Skip to content
DIS Robotics
ROBOTICS · ENGINEERING · ORIGINAL SOURCES
WORLD B-1 · TECHNICAL BRIEF
Technical BriefField Trial4 min read

A robot dog went into Sellafield. The interesting part was the swab.

RADAR86

A controlled trial with blue chalk reveals the work between sending a robot into a restricted area and trusting the sample it brings back.

Published
Updated
Primary Domain
Nuclear Robotics
Deployment Maturity
Field Trial
Sources
3 listed · 3 live links
Reading Time
4 min read
TeleoperationManipulationForce Control

DIS points readers toward original work. Source facts, DIS interpretation, and the editorial verdict are kept distinct below.

What Changed

Source-grounded summary

The most revealing detail in Sellafield's February 2026 robot-swabbing announcement is blue chalk. During a controlled deployment in a restricted area, Spot used a purpose-built attachment to sample a mock-contaminated surface. The chalk represented contamination. The site was real; the spill was simulated. Keeping those facts together gives the trial its proper scale. [1]

The tool came from the Robotics & AI Collaboration, or RAICo. Its shape, a Haply Robotics haptic controller and custom software translated an operator's hand movements into commands for the robot's manipulator. The task was to reproduce the dexterous motion involved in collecting a surface swab, then bring back a sample. [1] This is a case study of that documented trial, rather than a claim that robotic swabbing has since become routine across the nuclear estate.

Why It Matters

DIS interpretation

Sending a camera somewhere difficult is useful. Making controlled contact there introduces a second engineering problem: the robot changes the thing it is trying to measure. A swab has to touch the intended patch, move in the intended way and remain identifiable afterward. Reach alone cannot establish that the collected material represents that patch.

The sampling tool therefore deserves as much attention as the legs. Locomotion gets the instrument into the room; the contact interaction determines whether the trip produces useful evidence. The robot has effectively been promoted from mobile photographer to laboratory assistant. The promotion comes with paperwork, and the paperwork is part of the machine.

How It Works

DIS interpretation

Read the system as a chain: operator intent, manipulator motion, surface contact, recovered sample and interpretation. The announcement establishes the first controlled demonstration. DIS's analysis below identifies the interfaces that would need scrutiny before extending it to repeated operational sampling.

  1. Move the operator, not the exposure

    Sellafield separately reported off-site operation of Spot from Westlakes Science Park in March 2025. That milestone concerned secure remote operation; it should not be silently combined with the 2026 swabbing trial to imply that every part of the later task was operated from the same location. [2]

    The general opportunity is clear: put the operator where the work can be supervised without requiring them to stand beside the sample. The engineering question then shifts toward what the interface lets that person see, command and verify.

  2. Give the hand movement a physical destination

    The 2026 account identifies a shaped attachment, a haptic controller and custom motion-translation software. [1] Those are three different parts of the interaction. A controller can express intent, software can map it into motion, and the attachment can establish contact with the surface.

    We would want to know how contact force and motion were constrained, how the tool behaves on an uneven surface and what happens when the operator reaches an awkward orientation. The release does not provide a force-control law or a numerical sampling-repeatability result. Those remain questions, not hidden specifications.

  3. Treat the sample as the deliverable

    An image of a tool touching a wall cannot answer whether a sample is representative. A useful validation program would compare recovered material under defined contact paths, surface conditions and handling procedures. It would also test blank samples and carryover between tasks. These are proposed evaluation questions, not tests DIS observed.

    Blue chalk is valuable because it gives the team a controllable stand-in while exercising the task. It is not a tiny auditioning radionuclide. Its behavior cannot, by itself, establish performance for every real contaminant or surface.

The Mechanical Lens

DIS interpretation · source-supported dimensions only

Signature DIS section — only engineering dimensions available in the underlying material are shown.

Mechanism

The reported system combines a quadruped, manipulator and dedicated swabbing attachment. Tool geometry and accessible contact orientations deserve separate evaluation from walking capability. [1]

Sensors

Sellafield describes broader inspection payloads, including imaging, 3D scanning and radiological characterization. That capability list is not a specification of the exact swabbing configuration. [3]

Environmental constraints

Contact consistency, access, sample identification and contamination control are consequential system questions. The public trial report does not quantify all of them.

Deployment

DIS interpretation

  • Documented controlled swabbing demonstration at Sellafield, reported 3 February 2026. [1]
  • Potential support for health-physics tasks; wider routine deployment requires further operational evidence.

The Catch

DIS interpretation

LIMITATIONS & OPEN QUESTIONS

The controlled deployment used a simulated spill and mock-contaminated surface. A successful chalk collection does not establish equivalence to every operational radiological sampling procedure. The public announcement gives no sample-size distribution, contamination-recovery efficiency, cross-contamination measurement or full comparison against manual sampling. [1]

Sellafield's wider January 2026 account describes a progression from trials toward routine operations. It is the operator's account of its own program, not an independent audit. [3] The reasonable conclusion is that the integration has demonstrated a useful task under stated conditions. Quantified repeatability and representative sampling across operational conditions would make the next conclusion stronger.

HIT THE QUESTION BLOCK BEFORE YOU BELIEVE THE HYPE.

DIS Verdict

Editorial verdict — not source fact

DIS EDITORIAL JUDGMENT

The important advance is the attempt to close the loop between mobility, human dexterity and a physical sample. The trial deserves attention because it tackles a specific job that someone already needs done. The next milestone worth watching is evidence about the sample itself: what was recovered, how consistently, and with what contamination controls. That is where a memorable demonstration becomes a dependable instrument.

Sources

Original material

Follow the original material to check methods, test conditions, and limitations. Source claims are distinct from DIS interpretation.

Primary attribution label: Sellafield Ltd / Nuclear Decommissioning Authority