multitask, an ability female humans seem to have progressed further than the males of the species. Eventually AGI will transition to Artificial Super Intelligence (ASI) at which point computers will officially have superior intellect than humans. Secondly and the reason AI has progressed even faster than expected, is the evolution of High Performance Computing (HPC). The ability to throw more computing resources at AI has resulted in faster response times which closer mimic the speed at which humans can make decisions. It doesn't matter how brilliant a piece of AI software is, if it doesn't have the computing power to reach a decision in time, it can mean the difference between avoiding an object in the water and certain catastrophe.

So if you were to ask the question, "What is it that has enabled the notion of building autonomous warships?", the answer would invariably come down to "the evolution of AI". Sure, robotics plays a huge part too as without humans there is nobody to fix things. However when you render it down , the "advanced" in "advanced robotics" is AI. So once again we come back to the evolution of AI as being the catalyst for true autonomy. You can see in the above table that AI enables every task except for communication which requires nothing more than a method to send zeros and ones. Another technology that has also progressed in the last decade and worth noting is Electro-Optical. Electro-optical is a branch of electrical engineering, electronic engineering, materials science, and material physics involving components, devices (e.g. Lasers, LEDs, waveguides etc.) and systems which operate by the propagation and interaction of light with various tailored materials. What does this mean in Layman terms? It means it can track objects with remarkable precision and through the use of AI, understand what that object is and how it should interact with it. An example use of electro-optics on a warship is CIWS, short for close-in weapons system. General Dynamic's Phalanx is a CIWS system that uses electro-optics to track inbound missiles and destroy them using its coupled Vulcan Gatling canon.

Issues that need to be overcome before deploying autonomous warships

It goes without saying that to build such a complex platform such as a 10,000 tonne autonomous warship, there are going to be some issues to overcome. For the purposes of this article I am going to stick to technical issues and avoid the ethical, regulatory, legal and other issues. Also, in order to avoid this article being too long, I am going to keep to the top three technical issues that I see would need to be overcome.

1. Securing access to a floating warship. Physical security shouldn't be too hard to address as an autonomous ship doesn't need lots of doors and windows. You can also build in some deterrents that you couldn't otherwise on a manned vessel like electrification, mini-guns and high intensity sound. There are however problems that need to be addressed when it comes to cyber security. Not only would an autonomous warship have a primary and secondary (and probably tertiary) communication line to secure, with the advent of the Internet of Things (IOT), there are also thousands of networked devices on-board to secure. All it would take is for one of those networked devices like an engine temperature sensor to be exploited and someone has access to a cyber attack vector. That vector could then potentially be used to "hop" to other nodes and eventually into the control system where they could disable the manual override and take the ship for a joyride. Could you imagine the headlines "Cyber hacker gains control of warship from his living room"? The scariest prospect of building an autonomous warship with an arsenal of weapons is the possibility of it being hacked. Defining the right security protocols including a potential self-destruct mode would be essential.
2. Power failure. Where humans have the ability to break out some candles and continue on in a power failure, a floating computer has no such luxury. If the power failure is severe enough it may result in the warship colliding with another object. Power redundancy would need to be a big factor in the design of an autonomous warship. There would need to be enough backup power for the ship to shut down certain systems and potentially deploy an anchor especially if close to other vessels.
3. Communication failure. The possibility of losing communication with an autonomous warship is not only a little scary but also quite possible. Understanding how to deal with that situation would be a very important part of the design. It would need to factor in backup communication methods and also protocols for what to do in the event communication is down for extended periods. Basically you would need to make the ship think for itself like an adult.

Semi-Autonomous Warships

There is no doubt, autonomous warships will initially come with the potential of some very serious and embarrassing situations. So is there a stepping stone in between heteronomy and autonomy? Yes there is and if it was ever built it would look a lot like a video game just with some very serious consequences if you got things wrong.

Through the use of a Digital Twin, all the control and sensory of the physical ship would be managed and viewed from the land based control room

What you see in the image above is a simulation "cave" for a commercial cruise-liner. These simulators are nothing new and most large navies use them to train and on-board personnel. What would be new however, is using them to actually remotely operate a real warship. Instead of computer generated images that you see above, high definition cameras could be fitted to the ship relaying life like imagery to a land based control room. Effectively it would mimic being on-board the ship yet without the sea sickness. Through the use of a Digital Twin, all the control and sensory of the physical ship would be managed and viewed from the land based control room. It would be every boys' (and some girls') dream, the ultimate remote controlled boat.

