Hardware Implementation of Learning Reference Governors for Spacecraft Rendezvous and Proximity Maneuvering with Mobile Robots
Jonathan Heidegger, Samantha Romano, Abhiram Reddy Kondur, Anouck Girard, Ilya V. Kolmanovsky · 2024
Automated rendezvous, proximity operations, and docking (RPOD) are essential elements of on-orbit refueling and servicing, space station resupply, and debris removal missions. These operations require ensuring constraint satisfaction for safe mission completion while accounting for uncertainties in system dynamics and behavior. In this paper, the application of a Learning Reference Governor (LRG) is considered to address these challenges. An LRG is an add-on scheme that relies on learning instead of an explicit dynamic model of the system, which guarantees constraint satisfaction during and after learning part of the mission. Hardware testing and validation is a vital step in progressing algorithms from lab to deployment. This paper describes a hardware testbed built with mobile robots for verifying spacecraft controllers and its use for LRG. These holonomic mobile robots were programmed to emulate the relative motion of spacecraft in the presence of computational limitations, system noise, model mismatch, and sensor noise. This paper reports the outcomes of the implementation of LRG in this new hardware testbed.