Abstract: Nonlinear dynamical behaviours in engineering applications can be approximated by linear-parameter varying (LPV) representations, but obtaining precise model knowledge to develop a control algorithm is difficult in practice. In this paper, we develop the data-driven control strategies for event-triggered LPV systems with stability verifications. First, we provide the theoretical analysis of $\theta$-persistence of excitation for LPV systems, which leads to the feasible data-based representations. Then, in terms of the available perturbed data, we derive the stability certificates for event-triggered LPV systems with the aid of Petersen's lemma in the sense of robust control, resulting in the computationally tractable semidefinite programmings, the feasible solutions of which yields the optimal gain schedulings. Besides, we generalize the data-driven event-triggered LPV control methods to the scenario of reference trajectory tracking, and discuss the robust tracking stability accordingly. Finally, we verify the effectiveness of our theoretical derivations by numerical simulations.
Introduction
Many complex engineering systems—such as hypersonic aircraft or autonomous vehicles—exhibit nonlinear behaviors that change depending on operating conditions. These are often modeled as Linear Parameter-Varying (LPV) systems, where the system dynamics are linear but depend on a time-varying "scheduling signal."
Traditionally, controlling these systems requires a precise mathematical model. However, building such a model from first principles is difficult, and identifying one from data is computationally expensive and prone to errors. Furthermore, in modern networked control systems, communication bandwidth is a critical resource. To address this, event-triggered control is employed to avoid constant signal transmission. This paper introduces a direct data-driven framework for designing event-triggered controllers for LPV systems, bypassing explicit model identification while guaranteeing stability and tracking performance.
System Architecture
The proposed framework moves from raw data to a stable, efficient controller through a structured pipeline. The architecture integrates offline data collection with an online event-triggered control center to manage the unknown LPV plant.
Methodology
The authors propose several key steps to achieve robust control directly from data:
- $\theta$-Persistence of Excitation: A new condition is defined for LPV systems to ensure that collected data is "rich" enough to represent the underlying dynamics. This acts as a mathematical prerequisite for any data-driven analysis.
- Data-Based Representation: Instead of identifying system matrices $A$ and $B$, the framework uses an extension of Willems’ Fundamental Lemma to represent dynamics directly using matrices of collected input, state, and parameter data.
- Robust Stability Certificates: To handle noise and perturbations in the data, the authors utilize Petersen’s Lemma and the full-block S-procedure. These tools allow for the derivation of stability conditions (specifically Input-to-State Stability) as computationally tractable Linear Matrix Inequalities (LMIs).
- Event-Triggering Logic: A triggering mechanism is designed to update the control signal only when the difference between the current state and the last transmitted state exceeds a specific threshold.
- Integral Compensation for Tracking: An auxiliary integral compensator is introduced to allow the system to follow complex reference trajectories accurately.
Key Results and Insights
The effectiveness of the approach was validated through numerical simulations, including 2nd-order LPV stabilization and complex 2D plane tracking (e.g., following a "shape-8" trajectory). Key findings include:
- Direct Control: The core innovation is skipping the model identification step entirely; the data matrices themselves act as the model within the stability formulation.
- Communication Efficiency: The event-triggered approach significantly reduces network load. Simulations showed inter-event intervals of 0.1 to 0.3 seconds despite a 0.01s sampling rate.
- Robustness by Design: By treating data perturbations as bounded uncertainties, the resulting controller is inherently robust to the noise present in the training data.
- Computational Tractability: All design steps are reduced to Semidefinite Programming (SDP), which can be solved efficiently using standard toolkits like YALMIP and MOSEK.
Limitations and Future Work
While the framework provides strong theoretical guarantees for practical exponential Input-to-State Stability (ISS), some challenges remain. The number of decision variables in the LMIs increases with the number of scheduling parameters and state dimensions, which may affect scaling for very high-dimensional systems. Additionally, the current method assumes full state measurement and known upper bounds for noise. Future research could extend these methods to output-feedback scenarios and cases where the scheduling signal itself is noisy or delayed.
Summary updated