Engineers at Rutgers University have developed a prototype smart shoe that uses energy harvested from the wearer’s footsteps to power an onboard AI system, tracking walking patterns with 95.4% accuracy without ever needing a battery recharge.
Rutgers University-New Brunswick biomedical engineers have built a pair of ordinary-looking white athletic shoes designed to tackle a persistent challenge in wearable health tech. Instead of relying on a wall charger or a removable power cell, the prototype generates electricity from the pressure and friction of every footfall. The device counts steps, estimates calories burned, and classifies physical activity while running entirely on energy produced by movement.
How Footstep Energy Powers Onboard Artificial Intelligence
The system operates through an internal device embedded in the sole that produces electricity from physical pressure and friction, a mechanism known as the triboelectric effect. Because irregular electrical bursts cannot be used directly by microelectronics, the researchers designed a power-management circuit that increases usable energy by up to 120 times compared to conventional methods.
That harvested current feeds an accelerometer measuring foot movement along three axes. A tiny onboard processor runs an artificial intelligence algorithm that analyzes those measurements, breaking every 15-second block of movement into one of four categories: slow walking, fast walking, running, or climbing stairs. Results appear on a small screen attached to the shoe.
“When you are walking or running, you automatically have biomechanical energy available, so you can harvest this energy.”
Dr Simiao Niu, assistant professor of biomedical engineering in the Rutgers School of Engineering
Dr Simiao Niu noted that walking provides both the necessary data and the physical power to study it, adding, When you sit down, there is no energy available, but you don’t need gait monitoring.
Fuying Dong, a doctoral student, worked on the study published in Science Advances.
Solving the Energy-Intelligence Bottleneck in Wearable Devices
The design addresses a limitation common to modern health wearables. While smartwatches gather vast amounts of data, running artificial intelligence demands significant electrical power, draining batteries faster as devices grow smarter.
The sensor and AI system together consume just 86 microwatts. The analysis happens directly inside the footwear through an approach called edge AI, meaning raw movement data is never transmitted continuously to phones, computers, or cloud servers.
Fitting the algorithm onto a processor with limited memory required substantial streamlining. The team’s original machine-learning model examined 21 movement characteristics and reached 98.1% accuracy, but it exceeded the processor’s memory capacity. By isolating the variations along the three accelerometer axes that mattered most, the researchers built a slimmed-down model that achieves 95.4% accuracy while running 15 times faster and using one-sixth of the current.
Continuous Gait Monitoring Tracks Movement Disorders in Daily Life
Scientists evaluate gait—encompassing speed, stride, balance, and rhythm—as a primary biological marker for health conditions. Changes in walking patterns offer clues about disease progression, fall risk, and rehabilitation pace after injury.
Clinicians currently assess gait through brief observations in laboratories or clinics, capturing only narrow windows under artificial conditions. Continuous monitoring could eventually assist in tracking patients with Parkinson’s disease, brain injuries, spinal cord trauma, and other movement disorders during daily life.
However, the prototype remains in early development. Researchers built and tested the system using data gathered from just four healthy volunteers aged 23 to 26. The algorithm requires extensive clinical trials involving older adults and patients with specific neurological or physical conditions before it can be deployed to diagnose medical issues or evaluate treatments.