Autonomous Rendezvous with an Uncertain, Uncooperative Tumbling Target: the TumbleDock Flight Experiments

As the amount of orbital debris grows so too does the need for on-orbit repair and deorbit solutions to avoid cascading Kessler syndrome. While a number of options have been proposed for capturing defunct satellites and other high-value debris, methods for performing close-proximity rendezvous with these objects are also necessary. However, a significant portion of these objects are tumbling with unknown angular orientation and rate; the rendezvous procedure for these tumbling objects is complex and must be performed in real-time, precluding human teleoperation or offline, on-the-ground solutions. Therefore, autonomous rendezvous for these tumbling targets is highly desirable. A fully autonomous rendezvous pipeline was recently proposed by the authors, and has been extended in this paper to a working demonstration in microgravity on resource-constrained hardware. Utilizing factor graph-based SLAM to identify a target object’s rotation, nonlinear programming-based motion planning, and robust control for safe online-updateable reference trajectory tracking, this work overviews the TRACE (Tumbling Rendezvous via Autonomous Characterization and Execution) algorithmic pipeline in its entirety. The pipeline is shown in practice on a sample rendezvous case with a tri-axially tumbling target. The centerpiece of this work is on-orbit results following a two-year hardware implementation and microgravity testing campaign using NASA’s Astrobee robots. A number of implementation considerations are discussed, including augmentations to the Astrobees’ localization system. This work represents the first autonomous on-orbit rendezvous with an uncharacterized, uncooperative tumbling target, to the authors’ knowledge.