Electromagnetic Formation Flight

• Electromagnetic Formation Flight (EMFF) is a technique that uses controlled, time-varying magnetic dipoles on satellites to achieve propellant-free formation control.
• Decentralized control with oscillatory coil currents enables both translational and attitude maneuvers in nonholonomic satellite swarms.
• Experimental and simulation studies validate EMFF’s precision, demonstrating sub-centimeter to sub-meter accuracy and robust performance under state and input constraints.

Electromagnetic Formation Flight (EMFF) is a technology for controlling the relative positions and orientations of multiple satellites using only electromagnetic forces and torques. Each spacecraft is equipped with actively controlled coils that generate time-varying magnetic dipole moments, enabling mutual dipole–dipole force and torque interactions to perform formation maintenance and reconfiguration. Unlike thruster-based systems, EMFF provides fully propellantless actuation, and—if properly realized—enables both translation and attitude control within a satellite swarm or formation. This paradigm exploits high-bandwidth electromagnetic actuation and the internal momentum-coupling properties of closed multi-rigid-body systems, and is supported by recent developments in nonlinear control, decentralized architectures, and experimental validation.

1. Electromagnetic Force and System Modeling

EMFF relies on the down-to-the-coil modeling of each satellite as a rigid body carrying three orthogonal coils, which can be driven to produce any desired magnetic dipole moment vector $\mu_j(t)\in\mathbb{R}^3$. In the intersatellite context, the fundamental interaction is the magnetic dipole–dipole force and torque:

$$F_j = \nabla(\mu_j \cdot B),\qquad\tau_j = \mu_j \times B$$

where $B$ is the field at $j$’s location. For two satellites $i$ and $j$ separated by $r_{ij}$, the mutual force is

$$f(\mu_i, \mu_j, r_{ij}) = \frac{3\mu_0}{4\pi} \left[ \frac{\mu_i\cdot\mu_j}{\|r_{ij}\|^5}\,r_{ij} +\frac{\mu_i\cdot r_{ij}}{\|r_{ij}\|^5}\,\mu_j +\frac{\mu_j\cdot r_{ij}}{\|r_{ij}\|^5}\,\mu_i -5\frac{(\mu_i\cdot r_{ij})(\mu_j\cdot r_{ij})}{\|r_{ij}\|^7}r_{ij} \right]$$

and the corresponding torque is

$$\tau(\mu_i, \mu_j, r_{ij}) = \frac{\mu_0}{4\pi} \left[ \mu_j \times \left( \frac{3 r_{ij} (\mu_i\cdot r_{ij})}{\|r_{ij}\|^5} - \frac{\mu_i}{\|r_{ij}\|^3} \right) \right]$$

These expressions define the closed dynamical system for the full satellite swarm, with state vector including the positions, velocities, attitudes, and angular velocities of all members (Takahashi et al., 6 Jan 2026).

2. Nonholonomic Dynamics and Controllability

A satellite swarm actuated exclusively via EMFF is a closed mechanical system: all dipole–dipole interactions are internal. Consequently, both the total linear and angular momentum of the swarm are strictly conserved, imposing a nonholonomic velocity constraint. The total angular momentum is

$$L_{\rm tot} = \sum_{j=1}^n \left[ m_j (r_j - r_1) \times \dot r_j + J_j \omega_j \right]$$

This constraint implies that formation maneuvers must be designed to respect angular momentum conservation. The system kinematics can be reduced by projecting the state evolution onto the null space of the constraint, typically achieved by introducing a matrix $S$ such that $\zeta = Sv$, with $v$ the minimal set of free-motion velocities (Takahashi et al., 6 Jan 2026).

