The knee fails in
forty milliseconds.
Most ACL ruptures in field sport happen without contact — a plant, a cut, a landing. ARES is a compression sleeve worn below the knee that reads limb motion and muscle activity together, and flags the mechanics that precede a rupture while the athlete is still on the pitch.
Pick a movement. Watch the knee.
The figure below is driven by the same biomechanical model our sleeves run today: the shin is measured, and the knee, thigh and ankle are reconstructed from it using the anatomy's own constraints. Every number on the right is computed from that model, not scripted.
The limb reconstruction and activation model are the ones running on our hardware today. The movement inputs and the muscle traces are modelled for this demonstration rather than recorded from an athlete — see where we are today.
Motion tells you the knee is in danger.
Muscle tells you whether it's protected.
Two athletes can cut with an identical knee angle and carry completely different risk. The difference is underneath the skin — whether the hamstrings fire hard enough, and early enough, to resist the forward pull the quadriceps puts on the tibia. One sensor cannot see both.
Motion sensing on each shin
Inward knee collapse, flexion angle, limb speed and impact loading — the geometry of a dangerous position. Multiple sensors run at different sensitivities so a hard heel strike that overwhelms one is still captured cleanly by another.
Electrodes over each muscle belly
Hamstring-to-quadriceps balance and how quickly each muscle switches on — the protection that decides whether a dangerous position actually tears a ligament. Invisible to motion capture, and the reason ARES carries both.
How a step becomes a score.
A real sequence — each stage runs on the athlete, on the sleeve, or on the sideline.
-
01
Capture
Motion sensors and surface electrodes on each limb. The sensors themselves decide when a reading is ready, so timing comes from the hardware rather than from software guesswork — which is what keeps fast movements accurate.
200 readings per second, per limb -
02
Reconstruct
The sleeve turns raw motion into a drift-corrected limb orientation, referenced to the athlete's own standing posture captured when they put it on. From the shin alone, the body's own limits tell us what the knee and hip must be doing.
Full leg position, relative to standing -
03
Predict
An AI model reads the movement and muscle streams together, learning the build-up that leads to a dangerous landing or cut rather than judging each instant on its own. The pattern over time is what gives the warning.
Sequence model over a rolling movement window -
04
Surface
Risk lands on the ATHENA console as a per-athlete index with the limb, the movement and the contributing factor attached — so a strength coach sees which athlete, which leg, and why, in time to act.
Wireless to the sideline · target under 200 ms
An athlete will only wear what they forget they're wearing.
Every design decision is constrained by one rule: it has to survive a match, and no athlete should be able to feel it while playing.
Compression holds the sensors still
Graduated compression presses the electrodes against skin and stops the sleeve migrating during play — where the sensor sits is what the data is worth. A grippy inner cuff keeps it from sliding down.
Tucked behind the calf
A small sealed module drops into a hooded pocket behind the calf — away from impact, out of an opponent's way, and invisible under a sock. It clicks into the sleeve with a single connection, and a light strip shows charge and link at a glance.
On in seconds, boots still on
Athletes wrap and zip the sleeve on without taking their boots off. Left, right, front and the athlete's name are printed inside, so a rushed equipment manager cannot fit it the wrong way round.
One screen the coaching staff actually watches.
ARES sleeves feed ATHENA — a live squad view where every athlete carries a risk index, and any athlete can be opened into an anatomical view showing which muscles are working and which structures are loaded.
Roster and pitch
Every athlete on the park with a live risk index, sorted so the ones that need attention rise to the top. Session modes cover normal training through to rehabilitation loading.
Anatomical deep dive
A 3D model with per-muscle activity and per-ligament load — ACL, MCL, LCL, PCL, patellar tendon and Achilles — so a flag can be traced to the structure carrying it.
Confidence, not just numbers
Dropouts, weak signal and low battery degrade the display honestly rather than quietly showing a stale figure. A metric nobody trusts is a metric nobody uses.
What is built, and what is next.
The vision above is the complete product. This is its honest status — the motion platform is measured and working, and the muscle-sensing and prediction layers are the work in front of us.
| Subsystem | Status | Evidence |
|---|---|---|
| Motion capture and fusion Multi-sensor motion tracking per limb with on-sleeve orientation processing and wireless streaming |
Validated | Benchmarked against a SageMotion research sensor on the same limb: worst-case 4.5% error across all motion axes over a 199-second walking trial, correlation 0.85–0.96. |
| Limb reconstruction Real-time 3D leg model and gait asymmetry metrics |
Working | The model driving the console above, running live from two sleeves — cadence, range of motion and a loading split between legs. |
| Garment Compression sleeve, electrode carrier, sealed electronics module |
In development | Industrial design and construction resolved through to the material layers. Bonding the electrode carrier to compression fabric is the open question being tested. |
| Muscle sensing Surface electrodes over the muscle bellies and the balance metrics they feed |
In development | Electrode placement and carrier are designed and moving into build; muscle signals have not yet been captured through the sleeve itself. This is the next hardware milestone. |
| AI risk model Learns the movement pattern that precedes injury, from combined movement and muscle data |
In development | The approach follows published work on combined movement and muscle injury-risk prediction; training runs on the dataset the sleeves are being built to collect. |
Alzahrani, Aljohany & Alsirhani, “Real-time wearable biomechanics framework for sports injury prevention and rehabilitation optimization”, Scientific Reports (2025). A combined motion and muscle-sensing model across 50 athletes reported 92.3% accuracy, 90.5% recall and AUC 0.93 for injury-risk classification, with feedback in 188 ± 15 ms.
Those figures are the published result of that study, not ARES measurements. They are why we believe the approach works; proving it on our own hardware is the programme ahead.