Blighty BoffinsWe cog your gears of innovation

University of Bristol

Dr Sophie Chen

Builds robots that can be trusted near people and parts. Inspection, assembly, and the gentle art of not dropping a £40,000 optic.

Sophie works on compliant manipulation and inspection robotics for factories and laboratories. She has a particular interest in tasks that are dull, dangerous, or both.

She collaborates with manufacturers who have a bottleneck and a health-and-safety officer.

Compliance is the technical thread. A stiff robot in an uncertain environment either misses or breaks something; a compliant one can tolerate the several millimetres of positional uncertainty that any real fixture, part or human colleague introduces. This makes the control problem harder and the deployment problem very much easier.

She takes safety assessment as part of the engineering rather than a subsequent formality, and has learned to involve the safety officer in the second week rather than the second year. Teams that do this deploy. Teams that do not build excellent cells that remain behind a fence.

Selected projects

Named work, with a sentence of context each.

  • Confined-space inspection platform

    Developed a small compliant platform for inspecting structures with access too narrow for a person, including the recovery behaviour required for when it becomes stuck.

  • Optical component handling cell

    Automated the handling of high-value optics, with force-limited grasping and a deliberately conservative failure mode. No optics were lost, which was the specification.

  • Laboratory sample automation

    Automated a repetitive sample preparation step in a research laboratory, freeing roughly a day a week of technician time and improving consistency measurably.

Working with Chen

She will ask to see the task performed by a person, several times, preferably by different people. The variation observed in that half hour usually determines the entire system architecture.

Her projects include a private failure phase — a period in which the robot is permitted to be bad at the task, off the production line, without anyone drawing conclusions. She regards a partner's willingness to allow this as the single best predictor of success.

Looking for

Repetitive precision tasks and partners who will let a robot fail in private first.

Probably not for

Requests for a general-purpose humanoid, which she will decline with good humour.

Credits and contributions

  • Principal investigator on three industrially co-funded robotics projects
  • Named on two patents covering compliant end-effector design
  • Adviser on robot safety assessment for a manufacturing partner