Enabling Autonomous Robotics for Fusion Environments

Introduction

Fusion environments present extreme engineering challenges, including high radiation levels that prevent human access. Inspection, maintenance and intervention must therefore be performed using remote robotic systems.

To operate effectively, robots must perceive, interpret and navigate complex environments with minimal human input. However, existing systems often lack reliable vision capabilities in high-radiation conditions, limiting inspection coverage and increasing reliance on manual control.

PROJECT: Fusion Oriented Robotic Technology (FORT) – Automation Strand

PROGRAM: Fusion Futures Industry Capabilities (FFIC)

CLIENT: UK Atomic Energy Authority (UKAEA)

The Challenge

UK Atomic Energy Authority (UKAEA) launched the Fusion Oriented Robotic Technology (FORT) programme to advance technologies required for future fusion facilities.

Veolia was selected to develop a radiation-tolerant, vision-enabled robotic system capable of operating in fusion-representative environments, addressing the gap in robotic perception and autonomy.

The Solution

Veolia developed an integrated system combining radiation-tolerant cameras, machine vision and robotic automation.

At its core is a stereoscopic vision system using two radiation-tolerant cameras mounted on a robotic manipulator. This system enables:

• Real-time 3D environment reconstruction
• Object and material identification
• Collision detection and avoidance
• Dynamic path planning

The system generates real-time 3D point clouds, allowing robots to understand spatial relationships and safely navigate complex environments with reduced operator intervention.

Veolia has been supporting UKAEA's fusion programmes for over 25 years, delivering advanced remote handling expertise since the JET era.

UKAEA

Technical Highlights

• Radiation-tolerant stereoscopic vision (MAGICS cameras)
• Real-time 3D point cloud generation
• Machine vision and ML-based object identification
• Collision-aware robotic path planning
• Robot-agnostic architecture
• Modular, extensible software stack

System Development

Veolia delivered a fully integrated solution including hardware, software and system architecture.

Key features:

• Modular, scalable design for reuse across applications
• Developed using ROS2, OpenCV, Open3D, PyTorch and NVIDIA tools
• Tested in a fusion-representative mock-up environment simulating real reactor conditions

This ensured tight integration between perception and control systems, improving reliability and reducing integration risk.

Demonstration and Results

The system was demonstrated with UKAEA in March 2026 and focussed on:

• Real-time 3D environment mapping
• Object and material identification
• Dynamic collision-free path planning

Key outcomes

• Proven delivery performance: All milestones delivered within agreed timelines
• High technical quality: Software and documentation met UKAEA expectations
• Strong collaboration: Regular communication and responsive change management
• Improved autonomy and safety: Reduced operator dependency with collision-aware navigation

Impact

This project addresses a critical challenge in fusion: enabling reliable robotic perception in high-radiation environments. Through the FORT programme, Veolia strengthened capabilities in:

• Machine vision and intelligent automation
• Robotics integration
• Radiation-tolerant technologies

These advancements support future deployment across fusion, nuclear and other high-hazard industries.