You can't just add autonomy

There’s a phrase I now hear more often when discussing future submarines – we’ll just add autonomy. It sounds harmless, logical even, but it makes a misleading assumption shaping how we think about next-generation submarine design. Autonomy isn’t something we can simply add and, if we treat it that way, we risk getting it wrong.

 

Autonomy is already moving beyond isolated subsystems into core control and combat functions, filtering data, generating predictions and influencing decisions. As outlined in my ongoing work (to be presented at the RINA Warship Conference in Bath in June), this fundamentally changes the role of the operator and introduces Human–Autonomy Teaming (HAT) as a primary design driver. It also raises trust and reliability, not as abstract qualities, but as design requirements that must be engineered from the outset.

 

For decades, submarine design has followed a stable model: the platform senses, the crew interprets and the crew decides. Systems support that process, but they don’t challenge it. With HAT, autonomy becomes part of decision-making – control is no longer purely human and, critically, it’s not binary. This introduces a new layer of complexity that must be addressed through system architecture and platform design progressing in harmony.

 

Take the control room as an example. There’s a tendency to equate digitalisation with more screens and more data. But submariners don’t need more information, they need clarity. Trust in autonomous systems comes from transparency of intent, consistency of behaviour and clearly communicated confidence and limitations. If autonomy cannot do this under pressure, it is not adding capability, it is adding risk!

 

In my view, current approaches need refinement. We often design systems and then ask operators to adapt. In a constrained, high-tempo and unforgiving environment, that is not viable. Human factors must be treated as a core design input, ensuring interaction between the operator and the system drives performance, rather than undermining it.

 

Authority management is another area requiring care. In a HAT-enabled system, the question of who is in control becomes fluid. That fluidity must be engineered, not assumed, with clear boundaries, predictable transitions and unambiguous override mechanisms forming part of a reliable and trusted system architecture.

 

HAT cannot be treated as a software or integration problem alone; it is a naval architectural issue. It affects control spaces, system structures, function allocation and the relationship between vessel, crew and the wider operational network. To deliver this effectively, platform design, system architecture and human considerations need to evolve together. Increasing technical capability does not automatically translate into operational effectiveness.

 

Submariners already operate at the limits of human performance. Introducing autonomy without properly integrating it into the human system risks increasing cognitive load, reducing situational awareness and complicating error recovery at critical moments. Reliability, in this context, is not just about system uptime, but about predictable, understandable behaviour in demanding conditions.

 

Autonomy is widely expected to play a role in future submarine design. The key consideration will be how thoughtfully and effectively it is integrated, particularly in a way that fosters trust between human and machine. If we treat autonomy as something that can be added late, we risk building submarines that are technically advanced but operationally brittle. If we recognise HAT as a core design consideration, and ensure architecture, platform and human factors develop concurrently, there is an opportunity to deliver submarines that are more resilient, more usable and ultimately safer.

 

That’s the distinction that matters, because in a submarine, complexity doesn’t fail gracefully.

 

Author profile

James Gladman MRINA, chief engineer, naval architecture and platform design, Expleo UK.

 

This article appeared in Opinion, TNA May-June 2026.