Distributed Stabilization of State Interdependent Markov Jump LinearSystem-of-Systems with Partial Information

Guanze Peng, Juntao Chen, Quanyan ZhuarXiv 2020 · OpenReview

TL;DR

This paper tackles the challenge of stabilizing two interconnected systems where their operational modes depend on each other's states, and these modes are only partially observable. The authors propose a novel framework and design both centralized and distributed controllers that can achieve stochastic stability for these complex, interdependent systems, even with limited information.

What Problem Does This Paper Address?

Many real-world systems, like power grids or communication networks, are not isolated but rather interdependent. The failure or operational state of one system can directly influence the behavior and even the "mode" (e.g., normal operation, degraded state, failure) of another. This interdependency is particularly tricky when the exact operational mode of a system isn't directly known, but only inferred from noisy signals. Traditional control methods often assume full knowledge of system states and modes, which is unrealistic in these scenarios. This paper addresses this gap by developing a control strategy for two such interdependent Markov Jump Linear Systems (MJLSs) where the mode transitions are state-dependent and the modes themselves are only partially observed.

Methods

The core of the paper lies in developing a mathematical framework for these interdependent MJLSs. They first formulate an "integrated system" model that combines the two individual MJLSs, capturing their interactions.

  • Interdependent MJLS Model: The key novelty here is modeling the transition rates of one system's Markov jump mode as a function of the state of the other system. This captures cascading effects.
  • Partial Information: Instead of directly observing the system's mode (e.g., "System 1 is in mode A"), the controllers only receive an "observation signal" that provides partial, probabilistic information about the true mode.
  • Centralized Control: Initially, they design a controller that assumes access to the global state information of both systems. This serves as a baseline and helps derive stability conditions.
  • Distributed Control: Recognizing that global information might be impractical, they then develop distributed controllers. Each system's controller only needs its own state and observation, plus information about the other system's state partition and observation. This significantly reduces communication overhead.
  • Stability Conditions: The stability of the closed-loop system is analyzed using Lyapunov functions and the Dynkin's formula. The sufficient conditions for stability are derived in the form of Linear Matrix Inequalities (LMIs), which are standard tools for controller synthesis in this domain.

Datasets

Not discussed in the paper. The paper presents a numerical example with parameters defined by the authors, rather than using pre-existing datasets.

Key Insights and Ideas

  • State-Dependent Mode Transitions are Crucial: The paper highlights that simply having interdependent systems isn't enough; the mode transitions themselves being influenced by the other system's state is a critical factor in modeling complex failures or interactions.
  • Partial Observation Doesn't Mean No Control: A significant takeaway is that even with incomplete information about system modes, effective stabilization is achievable. This opens doors for controlling real-world systems where sensors are imperfect or communication is limited.
  • Distributed Control is Feasible and Beneficial: The development of distributed controllers is a major contribution. It shows that we can achieve system-wide stability without requiring every controller to know everything about every other system, drastically simplifying implementation and reducing communication burdens.
  • LMIs as a Powerful Synthesis Tool: The use of LMIs to derive sufficient stabilization conditions and synthesize controllers demonstrates their continued relevance and power in handling complex stochastic systems.

Key Results

The paper successfully derives sufficient conditions for the stochastic stabilization of the interdependent MJLSs under both centralized and distributed control schemes with partial information. A numerical example is provided, demonstrating the effectiveness of the designed distributed controllers. The results show that the distributed controllers can stabilize the system with satisfactory performance, even when system modes are not directly observable, and that this performance is comparable to, though slightly smoother with, full mode observation.

Limitations and Future Work

  • Limitations: The paper focuses on two interdependent MJLSs. Extending this to a larger number of interconnected systems would be a natural next step. The current work also focuses on stabilization; extending it to optimal control (considering state and control costs) is a significant area for future research.
  • Future Work: The authors explicitly suggest extending the stabilizing control design to optimal control design, considering both state and control costs for coupled MJLSs under incomplete information. This would provide a more comprehensive control solution for practical applications.

Summary updated

Abstract

In this paper, we study the stabilization of two interdependent Markov jump linear systems (MJLSs) with partial information, where the interdependency arises as the transition of the mode of one system depends on the states of the other system. First, we formulate a framework for the two interdependent MJLSs to capture the interactions between various entities in the system, where the modes of the system cannot be observed directly. Instead, a signal which contains information of the modes can be obtained. Then, depending on the scope of the available system state information (global or local), we design centralized and distributed controllers, respectively, that can stochastically stabilize the overall interdependent MJLS. In addition, the sufficient stabilization conditions for the system under both types of information structure are derived. Finally, we provide a numerical example to illustrate the effectiveness of the designed controllers.
BibTeX
@misc{peng2020distributed,
  author = {Guanze Peng and Juntao Chen and Quanyan Zhu},
  title = {{Distributed Stabilization of State Interdependent Markov Jump LinearSystem-of-Systems with Partial Information}},
  year = {2020},
  eprint = {2003.06493},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2003.06493},
}
Figure