Navies around the world are buying autonomous vessels faster than they can define what they want them to do. Walk the floor of any major defence exhibition and autonomous naval systems are everywhere, promising endurance, independence and multi-mission capability. Ask how navies actually plan to use them, and the answers are often less impressive.
Conversations on defence trade show floors can sometimes feel like the film ‘Groundhog day’. Take autonomous naval systems, for instance. In recent years, they have become ubiquitous, with maritime UxV (unmanned surface and underwater vehicles) manufacturers promising autonomy and endurance across multiple missions and environments. Ask a few basic questions, however – who controls it, how long it can truly operate independently, how navies plan to use and integrate it – and the notion of autonomy starts to look like the emperor’s new clothes.
The issue is not the technology itself. Industry has been experimenting with autonomy for years. The problem is that while navies articulate operational needs for UxVs, they often lack the concepts of operations (CONOPS) required to make them meaningful. As Julien Lalanne de St Quentin, Business Development Director of Naval Defence at Exail, told Warsight, CONOPS are what allows autonomy “to produce a credible effect.”
However, navies are procuring autonomous systems faster than they have developed the frameworks to define, certify, integrate and sustain them.
The War in Ukraine has played a central role in the exponential growth of the maritime UxV market. Images of the damage inflicted on Russia’s Black Sea Fleet, amplified by extensive coverage of how Ukrainian forces used these systems to push Russian ships back to port, have led many to ascribe near-unlimited potential to UxVs. The speed at which Ukrainian forces developed and adapted these systems has also led many navies to believe that this model can be replicated.
But the reality of the Ukrainian war differs, at times significantly so, from that of many other navies and nations. Ukrainian forces used these systems for highly specific missions in a confined theatre, where autonomy was not always the priority: many were remotely piloted and, particularly early on, not designed to return.
The risk is that these lessons are now being overgeneralised. What proved effective in a constrained, high-intensity conflict is shaping expectations for far more complex missions, without a corresponding evolution in how these systems are defined, tested, or integrated.
What is emerging on trade show floors reflects this shift. UxVs are being developed for a broader range of missions, from intelligence, surveillance, and reconnaissance (ISR) and mine countermeasures (MCM) to payload delivery, with far greater expectations of autonomy and endurance in complex environments. This shift brings added technical demands, notably the integration of multiple sensors for navigation and data collection, and the need to connect these systems effectively to the wider fleet.
Yet the frameworks required to make that complexity operational are lagging behind.
As Chirine Riaz argued in ‘Naval Drones: Between Expectations and Uncertainty’, published by the Fondation méditerranéenne d’études stratégiques (FMES) in December 2024, technology alone does not win wars. UxVs must be integrated “over the long term, across doctrinal, material, and operational dimensions.”
In practice, Lalanne de St Quentin told Warsight, this means that before procuring ‘autonomous’ systems, navies must first define what they expect autonomy to deliver. This conversation should determine the mission, the operating environment, the level of supervision – or autonomy – and how the system fits within the broader architecture.
As Jean François Pelliard and Chirine Riaz wrote in ‘Integrating Drones into Combat Navies: Challenges Ahead’ (FMES, May 2025), it also forces another fundamental clarification: what trade-offs navies are willing to make between accelerated acquisition and deployment – promising “immediate potential operational superiority” – and the longer-term approaches needed to define requirements and mitigate the most critical risks.
That tension is perhaps most visible in the still-common assumption that autonomous platforms can seamlessly perform multiple missions. Technology may allow payloads to be adapted quickly, but the trade-offs do not disappear: surveillance missions require discretion and endurance; strike missions demand speed and agility. A single platform cannot effectively serve both without compromises that degrade performance – a point Lalanne de St Quentin also made.
As Pelliard and Riaz put it, “a balance must therefore be struck between the long term and the short term.” The long term secures development, builds the necessary resources, and accounts for relatively fixed operating environments. The short term enables rapid scaling and sustained operational advantage by continuously integrating new technologies and enabling tactical innovation.
Short-term, rapid scaling requires flexibility and adaptation, two qualities that can often be achieved by leveraging technologies already proven in other domains. The commercial shipping sector offers a useful example. While large-scale adoption of autonomous vessels remains recent – Norway’s Yara Birkeland entered service in 2022 – the underlying technologies have been developed and refined over more than a decade.
In that domain, autonomy has not been driven by platforms, but by clearly-defined operational problems. As David Young, founder of Theyr, explained, factors such as fuel costs, weather routing, port scheduling and emissions compliance “have forced the industry to develop a granular understanding of what it means to operate a vessel with minimal human intervention.”
This has led to the development of advanced optimisation technologies. Theyr’s approach, for instance, is based on multi-objective optimisation: rather than optimising a route based on a single parameter such as fuel consumption, its T.VOS technology leverages artificial intelligence (AI) to process large volumes of meteorological, ocean and vessel data and simultaneously balance multiple, interdependent variables – time, energy, reliability, environmental conditions.
