← Paper Digest

Dual-Actuation Emulation of Scaled Relative Orbital Dynamics on a Free-Flyer

Elias Krantz, Pedro Roque, Gunnar Tibert, Huina Mao, Christer Fuglesang

November 2026

TL;DR A propeller module on the ATMOS air-bearing free-flyer physically applies scaled orbital (Clohessy-Wiltshire) accelerations, driven by its own MPC, while a separate spacecraft MPC flies the solenoid thrusters. On hardware the spacecraft controller tracks stable relative orbits to within 3 m at orbital scale (3 cm in the lab) and behaves like the same controller in a high-fidelity orbital simulation.

What they do

Air-bearing free-flyer testbeds reproduce the frictionless planar motion of a spacecraft, but not the relative orbital dynamics (the Clohessy-Wiltshire equations) a chaser experiences near a target. The authors add a second actuation system to the open-source ATMOS free-flyer: four bidirectional propellers that physically apply scaled orbital accelerations, leaving the eight solenoid thrusters to the spacecraft controller under test.

How it works

  • Two parallel MPCs. A spacecraft MPC with the CW dynamics in its model runs at orbital scale and commands the on/off thrusters. A CW dynamics MPC runs at laboratory scale and drives the propellers; it carries propeller forces in its state with a hard slew-rate bound and heavily penalizes net yaw torque, so attitude stays with the spacecraft controller.
  • Closed-loop reference. The CW reference is computed analytically and re-initialized with the velocity change from the commanded thrust, so the propellers track a continuously regenerated orbit instead of replaying forces open-loop.
  • Dynamic similitude. Time, length, mass and inertia factors of 50, 100, 8 and 80 map a 100 min low-Earth orbit onto a 2 min lab period within about 15 m² of floor. States are normalized so the solver converges at orbital scale; mean solve times are 1.1-1.4 ms.

Results

  • Scenario: a 152 kg chaser holds three unstable waypoints with yaw maneuvers in between, then flies two stable 2:1 relative orbits.
  • On the stable orbits, spacecraft thrust drops to occasional corrections below 50 mN (orbital scale) while the propellers supply up to about 230 mN. Spacecraft tracking error stays below 3 m (3 cm in the lab); the propeller controller tracks its reference within 8 m (8 cm).
  • The same unmodified controller in the Basilisk simulator tracks within 20 cm, with thrust essentially zero on the stable orbits.

Limitations

One scenario is tested, and agreement with simulation is judged from controller behavior. Residual errors are attributed to imperfect frictionless support and model mismatch, which neither controller estimates. The planar floor omits the cross-track mode, and scaled forces must fit both actuators’ envelopes, which bounds the admissible scales and maneuvers.

Figures

From Figure 2: Dual-actuation control architecture. The spacecraft MPC (orbital scale) tracks the mission trajectory ...
From Figure 2: Dual-actuation control architecture. The spacecraft MPC (orbital scale) tracks the mission trajectory with the solenoid thrusters; its commanded thrust updates the Clohessy-Wiltshire reference generator, whose reference the CW dynamics MPC (laboratory scale) tracks with the bidirectional propellers. Right: the ATMOS free-flyer with the new propeller module.
From Figure 3: Hardware results on ATMOS, shown at orbital scale. Left: spacecraft MPC; right: CW dynamics MPC. The c...
From Figure 3: Hardware results on ATMOS, shown at orbital scale. Left: spacecraft MPC; right: CW dynamics MPC. The chaser holds three unstable static formations, then from 110 min follows two stable relative orbits. Rows show measured and reference positions, position tracking error and commanded body forces. On the stable orbits the spacecraft thrust nearly vanishes while the propellers supply the orbital accelerations.
From Figure 4: Basilisk simulation reference at orbital scale. Only the spacecraft MPC runs, since the simulator's no...
From Figure 4: Basilisk simulation reference at orbital scale. Only the spacecraft MPC runs, since the simulator's nonlinear orbital dynamics replace the CW dynamics MPC. Without modeled disturbances the controller tracks the reference closely and coasts on the stable orbits (from 110 min).
Abstract

Ground-based free-flyer testbeds reproduce the planar motion of a free-floating spacecraft using air bearings, but not the relative dynamics of spacecraft in orbit. We present a dual-actuation architecture that physically emulates Clohessy-Wiltshire dynamics on an open-source free-flyer. Two model predictive controllers run in parallel: one drives the solenoid thrusters with an orbital dynamics model, representing the spacecraft under test; the other drives bidirectional propellers that physically apply the orbital accelerations, continuously updating its reference to account for thruster-induced velocity changes. A dynamic similitude maps the orbital scenario onto the limited floor area and actuation rates of the testbed, letting the spacecraft controller run directly at orbital scale. We validate the architecture on hardware and against a high-fidelity orbital simulation, where the orbital accelerations instead arise naturally, both driven by the identical spacecraft controller. Their agreement indicates that the testbed reproduces the relative orbital motion the controller would encounter in flight.

Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.