Foot sensors are essential for legged robots to interact with their environment, and ground reaction force (GRF) measurement is a widely used method for evaluating their dynamics. Traditional 3D force sensors rely on load cells with strain gauges, which are implemented using either three perpendicular sensors or a monolithic structure. While these rigid sensors provide linear responses with minimal impact on positioning, they are challenging to integrate into small-scale robots and cost-effective sensors. Therefore, we propose a novel foot sensor that utilizes a magnetic field with the rubber O-ring as a compliant element to measure a 3D ground reaction force-like profile, providing a cost-effective solution for foot-ground interaction sensing. We validate the reliability of the sensor design through quantitative analysis of the collected data. Finally, we train a CNN model and show that it can robustly predict the inclination of an unseen slope given a short time series of the sensor readings.
Lertvittayavivat et al. (Fri,) studied this question.