CABLE-BOT_
A platform hung on four cables inside a 3 m frame, positioned by winch length alone. We wanted to know how much of a cable robot’s stated workspace is actually usable when the cables have to stay in tension.
- Usable workspace
- 2.0 m
- largest square holding < 5 mm error, from the 40-point map
- Position error
- 3.20 mm
- RMS over the central 32 of 40 grid points, laser line and printed grid
- Payload
- 2.0 kg
- calibrated masses at the worst pose
- Build
- 25 d
- Brief → 40-point error map measured
- △3 · 03-06
- Published the 40-point error map with the 11 mm edge failure left in it.
- △2 · 02-24
- Tension floor added to the planner. Slack events 7 → 0 over 200 moves.
- △1 · 02-17
- Winch drums recut with a helical groove at 1.2× cable diameter. Centre error 9 → 3.2 mm.
BDR-25-01-A — reserved. Four winches at the corners of a 3 m frame. The usable square inside it is 2 m on a side.
- Frame span
- 3000 mm
- Edge error
- 11 mm · RMS over the outer 8 grid points, same measurement pass
- Settle
- 260 ms · encoder, ±5 mm band, median of 200 moves
- Peak acceleration
- 6.40 m/s² · platform IMU, 95th percentile of 200 moves
- Minimum tension
- 18 N · in-line load cell on one cable, worst of 40 poses
- Cables
- 4
- Parts
- 132
What we were asked for.
Cable robots are quoted by their frame size, and a 3 m frame sounds like a 3 m machine. We wanted our own number for how much of that is usable, measured on a grid, before we ever proposed one to anybody.
The build was deliberately cheap: 8 mm plywood gussets, off-the-shelf winch drums, Dyneema. The point was the error map, not the machine.
The targets, and what we actually hit.
| Constraint | Target | Achieved | How measured |
|---|---|---|---|
| Position error in the central 2 m | < 5 mm | 3.2 mm RMS | 32 grid points, laser line against a printed grid |
| Position error at the frame edge | < 5 mm | 11 mm RMSMissed | the outer 8 grid points of the same 40-point pass |
| Cables in tension everywhere in the workspace | > 15 N | 18 N | in-line load cell, worst of 40 poses |
| Payload | 2 kg | 2 kg | calibrated masses held at the worst pose for 60 s |
| One person puts the frame up | < 30 min, no second pair of hands | 25 min | timed, three separate setups, one person each time |
Targets were set at kickoff, before the first part was cut. A target we missed stays on the sheet with the number we actually got.
How it moves.
Four winches at the frame corners, Dyneema to a 320 mm platform, position from drum encoders and a length model. No camera, no external metrology in the loop — the whole question was how far pure kinematics gets you.
The drums are the part we recut. The first set had a plain bore and the cable would lay on top of itself unpredictably, which puts a whole cable diameter of error into the length model at random. Helical grooves at 1.2 times cable diameter fixed it and took position error from about 9 mm to 3.2 mm in the centre.
Measured, and where it fails.
Where it fails
- Anywhere near a frame corner the cable geometry goes nearly parallel and the position error triples. This is geometry, not tuning, and no controller fixes it.
- A four-cable planar arrangement cannot resist a moment about the vertical axis. The platform yaws under any off-centre load and we do not correct it.
- Cable stretch is modelled as a constant. Over a 90-minute run the platform drifts about 4 mm downward and needs re-homing.
What it did.
A 3 m frame gives a 2 m usable square. Inside it, 3.2 mm RMS over 32 grid points; at the edge, 11 mm, against a 5 mm target we missed and published anyway, because the shape of that error is the actual finding.
Minimum cable tension held at 18 N across all 40 poses once the planner enforced a tension floor, which took slack events from 7 in 200 moves to none. The platform carries 2 kg and settles in 260 ms.
We would propose a cable robot for a job that needs coverage over a large area at centimetre accuracy — camera rigs, sensor sweeps, warehouse-scale pick-and-place. We would not propose one where the payload has to arrive at a fixture.
Elapsed days between recorded milestones, drawn to scale from the dates. The overall dimension runs from the brief to the last entry.
What changed since.
| Rev | Date | Change | By |
|---|---|---|---|
| △3 | 2025-03-06 | Published the 40-point error map with the 11 mm edge failure left in it. | DZ |
| △2 | 2025-02-24 | Tension floor added to the planner. Slack events 7 → 0 over 200 moves. | RM |
| △1 | 2025-02-17 | Winch drums recut with a helical groove at 1.2× cable diameter. Centre error 9 → 3.2 mm. | DZ |
What we would do next
Eight cables, not four. A spatial arrangement resists the yaw moment the planar one cannot, at the cost of four more winches and a much harder tension distribution problem.
Put a laser rangefinder or a fiducial camera on the platform and close the loop on measured position rather than modelled cable length. Every error in this record is a length-model error.
Model cable creep as a function of tension and time. A constant is wrong by about 4 mm after ninety minutes and that is the whole drift budget.
Who built it, and when.
- Entry
- R&D
- Status
- ● SHIPPED
- Brief
- 2025-02-03
- Demo
- 2025-02-28
- Elapsed
- 25 d
- Built by
- DZ, RM
- MEC
- Mechanical design
- FAB
- Fabrication and finishing
- ELE
- Electronics and PCB
- FRM
- Firmware
- CTL
- Controls and motion