Model-Free Magnetic Servoing Control: Leveraging Raw Magnetic Data for Robotic Manipulation

This article introduces a novel, model-free magnetic servoing control approach for robotic manipulation, typically employed for specified 6-DoF pose following or trajectory tracking, using raw magnetic data acquired from a magnetometer array. Conventional closed-loop magnetic servoing control of robot manipulators requires an accurate model that correlates robot motion with real-time magnetic field measurements. However, this modeling is complex due to the high degree of nonlinearity in magnetic field calculations. Moreover, measurement errors or model inaccuracies can adversely affect control outcomes. In this study, we attach two orthogonal magnets to the robot end-effector to facilitate its 6-DoF control. To enhance control stability and convergence rate, we make a Jacobian consistency assumption and implement closed-loop control that incorporates a moving window of historical error and actuation data. Furthermore, an adaptive extended Kalman filter is utilized to dynamically estimate the Jacobian matrix and update the noise covariance matrices. As a result, the magnetic servoing control can be executed without any prior knowledge of the magnetic model. Experiments are finally conducted by tracking specified 6-DoF poses and different trajectories with the robot end-effector. The results validate the stability and efficiency of our proposed method.