Abstract
The present paper deals with data-driven event-triggered control of a class of unknown discrete-time interconnected systems (a.k.a. network systems). To this end, we start by putting forth a novel distributed event-triggering transmission strategy based on periodic sampling, under which a model-based stability criterion for the closed-loop network system is derived, by leveraging a discrete-time looped-functional approach. Marrying the model-based criterion with a data-driven system representation recently developed in the literature, a purely data-driven stability criterion expressed in the form of linear matrix inequalities (LMIs) is established. Meanwhile, the data-driven stability criterion suggests a means for co-designing the event-triggering coefficient matrix and the feedback control gain matrix using only some offline collected state-input data. Finally, numerical results corroborate the efficacy of the proposed distributed data-driven ETS in cutting off data transmissions and the co-design procedure.
Introduction
In Networked Control Systems (NCS), multiple subsystems exchange information over a shared communication channel to achieve a global goal. Traditionally, this is achieved via periodic sampling, where updates are sent at fixed time intervals. However, in large-scale systems like power grids or robot swarms, this "constant chatter" can lead to significant waste of bandwidth and energy.
Event-Triggered Systems (ETS) offer a solution by transmitting data only when a specific "event"—such as a significant state deviation—occurs. However, designing these triggers typically requires an exact mathematical model (e.g., state-space matrices) of the system. In practice, systems are often too complex to model accurately, or their parameters may change over time. This paper addresses the challenge of designing efficient, distributed event-triggered control for unknown discrete-time systems using only raw input-output data.
Methodology
The proposed approach integrates data-driven system representation with a novel triggering mechanism to ensure stability without requiring a physical model.
1. Distributed Dynamic ETS
The authors developed a new triggering law that utilizes both local state information and data from neighboring subsystems. Unlike static triggers, this mechanism includes a dynamic variable ($\eta$) that serves as a flexible threshold. This "memory" mechanism allows for longer intervals between transmissions by accounting for past stability, thereby reducing network load without compromising performance.
2. Data-Driven System Representation
Instead of relying on physics-based equations, the framework employs a data-driven representation based on Willems’ Fundamental Lemma. This allows the system to be characterized using a collection of offline-collected state and input data, even in the presence of measurement noise.
3. Stability Analysis and Co-design
To ensure system stability, a discrete-time looped-functional mathematical framework is used. This approach is less conservative than traditional Lyapunov methods, allowing for more efficient control design. The final result is a set of Linear Matrix Inequalities (LMIs) that enable the simultaneous co-design of event-triggering parameters and feedback control gains using standard optimization software.
Experimental Evaluation
The method was evaluated using a simulated network of three coupled inverted pendulums connected by springs.
- Data Collection: 200 offline measurements were generated by applying random inputs and recording states, with added random noise to simulate real-world sensor inaccuracies.
- Comparison: The data-driven approach was compared against a model-based version to quantify the performance trade-offs associated with model uncertainty.
Key Results and Insights
- Communication Efficiency: In the pendulum simulation, the system maintained stability with only 109 transmissions out of 450 possible sampling instances.
- Distributed vs. Decentralized: The distributed strategy (using neighbor information) outperformed decentralized methods (local information only), requiring significantly fewer transmissions (109 vs. 148) for the same control performance.
- Robustness to Noise: Despite sensor noise and the lack of a physical model, all subsystem states successfully converged to zero.
- The "Data Tax": Results indicate that data-driven control is inherently more conservative than model-based control. The lack of exact physical knowledge necessitates more "cautious" control parameters, which can lead to slightly slower convergence times.
Conclusion and Future Work
While the proposed method effectively reduces communication overhead in unknown network systems, some challenges remain. The computational complexity of solving LMIs increases with the amount of collected data, which may limit scalability for massive networks. Additionally, while control execution is distributed, the design phase currently requires global network information. Future research will focus on making the design phase fully distributed and extending the framework to provide $L_2$-gain performance guarantees for better disturbance rejection.
Summary updated