Gaussian Process Regression-Based Wheel-Legged Robot Contact Force Estimation With PVDF Piezoelectric Film sensors

Wheel-legged robots combine the advantages of both wheeled and legged locomotion, enabling them to maintain balance, navigate, and perform stable tasks across complex terrains. This paper proposes an innovative method for contact force estimation in wheel-legged robots, utilizing PVDF (Polyvinylidene Fluoride) piezoelectric film sensors and Gaussian Process Regression (GPR). The PVDF sensors are embedded in the inner surface of the foot-end tire, where they effectively detect dynamic changes in pressure and strain. The GPR model is employed to capture the complex nonlinear relationship between sensor readings and the contact forces along the three axes (Fx, Fy, Fz). The experimental setup includes a custom testing platform with a six-axis force sensor to measure actual contact forces, while the PVDF sensors provide force estimates. Results show that the GPR-based model offers high accuracy in estimating contact forces, with normalized root mean square errors (NRMSE) of 11%, 15%, and 16% for the Fx, Fy, and Fz axes, respectively. This sensor-based approach is lightweight, easy to install, and highly sensitive, making it a promising solution for contact detection in complex robotic systems and terrain classification tasks.