Development of an optical technology-based sensor and deep learning models to estimate user-interaction forces in smart walkers

In response to the rising prevalence of mobility-related pathologies and the need to develop technologies capable of constantly monitoring the physical interaction with the user, this study introduces an innovative sensor for robotic walkers. Integrating visible light-sensitive photodiodes and addressable RGB LEDs, the sensor utilizes overlapping signals from embedded emitters and receivers in a transparent flexible waveguide layer to measure tactile deformation. This novel approach overcomes limitations of traditional sensors (i.e. strain gauges, resistive or piezoelectric sensors and, high-resolution triaxial cells) such as fragility and high costs, offering a more accessible alternative. Additionally, deep learning methods demonstrated accurate estimation of the applied normal force with an average error of 2.2%, while the random forest model achieved a classification accuracy of 98% for the contact zones. Designed for seamless integration, the sensor promises to enhance functionality and reduce costs in robotic assistive devices, opening new possibilities in mobility care.