Back to the build log

How we designed a prosthetic arm

The brief

I was accepted onto the Imperial College London STEM Challenge and put in Team F. Each team was invited to tender for a human-centred prosthetic design: a new prosthesis for an adult with a unilateral upper limb difference. Our team had an extra requirement. The arm had to let someone who loves gaming use a standard, unmodified handheld controller.

Constraints:

  • A budget of up to £5,000. That was a ceiling, and we would lose marks for over-engineering.
  • Built for a more economically developed country
  • Offering something no prosthesis on the market has

We were judged on:

  • Application of STEM (35%)
  • Innovative thinking (25%)
  • Requirements met (20%)
  • Teamwork (20%)

Day one (research)

What we did

  • We started with prosthesis anatomy (socket, suspension system, control system, joints and terminal device) and the four main types: passive, body-powered, myoelectric and hybrid.
  • Prithvi, President of Imperial’s Prosthetics Society, gave a talk about the society’s arm, leg and exoskeleton projects. His advice was to design around the user: start with what already exists, ask whether you can make it, then adapt it to the person’s needs.
  • The day was for research only, so nobody started designing yet. We put everything we found into a group A3 research poster.

My research

  • EMG signal conditioning for myoelectric control. I looked at how muscle signals are picked up and turned into movement. MyoWare 2.0 sensors (about £35 each) detect signals in the microvolt range. Before a microcontroller can use them to drive servos, they have to be amplified, band-pass filtered (keeping 20 to 500 Hz, which removes noise and movement artefacts) and rectified.
  • Materials. I compared TPU, PETG, carbon-fibre PETG, silicone, EVA foam and silver-coated nylon on heat tolerance, stiffness, grip, comfort and cost. TPU softens at around 80 °C, carbon-fibre PETG is stiff enough for a structural skeleton, and silver-coated nylon conducts but costs £50 to £100 per metre.
  • Custom sockets. I looked at 3D scanning the residual limb, either with an optical scanner or more cheaply with LiDAR through a phone app, then cleaning up the mesh in SolidWorks or Fusion 360 to make a socket that fits.
  • Practical design rules. Keep the electronics near the top of the forearm so the weight sits close to the body, shield the battery from servo heat and punctures, and lubricate the tendon channels to cut friction.

Early concepts

We came up with two control concepts:

  • Concept A: a 3D-printed frame that snaps onto the controller. Tilting the wrist or rotating the forearm pushes levers onto the buttons. It’s simple and has almost no latency.
  • Concept B: EMG signals mapped to buttons in a switchable mode, so one limb could work half of the controller.

Skills: technical research, reading datasheets, signal processing fundamentals, materials comparison, concept generation.

What I learnt: a prosthesis is a whole system. The control method, the socket and the materials all affect each other, so a small change in one shows up in the others. Getting a clean muscle signal also takes far more processing than I expected.

Day two (design)

What we did

  • Professor Ravi, a professor of Biomechatronics, gave a talk, and Jane, an education consultant, ran a session called “Why Bother?”.
  • We learnt the Double Diamond design process (Discover, Define, Develop, Deliver), and the difference between divergent thinking (generating lots of ideas) and convergent thinking (narrowing down to the best one).
  • We analysed two existing products, the Xbox Adaptive Controller and the Open Bionics Hero Arm. From them we took three requirements: our design should be customisable, lightweight and modular.
  • We moved from concept sketches to 3D CAD models and wrote an A4 specification sheet, with a technical data table covering materials and operating conditions.

