Systematic Comparison of Diverse Control Architectures in an Adversarial Multiagent Orbital Scenario

Increasing orbital congestion and the growth of decentralized distributed missions necessitate an increased demand for efficient controls. Space operations, including satellite constellation coordination and in-service assembly and maintenance (including docking) require effective close proximity controllers. These applications may include close range maneuvering relative to non-cooperative targets, spanning a spectrum of unresponsive to adversarial. The absence of analytic solutions for dynamic space systems facilitates the need to develop unconventional solutions for rendezvous operations. Systematic performance comparison tools across control methods are needed to systematically understand the spaces in which diverse control techniques (frequency domain, state-space, and data-driven control) perform best. Considering the Lady-Bandit-Guard problem with a scoring system using the Kerbal Space Program Differential Games (KSPDG) environment, various controllers can be compared in repeatable scenarios. In addition to performance, each control method is evaluated for computational efficiency. This paper examines three particular control methods: PID, LQR, and Game Theory, previously developed for the aforementioned KSPDG competition. The resulting analysis guides the real-world applicability of various control methods, while also presenting promising paths that can be followed for future controller developments.