Modeling of flange-mounted force sensor frequency response function for inverse filtering of forces in robotic milling

Abstract

Force/torque sensors have great potential for use in industrial robotic applications. However, the Frequency Response Function (FRF) of the robot flange-mounted sensor is influenced by the dynamics of the robot arm, which distorts the force sensor measurements. The sensor FRF is required for developing an inverse filter to compensate for the distortion. However, the sensor FRF changes as a function of the robot end effector position, and therefore must be accurately modeled. This paper proposes a Gaussian Process Regression (GPR) model-based approach to predict the FRF of a force sensor mounted on the end effector of a six degree-of-freedom (6-dof) industrial robot. The force sensor FRF as a function of the robot position in the workspace is characterized using impact hammer experiments. GPR models that predict the magnitude of a specific frequency bin are developed using the impact hammer results. The GPR modeling approach is evaluated using cross evaluation metrics. The predicted sensor FRFs are used for inverse filtering the milling forces measured by the flange force sensor during robotic milling tests. The filtered forces are shown to be in good agreement with forces measured by a ground-mounted quartz-based piezoelectric force dynamometer; specifically, the inverse filter lowers the maximum peak-to-peak error in the force measured by the flange sensor from 146 N to 25.4 N. Hence, the proposed methodology is shown to be valid for improving the accuracy of robot end effector-mounted force/torque sensors.