Abstract
After analyzing dynamic motions performed in nature, it is clear that animals adapt their footing in real time to compensate for disturbances such as slipping or stumbling, by relying on ground reaction force feedback. However, in robotics, this feedback is not commonly utilized due to the costs, as well as the size and mass of available ground reaction force sensors. The aim of this letter is to present a prototype of a novel ground reaction force sensor that makes use of low-cost load cells mounted at fixed angles to each other. The angle of the load cells was chosen to maximize crosstalk between the sensors while minimizing the overall size of the sensor. Due to the crosstalk, it drastically simplified the mechanical design of the sensor. Through the use of neural networks, two-axis ground reaction forces were estimated with a high accuracy of above $98.5\%$ . Future research involves extending the concept to measure three-axis ground reaction forces in real time.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
