End-effector contact force estimation for the industrial robot in automated fiber placement processes with dynamic end-load variations

Abstract

Contact force estimation makes the interaction between the robot and the environment perceptible and controllable without the need for additional sensor devices. For robots used in Automated Fiber Placement (AFP), contact force estimation is primarily aimed at grasping the magnitude and real-time variations of the compaction roller force, which is one of the most critical process parameters in AFP, directly related to layup accuracy and quality. However, contact force estimation during AFP faces challenges: the consumable carbon fiber prepreg material is mounted on the end-effector, leading to continuous changes in load dynamics parameters. Ignoring this situation renders contact force estimation unreliable, making real-time end-load calibration necessary. This paper first establishes a linear dynamic model of the robot and obtains the Variable Parameter Set (VPS) related to end-load. Then, combining distance sensors and weighted moving average, the volume changes in material rollers are measured to perform fast load calibration of the end-effector and update the VPS. Subsequently, a joint hybrid friction model is established and calibrated. Based on this, a Disturbance Kalman Filter (DKF) observer is established using inverse dynamics and discrete Kalman filter algorithms for real-time estimation of contact force. The calibration routine of the DKF observer considers the noise in measured motor currents and errors in velocity data from individual joints. Vertical lifting experiments and actual AFP experiments demonstrate that fast load calibration and DKF observer can provide robust and accurate estimation of end-effector contact force. Compared to force sensors and the DKF observer without FLC, the RMSE is reduced by over 49 %.