CyberRow: a rowing machine that lets the trunk move
A two-lever rowing machine whose handles travel in a straight line, and a new score for how steady a rower’s trunk is.
People with a spinal cord injury can row. Electrical stimulation (FES) makes the paralysed leg muscles contract, and the rowing strengthens bone and muscle that would otherwise waste away. But the machines are adapted standard rowers: the upper body is strapped to a rigid backrest, the stroke is short, and the trunk — the part that most needs training for everyday life, like moving from a bed to a chair — does almost nothing.
This was my MEng group project at Warwick: build a machine that lets the trunk move, and find a way to measure whether it is working.
The machine
Two levers instead of a cable, one for each hand, so the arms push and pull. A single lever can only swing its handle on an arc, which is nothing like an oar. So each lever drives a Chebyshev lambda linkage, a 19th-century mechanism that turns a swing into a near-straight line. The handle travels flat, like an oar handle on water, and the rower’s trunk is free.
It is full of sensors: load cells in the handles, the seat and the footplate; a motion sensor on the linkage, fused with a Kalman filter to know where every part of the machine is; and from that, the power the rower puts in. A digital twin and a game-style interface got as far as a proof of concept.
My part
I was the product owner: I ran most of the team meetings and kept the mechanical, electronic and analysis sides fitting together. On the technical side, I built everything that turned a trial into numbers.
- A Trunk Stability Score. There was no measure of trunk control for someone moving and holding on at the same time. The score combines how much the trunk sways side to side (from motion capture) with how hard the hands push sideways to steady it (from the load cells), into one number from 0 to 1. Higher is steadier. One number means a patient’s progress can be tracked over months.
- A balance check borrowed from walking robots. The zero moment point is where the forces on the body balance out. If it stays within 10 cm of the centre of the seat, the rower is stable.
- The muscle study. Surface EMG on the biceps, triceps, rear shoulder, back and thigh; a MATLAB pipeline to filter and normalise the signals; and the fiddly part, lining up three sensor systems that each ran at a different speed (200, 50 and 1,000 samples a second).
What came out
On an able-bodied rower, in one trial, the trunk swayed about 1 cm, the stability score was 0.75, and the balance point never left the stable zone. The muscles fired as they should: the thighs in the drive, the biceps on the pull and the triceps half a stroke later on the push. So the push-pull levers do work the arms both ways.
That shows the method works. It does not yet show the machine beats the standard one: we tested one able-bodied person and made no head-to-head comparison. The weights inside the score were set by hand, and the left handle’s load cell was unreliable. The next steps were trials with spinal cord injury patients, adding FES, and a game to make the training something people want to do.
Why it is interesting
It is where the idea behind my later work started: a single number that follows one person over time, so their progress is measured against themselves.
Built with Adam Benaben, Chinmay Bharatia, Fred Carver, Diya Kalyan and Rohan McCall. With thanks to Prof. Michael Chappell, Dr Neil Evans, Prof. Brian Andrews, Dr Panagiotis Tsakonas and Jim Flood.