So what technologies would be required for a semi-autonomous warship:

As you can see, the technologies that would be used on both an autonomous and a semi-autonomous warship are similar however there is one noticeable difference. A semi-autonomous warship would require simulation and Digital Twin at its core. A semi-autonomous version would still benefit significantly from AI however default control would be provided by humans in the land based control bridge. AI would take control in the event of communication severance, system failure (on-board or in the control bridge) or when the warship suspects a significant behavioural or protocol anomaly such as cyber attack, firing a missile at a friendly, non-essential self destruct, etc. For both autonomous and semi-autonomous warships, manual override would most likely be built into the design however it would require multi factor enablement.

Issues that need to be overcome before deploying semi-autonomous warships

When it comes to technical issues, a semi-autonomous warship is not that dissimilar to an autonomous one. The biggest challenges are cyber security and what to do in the event of power or communication failure. Communication is key with either variants however probably even more important for a semi-autonomous warship. A semi-autonomous warship would need to rely heavily on communication resilience so would have at least two and most likely three modes; Cellular (GSM, 3G, 4G, 5G, 6G) when in close proximity to the coast, two satellite connections (each paired to a different satellite) and as a last resort HF Radio (now supporting 120 kbps through MIL-STD-188-110C compliant modems). Even with multiple connection options there would often be drops in communication and therefore AI would constantly be in and out of autopilot mode to keep things running smoothly. There is however a potential solution to the communication problem. The land based control bridge could actually be housed in a ship that follows the warship. That way you could keep in communication range however still stay out of harm's way by staying a safe distance behind as seen in the image below.

Using AI in a Digital Shipyard

AI is one of the most powerful enablers of technology shift and drift. Not only is it the key ingredient in autonomous vehicles such as cars, drones and ships but it will also play an important role in the future of manufacturing. Shipbuilding is one such beneficiary as it requires a number of core capabilities that can be significantly enhanced or even completely transformed with AI. Let's take a look at a few examples

1. Driving the Digital Twin. AI is a significant enabler in operating digital twins at scale. Where less complex algorithms can drive digital twins of smaller systems such as turbine engines, digital twins of sizable platforms like a ship require the capabilities of AI and Machine Learning (ML). Coupled with High Performance Computing (HPC), AI can analyse sensory data collected from the physical twin to understand things like potential issues with systems, location of movable equipment, supply levels of food and water and even the heart rate of of those on-board

1. Optimizing production engineering. AI is particularly good at processing large amounts of information in a short period of time to reduce time to build. Below is an example of where AI can be used to optimize the bolt layouts of sheet metal in the manufacturing of ships. This view through an Augmented Reality (AR) headset is the output of AI interacting with the Digital Twin to determine the best place to insert bolts based on the location and density of materials it is about to attach to. The headset user can therefore use a laser pen to mark each spot on the sheet metal for the production engineer to place the bolt. If you do this manually it can take 5 minutes for each bolt however with AI it is complete within less than five minutes for all the bolts.

1. Validating the build against the design. Another useful application of AI in shipbuilding is validating that what has been built is what was was designed. Using an AR headset to capture what is being seen of the physical build, AI can validate that input against the Digital Twin and report back on any anomalies to the user as seen in the view below.

1. Automating Engineering Change. With the cognitive power of AI comes the ability to process the same information that humans can when managing change. When AI is let free to interact with the history of an organisation's shipbuilding data, it puts itself into Machine Learning mode. This becomes exponentially powerful as based on thousands of previous changes it can determine possible solutions for an engineering fault. By inputting the symptoms of an engineering fault i.e. "Too much vibration" or "Excessive heat", AI can go away, crunch some data and come up with possible engineering alternatives, saving hundreds of man hours.

Welcome to the age of Autonomy where AI is King

AI can bring countless use cases to life within manufacturing industries. Its ability to not only expedite product build but also increase its quality is second to none

AI is nothing new but it has evolved at great pace over the last 10 years. So much so that it has opened the door to capabilities that have pushed the envelope on what we thought was going to be possible in the first half of this century. Small autonomous drones are one thing but a self thinking 10,000+ tonne warship is a completely different proposition. AI can't take all the credit though as it takes the advancement of not only software but also the hardware. High performance computing has also come a long way and is AI's best friend when it comes to enabling these truly amazing technologies. However the benefits of AI don't stop at autonomy. AI can bring countless use cases to life within manufacturing industries. Its ability to not only expedite product build but also increase its quality is second to none. In the operate and maintain phases of a warship life-cycle, AI also has the ability to dramatically improve effectivity and efficiency of platform maintenance and therefore availability. When it comes to advancing defence platforms, AI is King and it will continue to expedite the realization of what was once just the subject of science fiction.