The multi-satellite EMFF system is fundamentally nonholonomic. Despite each satellite offering only three directly actuated degrees of freedom (the three coil currents), the controllable directions expand via momentum coupling and nontrivial Lie-bracket structure. Small-time local controllability (STLC) is guaranteed if the Lie algebra generated by the available vector fields and their commutators spans the full accessible space. This has been shown via the Sussmann–Coron theorem for EMFF swarms, establishing that the relative positions and attitudes of the formation are controllable using only internal magnetic actuation and appropriate time-varying controls (Takahashi et al., 6 Jan 2026).

3. Time-Varying Kinematic Control and Lie-Bracket Excitation

The restrictive momentum-conservation-induced nonholonomy of EMFF is addressed by time-varying feedback control laws leveraging the system’s Lie-bracket structure. Explicitly, controllers inject high-frequency oscillatory coil currents (sinusoids or square waves) to excite directions that are otherwise inaccessible by static actuation due to the constraints. The nominal control is

$u_c = [M]S u, \qquad u = -K(v-w(t,q))$

where $K>0$ is a stabilizing gain, $w(t,q)$ is a small, highly oscillatory time-periodic trajectory (frequencies $\omega_1$, $\omega_2$, $\omega_3$ governed by a parameter $\epsilon$), and $S$ projects input into the null space of the angular momentum constraint. The reference $w(t,q)$ is designed per Coron–M’Closkey–Murray conditions to drive the system along appropriate Lie brackets.

Average system behavior under high-frequency control converges exponentially to the desired formation states, as demonstrated by time-varying Lyapunov analysis and averaging theory. The design achieves fully fuel- and reaction-wheel-free formation acquisition and maintenance with locally exponential stability (Takahashi et al., 6 Jan 2026).

4. Alternating Magnetic Field Forces (AMFF) and Decoupling

For practical decentralized control, alternating magnetic field forces (AMFF) are used to decouple pairwise EMFF forces. Each satellite generates its coil moments as a sum of amplitude-modulated sinusoids:

$$u_i(t) = \sum_{j \neq i} p_{ij,k} \sin(\omega_{ij} t)$$

where each pair $(i,j)$ is assigned a unique frequency $\omega_{ij}$, and $p_{ij,k}$ is the amplitude specified for control period $k$. Fundamental orthogonality ensures that, after averaging over integer multiples of periods, the time-averaged force between each pair depends only on their own pair of amplitudes:

$$\frac{1}{T} \int_{kT}^{(k+1)T} f\left(r_{ij}, u_i(t), u_j(t)\right) dt = \tfrac{1}{2} f\left(r_{ij}, p_{ij,k}, p_{ji,k}\right)$$

This mechanism is central to decentralized EMFF: feedback can be computed locally for each pair, and the system as a whole decomposes into independent virtual “links” (Kamat et al., 8 Jan 2026, Kamat et al., 2024, Kamat et al., 25 Aug 2025). The allocation from desired average force to coil amplitude is constructed in closed form for arbitrary $r_{ij}, f_{ij}$ pairs.

5. Decentralized Control, State/Input Constraints, and Control Barrier Functions

EMFF is well suited to decentralized architectures in which each satellite or pair of satellites independently computes the amplitudes required to achieve a prescribed time-averaged force, based only on neighbor-relative positions and velocities. Decentralized virtual spring–damper control laws of the form

$$f^*_{ij,k} = - \left( \frac{2 m |r_{ij,k}|^4}{c_0} \right) \alpha_{ij} \left[ (r_{ij,k} - d_{ij}) + \beta v_{ij,k} \right]$$

realize consensus- or formation-tracking objectives (Kamat et al., 8 Jan 2026).

Extensions to constrained EMFF introduce collision avoidance, maximum intersatellite velocity, and coil power (input) limits enforced via higher-order control barrier functions (CBFs). Multiple CBFs (for collision, speed, and power) are composed via a log-sum-exponential “softmin” to yield a single relaxed barrier function $h(x,\zeta)$. The feedback law integrates optimal control (typically MPC) for formation performance and an online quadratic program to ensure constraint satisfaction at all times. Closed-form expressions for the safe optimal surrogate control action are provided (Kamat et al., 2024, Kamat et al., 25 Aug 2025).

