SAFETY.md, Third Eye Kit
Status & disclaimer
Third Eye Kit is a research and engineering project. v1 is a tested prototype; v2 is a design and reference specification that has NOT been clinically validated and is NOT a certified medical device. The powered exoskeleton, the hot/cold-surface warning, and the health features are novel and unproven. Nothing in this repository is medical advice. Any build or human trial must happen under appropriate technical and clinical supervision.
The mobility aid and the user's own hearing remain primary at all times. This system is designed to augment them, never to replace them.
Not a medical or monitoring device
- The health & emergency bracelet stores emergency information and contacts and can alert family, it does not monitor vitals, diagnose, or treat anything.
- The hot/cold-surface warning is a supplementary aid, not a guarantee; it does not remove the need for normal caution.
- Do not make medical claims for these features in any publication.
Core safety behavior
- Warn → gently steer → never lock. The exoskeleton warns on the skin first, then applies only a small, torque-capped nudge the wearer can always walk through. It never blocks or locks a joint.
- Fail transparent. On any fault, power loss, or emergency-stop, the actuators drop to zero torque and freewheel, the failure mode is always "limp," never "trapped."
- Nothing fails silently. Sensor or system faults are announced within ~50 ms with a distinct signature.
- The human always wins. The steer torque is hard-capped below what could destabilize a gait; a hardwired emergency stop cuts actuator power independent of any software.
Contraindications & limits
- Requires individualized clinical assessment before use for people with certain balance, orthopedic, cardiovascular, cognitive, or skin-integrity conditions.
- Not validated for dense/dynamic crowds or fast traffic.
- Indoors, GPS and live city data do not apply; the wearer relies on the sensors, gloves, and haptics.
Staged validation milestone (must be cleared in order before any blind user relies on the powered steer)
| Milestone | Test | Must prove |
|---|---|---|
| G0 | bench / hardware-in-the-loop | every fault path → zero-torque freewheel within budget; torque caps are hard |
| G1 | able-bodied, blindfolded, harnessed | steer is perceivable, overridable, non-destabilizing |
| G2 | supervised low-vision pilot (fall protection + O&M clinician) | acceptable false-positive rate; cues decoded; trust & load acceptable |
| G3 | limited real-world, mobility aid retained | safe on real drop-offs, curbs, glass; no unsafe surprises |
| G4 | broader field | sustained safe use; retention (people keep wearing it) is the real success metric |
Honest note: nobody has safely robot-steered a blind person's gait before. This is novel and unproven. Do not skip the milestone.
Regulatory (orientation, not legal advice)
A powered wearable that physically assists and steers a person is likely regulated as a medical device and a physical assistant robot (jurisdiction-specific, confirm with professionals). Design against ISO 13482, ISO 14971, IEC 60601, IEC 62304, ISO/TS 15066. Comparable exoskeletons (ReWalk, Ekso) reached market via medical-device clearance. There is real product-liability and insurance exposure.
If you publish this
Keep this file and the README disclaimer visible and unedited. Present the project clearly as a research/design project and prototype, not a product for sale or clinical use.