TECHNOLOGY

Wearables Try to Guess Foot Contact Before Stairs

controlled laboratoryFri Nov 20 2026

A new study asked a simple question. Can wearable sensors guess how a foot will meet a step before it lands? The focus was the shift from flat ground to stairs. That shift can be hard for walking robots and helper devices. Cameras have been used before to identify surfaces while people walk. But guessing the exact foot contact was still not well studied.

The team tested eight people in a lab. Each person wore a small camera on the right lower leg. They also wore pressure-sensing insoles in their shoes. They did about ninety flat-to-stair transitions. Two computer models were trained. One estimated where pressure would sit along the foot from heel to toe. The other estimated when the foot would hit the surface. The models were adjusted for each person. They made updates every 16.67 milliseconds. They looked ahead up to 250 milliseconds before contact. Testing used trials set aside from the same person.

The forecasts were fairly close. With a 150 millisecond warning window, the pressure estimate missed by about 29.42 millimeters on average. The timing estimate missed by about 21.14 milliseconds. At 100 milliseconds, those errors were 26.82 millimeters and 20.08 milliseconds. At 50 milliseconds, they dropped to 23.72 millimeters and 17.73 milliseconds. Faster torso movement and faster toe swing did not clearly change the mistakes. Foot strikes that landed farther forward made pressure guesses less accurate, but timing guesses stayed about the same. A separate speed check showed the image processing and both models could run together at about sixty cycles per second. That worked on a normal laptop and on a small edge computer called Jetson Orin Nano. The findings suggest camera data gathered before contact may help legged machines plan steps more smoothly.

actions