Remote Magnetic Levitation Using Reduced Attitude Control and Parametric Field Models

Neelaksh Singh, Jasan Zughaibi, Denis von Arx, Bradley Nelson, Michael Muehlebach

TL;DR

This paper presents a novel approach to remotely levitate and control centimeter-scale rigid bodies in air using electromagnetic navigation systems (eMNS). By developing a platform-agnostic parametric model and a robust reduced-attitude feedback controller, the researchers demonstrate precise, high-bandwidth control of levitating objects, paving the way for advanced medical applications.

What Problem Does This Paper Address?

Current electromagnetic navigation systems (eMNS) are primarily used for quasi-static manipulation or open-loop control in medical procedures like endovascular interventions. While effective for some tasks, this approach is limited in dynamic scenarios, such as those involving physiological motion. Achieving fast, closed-loop control is crucial for enhancing precision, energy efficiency, and expanding the operational workspace. However, developing such control systems faces challenges: creating accurate models of magnetic fields and their interactions with levitating objects is computationally intensive, and controlling the full six degrees of freedom (DoF) of an object is complicated by inherent physical limitations, such as uncontrollable rotations. This paper tackles these issues by proposing a method for high-bandwidth, feedback-controlled magnetic levitation that is both object- and platform-agnostic.

Methods

The core of this work lies in two key innovations:

  • Parametric Field Model: Instead of relying on computationally expensive simulations or pre-computed lookup tables, the authors developed a compact, analytical parametric model. This model maps coil currents to the forces and torques acting on a levitating object. It uses a point-dipole approximation and is derived from a multipole expansion model (MPEM), which allows for efficient, real-time calculation of the magnetic interaction matrix. Crucially, this model is "levitator- and platform-agnostic," meaning it can be adapted to different eMNS setups and levitating objects by simply updating the object's magnetic dipole parameters.
  • Reduced Attitude Control: Controlling the full six DoF of a rigid body is challenging due to inherent uncontrollability in certain rotations. For objects with a single magnetic dipole, rotations about the dipole axis are uncontrollable. To address this, the paper employs a "reduced attitude" representation that focuses on the controllable subspace of the object's orientation. A nonlinear, time-invariant feedback controller is designed to stabilize this reduced attitude and the object's translational motion. Translational motion is handled by linear quadratic regulators (LQR), while the reduced attitude is controlled using a nonlinear controller that guarantees asymptotic stability. The authors also introduce integral actions to compensate for model inaccuracies and disturbances.

Datasets

Not discussed in the paper. The experiments utilize custom-built levitators and eMNS platforms, with data collected through a Vicon motion-capture system.

Key Insights and Ideas

  • Platform-Agnostic Modeling is Achievable: The paper demonstrates that a parametric model can effectively capture the complex magnetic interactions without requiring system-specific calibration for every new object or eMNS configuration. This significantly lowers the barrier to entry for applying eMNS in new scenarios.
  • Reduced Attitude Control is Sufficient for Robust Levitation: By focusing on the controllable degrees of freedom for attitude, the controller achieves stable levitation and trajectory tracking, even for large spatial rotations, overcoming a fundamental limitation of controlling rigid bodies.
  • High-Bandwidth Control is Critical for Dynamic Performance: The experiments highlight that even small delays in the control loop can significantly degrade performance, emphasizing the importance of fast sensing, computation, and actuation for achieving stable and responsive magnetic levitation.
  • eMNS Can Achieve Macroscopic Dynamic Control: The work pushes the boundaries of eMNS beyond microrobotics and quasi-static applications, showing its potential for controlling centimeter-scale objects with significant dynamic capabilities.

Key Results

  • Successful Levitation of Multiple Objects: The framework was successfully used to levitate three distinct 3D-printed objects with varying masses and inertia properties across two different eMNS platforms (OctoMag and a 13-coil system).
  • Precise Trajectory Tracking: The system demonstrated accurate tracking of both position and orientation setpoints. For example, in a 28-minute figure-eight trajectory experiment, position RMS errors remained below 0.6 mm.
  • Superiority over PID Control: The proposed reduced-attitude controller significantly outperformed a baseline decoupled PID controller, especially during large-angle tracking experiments. The PID controller became unstable at approximately 65° spatial rotations, while the proposed controller remained robust.
  • Robustness to Disturbances: The controller showed robustness to uncompensated 3D cross-coupling and was able to maintain stable levitation even when subjected to a manually induced yaw spin of 5 rotations per second.
  • Low Latency Requirements: Artificial injection of delays showed that even an additional 3-5 ms could significantly degrade control performance, underscoring the need for a low-latency architecture.

Limitations and Future Work

  • Motion Capture Dependency: The current system relies on a motion-capture system for state estimation, which limits the full exploitation of the controller's capabilities for unbounded rotations. Future work aims to extend state estimation to handle full rotations.
  • Hardware Limitations: The effective bandwidth of the coil drivers is a critical factor. While the current setup achieves a respectable bandwidth, advancements in driver technology or coils with reduced inductance are needed for even faster control, particularly for lighter objects.
  • Clinical Translation: The authors suggest exploring whether this levitation approach can be applied to clinically ready, human-scale eMNS systems, such as Navion.
  • Unknown Mass/Magnetization: While the controller demonstrated robustness to unknown mass (up to 16%) and magnetization mismatches (±5%), further investigation into handling larger uncertainties could be beneficial.

Summary updated

Abstract

Electromagnetic navigation systems (eMNS) are increasingly used in minimally invasive procedures such as endovascular interventions and targeted drug delivery due to their ability to generate fast and precise magnetic fields. In this paper, we utilize the OctoMag eMNS to achieve remote levitation and control of a rigid body across large air gaps which showcases the dynamic capabilities of clinical eMNS. A compact parametric analytical model maps coil currents to the forces and torques acting on the levitating object, eliminating the need for computationally expensive simulations or lookup tables and leading to a levitator agnostic modeling approach. Translational motion is stabilized using linear quadratic regulators. A nonlinear time-invariant controller is used to regulate the reduced attitude accounting for the inherent uncontrollability of rotations about the dipole axis and stabilizing the full five degrees of freedom controllable pose subspace. We analyze key design limitations and evaluate the approach through trajectory tracking experiments. This work demonstrates the dynamic capabilities and potential of feedback control in electromagnetic navigation, which is likely to open up new medical applications.
BibTeX
@inproceedings{singh2025remote,
  author = {Neelaksh Singh and Jasan Zughaibi and Denis von Arx and Bradley Nelson and Michael Muehlebach},
  title = {{Remote Magnetic Levitation Using Reduced Attitude Control and Parametric Field Models}},
  booktitle = {Repository},
  year = {2025},
  doi = {10.3929/ethz-c-000791286},
}
Figure