Key decisions and why

  • Hybrid myoelectric and body-powered control. Relying entirely on muscle signals would have cost more and drained the battery faster. A hybrid uses less power and still gives motorised precision where it’s needed.
  • Three digits instead of a five-finger hand. We modelled a five-digit anthropomorphic hand and a simpler three-digit terminal device in CAD. Fusing the middle, ring and little fingers into one block, driven by a single high-torque servo, cut the part count by about 33% and gave a much larger contact area for gripping the controller.
  • The three digits:
    • a hinged thumb with a socket for rotary motion, to work the joysticks
    • a hinged index finger with a platinum-cure silicone tip that grips buttons like skin and presses them physically, so there is no software latency
    • a padded third digit with a bistable spring steel strip (the same mechanism as a slap bracelet) that wraps around the controller handle

Skills: design process methodology, user-centred design, CAD modelling, concept evaluation, writing a technical specification.

What I learnt: the most complex design isn’t always the best engineering answer. A copy of a human hand would have looked more impressive, but the simpler design was cheaper, used fewer parts and held a controller better.

Day three (consolidation)

What we did

  • We finalised the design, built our tender presentation within the 11-slide limit, and gave every team member a part to present.

Key decisions and why

  • Component choices and trade-offs:
    • Teensy 4.0 over Arduino Nano: the Teensy has native USB and a 600 MHz processor, while the Nano would need an extra chip to work as a game controller.
    • DS3225 servos (25 kg/cm) over MG996R (10 kg/cm): the cheaper servo has less than half the torque and would strip its gears under load.
    • Carbon-fibre PETG over PLA: PLA is brittle and cracks under tendon tension.
    • Silver-coated nylon dropped: it cost too much for what it added.
  • A costed bill of materials under £500:
    • 2× MyoWare 2.0 EMG sensors
    • 4× DS3225 servos
    • a Teensy 4.0 microcontroller
    • CF-PETG filament
    • TPU and silicone for socket liners and fingertips
    • a thermoformed Aquaplast inner socket
  • FDM 3D printing over injection moulding: about £200 to set up, against roughly £5,000 of tooling that would have used up the whole budget.
  • A quick-disconnect “snap-on/snap-off” collar, so the user can swap terminal devices in seconds without needing a second prosthesis.

Benchmarking analysis

We compared our design with the Open Bionics Hero Pro Arm (548 g, about $5,999), the Joys Pro, the PlayStation Access controller (£79.99) and a body-powered gripper, looking at mechanism, materials, weight, pros and cons. Ours came in at roughly a tenth of the price of the nearest commercial prosthesis, at an estimated 600 g.

Impact and limitations

  • The bistable wrap might not grip the handle tightly enough, because slap bands tend to curl into a circle instead of clamping. The EVA foam pads would also wear over time.
  • From our predictions, we expected to score well on grip actuation and socket comfort, and less well on the Box and Block Test, a standard test of hand dexterity.

Skills: engineering trade-off analysis, component selection, cost estimation, bill of materials, benchmarking, critical evaluation of our own design.

What I learnt: every choice costs something, whether that’s money, weight, power or reliability. Explaining why you rejected an option matters as much as explaining the one you picked.

Day four (pitch)

  • We practised our pitch with our mentor, Fatoumata D.
  • Then we pitched PSthesis5 as a formal tender to a panel of Imperial College London staff, who scored it against the assessment criteria and questioned our design decisions in a live Q&A.

Skills: public speaking, technical communication, defending design decisions under questioning, teamwork.

Skills I developed

  • Engineering: EMG signal conditioning, materials selection, component selection, CAD modelling, design for cost and manufacture
  • Design: Double Diamond process, user-centred design, concept generation and selection, benchmarking
  • Professional: writing a tender, costing a project to a budget, presenting to a panel, working to a deadline in a team of seven

What I’d do differently

  • Use shape-memory alloy or Aquaplast for the third digit so it wraps exactly around the controller handle.
  • Make the fingertips conductive so they work reliably on capacitive touchpads.
  • Build and test a physical prototype, measuring grip strength, comfort and weight instead of relying on predictions.

Thank you

Thank you to Imperial College London, the Imperial Outreach team, our mentor Fatoumata D, Prithvi from the Prosthetics Society, Professor Ravi and Jane for their talks, and my Team F teammates.