Data Driven Modeling and Graph-Theoretic Synchronization for Jet-Powered Robotic Propulsion Systems

Abstract

This paper investigates synchronization control for a robotic power system driven by commercial Micro Turbine Engines (MTEs). To address the challenge of controlling commercial engines with undisclosed parameters and complex nonlinear throttle-to-speed responses, we propose a framework integrating regularized data-driven identification with graph-theoretic synchronization. First, high-fidelity implicit models for two distinct engine groups (SW300B and SW400P) are established utilizing Nonlinear AutoRegressive models with eXogenous inputs (NLARX) and regularized neural networks, trained on experimental bench data. Subsequently, a generalized relative-coupling control law is derived within a Laplacian topology, supported by Lyapunov analysis to guarantee the Uniform Ultimate Boundedness (UUB) of the closed-loop system. Validated via simulations, the architecture reduces inter-engine speed mismatches by 35.5% and 51.3% for the SW300B and SW400P groups, substantiating its effectiveness for coordinated propulsion.