top of page

Robots & Autonomous Systems

Research theme 05 · Led by Prof. Yiannis Papadopoulos

In short

Vehicles, vessels and aircraft that make their own decisions, often working with others that weren’t known at design time. This is where the group’s methods meet physical consequences, and where assurance has to continue while the system runs.

Why it matters

Conventional certification rests on an assumption that autonomous systems break: that the operating conditions can be listed in advance. A crewed vessel has a crew to handle the situation nobody wrote down. An uncrewed one has only what its designers anticipated, and the open sea doesn’t stick to that list.

 

Cooperation makes it harder still. When several autonomous units share a mission, each one is part of the others’ environment. The combination may never have been analysed, members can join or leave mid-mission, and behaviour comes from the interaction rather than from any single design.

 

The group’s answer is to move part of the safety argument into the running system. That’s the idea behind EDDIs: make selected safety knowledge executable, so a system can check its own condition against it while it operates.

Theme lead: Prof. Yiannis Papadopoulos

Professor of Computer Science · Lead, DEIS

 

Leads the group and this theme, working on dependability, system safety and runtime assurance for autonomous systems.

 

View profile →

Autonomous vehicles

Safety analysis and assurance for road vehicles whose perception and decision-making are learned, including how a safety case survives an over-the-air update to the model.

Maritime autonomy

Uncrewed and reduced-crew vessels, where long missions, little chance to intervene and a harsh environment put most of the burden on autonomous decision-making.

UAV, ASV and AUV systems

Aerial, surface and underwater platforms working alone and in groups, including inspection and survey work where the platform is out of contact for long periods.

Runtime assurance

Monitoring and safety arguments that continue while the system operates — EDDIs, health and condition monitoring — and the decision to slow down or stop when conditions leave the assured envelope.

Open questions

• How to assure a cooperating group of systems that didn’t exist when any of its members were certified.

• What a regulator should require of a safety argument made at runtime, given that no such framework is agreed yet.

• What a vessel should do when it detects it has left its assured envelope but stopping is itself unsafe.

 

Taking a platform autonomous? We work with vehicle, maritime and aerospace partners on assurance for systems that operate without a person in the loop.

 

Talk to this theme →

​

Related themes:

01 · Safety of AI

03 · Security and Safety-Security

04 · Model-Based Safety & Dependability

​

← All research themes

bottom of page