Autonomous Spacecraft Rendezvous using Tube-based Model Predictive Control: Design and Application

Abstract

As the concentration of large space debris increases, how rendezvous maneuvers involving these typically non-cooperative, freely-tumbling bodies are planned and executed is evolving. The rendezvous must be carefully planned, employing up-to-date in situ data to identify the inertial and motion parameters of the target body, and executed in a manner which accounts for the remaining uncertainty in these parameters. This paper presents an extension of the TRACE pipeline used in the ROAM/TumbleDock Astrobee experiment campaign, which sequences the target state estimation, motion planning, controller design, and maneuver execution tasks while additionally providing logical loop-back avenues to previous tasks, increasing the chances of a successful maneuver. The pipeline's performance is analyzed in simulation, utilizing: target state estimates generated in a previous activity on a dedicated on-ground testbed; online motion planning, based on non-linear programming and warm-started using a trajectory library generated offline with a novel GPU-based method; and Tube-based Model Predictive Control to robustly track the planned trajectory. Tube-based Model Predictive Control is an actively evolving subject, distributed over multiple publications and various research interests. The necessary theory and considerations for practical implementation of the method are consolidated; its use, features, and limitations in the proposed task are demonstrated.