Toward Adaptive Planar Magnetic Levitation Systems for Variable Load Transport: Design and Estimation Strategies

Abstract

Magnetic levitation offers unique advantages for industrial automation—frictionless actuation, silent operation, and fully decoupled multi-degree-of-freedom (DOF) motion—but these benefits are often offset by practical challenges in real-world deployment. In particular, sensitivity due to misalignment in mover and stator magnetic fields, manufacturing tolerances, and external disturbances from variable payloads. These limitations make it difficult to deploy levitation platforms in dynamic environments where the mass and location of transported objects cannot be tightly controlled. Motivated by a growing demand for intelligent transport systems in flexible manufacturing lines and collaborative robotic workcells, this research presents the development of an adaptive control strategy for planar magnetic levitation systems capable of estimating unknown payload mass and center-of-mass location data in real time. Performance in the face of variable payloads maintains stability and preserves control performance across a wide range of operating scenarios.

An adaptive linear quadratic regulator (ALQR) framework was implemented on two planar Halbach-array-based mover units—one with a 6-pole configuration and another with a 5-pole layout. The control architecture incorporates online parameter estimation to recover payload characteristics from force and torque measurements without requiring extensive prior calibration or external measurements of the load. This enables the system to dynamically compensate for mass disturbances and maintain control accuracy across varying operating conditions.

Experimental validation shows that the 5-pole array achieves an average payload mass estimation error of less than 2% of the true value, with the payload’s center of mass localized within ±1.9cm. For the 6-pole mover array, the average mass estimation error is below 4%, and the location estimation error is within ±0.5cm. These results varied across multiple stator locations and mover configurations. A series of experiments—including sequential payload loading, off-center placement, and mid-trajectory dynamic loading—demonstrated the estimator’s robustness and its resilience to transient noise and visual occlusion artifacts. To support consistent estimator performance across the stator, a compensation method was developed to account for geometric deviations in the stator. This included a payload mass offset profile derived through iterative calibration, which enables accurate levitation and estimation throughout the workspace without relying on predefined lookup tables. When tracking motion trajectories, the adaptive controller and mass offset profile significantly reduced steady-state error in the payload mass estimation following load application—from 25% to below 3%—and improved vertical and yaw response times when compared to a fixed-parameter LQR control scheme.

Together, these contributions demonstrate a magnetic levitation platform suitable for collaborative and flexible automation tasks. The system adapts to unknown payloads in real time, estimates key inertial parameters during operation, and maintains performance without requiring redesign or calibration for each payload. These capabilities form a foundation for future research in multi-agent manipulation, intelligent load sharing, and sensorless coordination in levitated transport systems.