6. Experimental and Numerical Demonstrations

A substantial body of recent experimental and simulation work validates EMFF and AMFF architectures:

• Closed-loop ground-based demonstration: A three-satellite, air-track platform experimentally achieves both attraction and repulsion maneuvers, two-distance formation objectives, and multi-setpoint maneuvers with sub-centimeter accuracy and settling times on the order of tens of seconds. Simulations corroborate the observed transient behavior to within 15% in key performance metrics (Kamat et al., 8 Jan 2026).
• Numerical simulation (constrained EMFF): Three-satellite deep-space and LEO simulations implement the relaxed CBF-based feedback, enforcing collision, speed, and power constraints while asymptotically attaining desired formations. Collisions and physical limits are strictly maintained, and actuator saturation is respected throughout, even when unconstrained LQR would violate them (Kamat et al., 2024, Kamat et al., 25 Aug 2025).
• Fully reaction-wheel-free 3D simulation: A three-satellite swarm simulated with time-varying kinematic control (no attitude actuators) achieves sub-meter position errors and Modified Rodrigues Parameter (MRP) attitude errors $<10^{-3}$. The control inputs remain well within hardware limits, and time histories confirm exponential convergence (Takahashi et al., 6 Jan 2026).

A representative table summarizing major reported experimental and simulation setups appears below.

Reference	Platform/Domain	Satellite Count	Test/Control Features
(Takahashi et al., 6 Jan 2026)	Simulation (3D)	3	Full-coupled, nonholonomic, no reaction wheels
(Kamat et al., 8 Jan 2026)	Air track lab	3	AMFF, 1D, closed-loop, decentralized
(Kamat et al., 2024)	Simulation (3D)	3	CBF safety, constraints, MPC
(Kamat et al., 25 Aug 2025)	Simulation/LEO	2–4	LEO gravity, CBF, formation tracking

7. Practical Considerations, Limitations, and Future Directions

EMFF provides a propellant-free, potentially fully internal means of formation maintenance and reconfiguration, significantly reducing system complexity by eliminating thrusters and reaction wheels. However, several fundamental and implementation-specific limitations are identified:

• High-bandwidth actuation: The requisite fast periodic excitation imposes stringent demands on coil drivers and thermal/battery management.
• Constraint-driven performance bounds: Convergence rate and achievable maneuvers are limited by the small parameter $\epsilon$ (controlling oscillation frequency) and amplitude bounds set by hardware limits.
• Magnetic environment sensitivity: Geomagnetic field variations in LEO necessitate accurate real-time field estimation or measurement to preserve model validity.
• Attitude coupling and extension: Most practical studies to-date focus on translation or single-degree-of-freedom motion; full 6+ degree-of-freedom maneuvers involving attitude require more complex tri-axial control and frequency allocation.
• Scalability: Frequency-multiplexed AMFF, while theoretically scalable to large $n$, hits practical constraints as coil bandwidth and frequency resolution saturate (Kamat et al., 8 Jan 2026).

A plausible implication is that hybrid architectures—incorporating coarse-control methods (e.g., reaction wheels) for large maneuvers, with EMFF providing fine, fuel-free relative control—may address some scalability and bandwidth challenges (Takahashi et al., 6 Jan 2026).

• "Time-Varying Kinematics Control for Magnetically-Actuated Satellite Swarm without Additional Actuator" (Takahashi et al., 6 Jan 2026)
• "Experimental Demonstration of a Decentralized Electromagnetic Formation Flying Control Using Alternating Magnetic Field Forces" (Kamat et al., 8 Jan 2026)
• "Electromagnetic Formation Flying with State and Input Constraints Using Alternating Magnetic Field Forces" (Kamat et al., 2024)
• "Electromagnetic Formation Flying Using Alternating Magnetic Field Forces and Control Barrier Functions for State and Input Constraints" (Kamat et al., 25 Aug 2025)