Crucially, however, this type of optimisation only works because the problem is well defined. The system must know what it is optimising for, under what constraints, and in what environment. In commercial shipping, those parameters are relatively stable. In defence, they are often not.
Transposing these technologies into naval programmes therefore raises a fundamental challenge. As Lalanne de St Quentin told Warsight, autonomous systems are expected to operate across a wide range of environments, some of them highly constrained – including high latitudes, extreme temperatures, strong currents, and severe sea states – each of which fundamentally alters the optimisation problem. Without clearly defined CONOPS, these variables cannot be meaningfully prioritised or even properly modelled.
The technologies enabling rapid adoption already exist, but fielding them rapidly is only the first part of the problem. The harder question is how to integrate, validate and sustain them across the full range of conditions navies operate in.
If the short-term challenge lies in rapidly fielding autonomous systems, the longer-term question is how to integrate them into complex naval architectures and demanding operating environments. It requires not only technological maturity, but the ability to test, validate and sustain these systems across a wide range of conditions.
It is in this context that exercises have become a central tool.
In recent years, several nations have developed national and bilateral exercises centred on UxVs. Initiatives such as Task Force 59 in the US (launched in 2021) or Project JUMPSTART in Australia (2024) were initially designed to assess whether these systems could operate effectively, and under what conditions. As the War in Ukraine underscored their operational relevance, the focus has shifted toward understanding how they can be integrated into naval fleets. REPMUS – launched in 2010 as a joint initiative between the University of Porto and the Portuguese Navy, and a NATO exercise since 2014 – is now the largest international exercise dedicated to testing UxVs.
These exercises are essential. They provide a rare opportunity to test systems across missions and environments, observe their behaviour, and identify areas for improvement. But they also have clear limits. By nature, they take place at fixed times and locations, and cannot replicate the full range of operating conditions on demand. Weather, currents and sea states remain largely outside planners’ control.
As Pelliard and Riaz note, such exercises demonstrate that even modest capabilities can deliver meaningful tactical gains. “But outside of operational urgency,” they argued, “there remains a programmatic gap when it comes to integrating these systems beyond the duration of an exercise and deploying them at scale.”
The US DARPA NOMARS programme – aimed at demonstrating a fully uncrewed surface vessel – offers a rare illustration of what filling that gap requires. Theyr spent more than two and a half years adapting its T.VOS optimisation technology to NOMARS’s evolving demands, in a process that was neither linear nor fixed: objectives were defined from the outset but allowed to evolve through real-world development. The result is instructive precisely because it is unusual for operational concept and technology to mature together. Replicating that model at scale is in fact likely to run directly into a fragmented regulatory environment governing how autonomous systems are tested and certified.
As Lalanne de St Quentin explained, approaches to testing autonomy at sea remain fragmented. In France, civilian frameworks tend to certify platforms against technical requirements – architecture, redundancy, communications, sensors – largely independent of how they are used. In practice, this means certifying an operational envelope defined by mission type, environment, level of supervision and degraded modes.
Elsewhere, including in the United Kingdom, systems are authorised case by case, with industry required to demonstrate the safety and relevance of a specific use scenario. The result is a patchwork of regulatory models that complicates both testing and deployment.
This is where simulation is beginning to play a critical role. BMT’s ROC-Sim, developed at the request of the Royal Navy and DSTL, aims to bridge this gap. As Jesse Loynes, Maritime Autonomous Systems Consultant at BMT, and Will Alexander, Global Business Development Lead – Maritime Autonomous Systems at BMT, told Warsight at Oceanology International, the system enables operators to assess “when systems are safe to perform and when they aren’t,” while also allowing mission rehearsal across a wide range of conditions.
Crucially, Loynes and Alexander continued, “it makes it possible to test scenarios that would be impractical [or prohibitively expensive] to replicate at sea, from degraded communications to severe weather and high sea states.” But beyond testing, simulation also enables something more fundamental: the ability to interrogate assumptions, refine concepts of operations, and explore how systems behave within realistic command-and-control structures and human–machine teams.
Unlike live exercises, which remain episodic and environment-dependent, simulation provides a controlled and repeatable environment in which systems, operators and CONOPS can be developed iteratively. In doing so, it generates the kind of traceable evidence increasingly required to support certification and regulatory approval.
The rush to deploy autonomous systems has created a structural paradox: platforms are preceding the very concepts that justify them. As Julien Lalanne de St Quentin noted, in an ideal world, the Concept of Operations (CONOPS) should shape a system’s architecture from the start. In practice, the reverse is common: decisions accumulate, and the operational concept arrives later, as a rationalisation rather than a foundation.
The result is a structural mismatch. Many programmes are developing autonomous systems without clearly defined mission sets, making it difficult to determine what capability is actually required. As David Young warned, when a platform precedes its operational concept, “that is how you end up with capability without concept.”
All told, autonomy is no longer a technological problem – it is an architectural one.
