Robust <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si1.gif" overflow="scroll"><mml:msub><mml:mi>H</mml:mi><mml:mo>∞</mml:mo></mml:msub></mml:math> filters for Markovian jump linear systems under sampled measurements

Hao Zhang, Huaicheng Yan, Junhao Liu, Qijun ChenJournal of Mathematical Analysis and Applications 2009 · DOI

TL;DR

This paper introduces a novel method for designing robust H∞ filters for Markovian jump linear systems that receive measurements at discrete time intervals. The key contribution is a framework that guarantees both stability and a specific performance level for the filtering error, even when system parameters and transition rates are uncertain and belong to known polytopes.

What Problem Does This Paper Address?

Many real-world industrial systems exhibit complex dynamics that change abruptly and unpredictably. These "Markovian jump linear systems" are characterized by discrete "modes" that transition randomly over time, governed by a Markov process. Adding to this complexity, the measurements from these systems are often not continuous but are sampled at discrete time points. Furthermore, the exact parameters of these systems and the rates at which they switch modes can be uncertain, only known to lie within certain bounds (polytopes).

Traditional filtering techniques, like the Kalman filter, struggle with these uncertainties and sampled data. The H∞ filtering approach offers a way to design filters that are robust to noise and uncertainties by minimizing the worst-case impact of disturbances. However, designing such filters for Markovian jump systems with sampled measurements and parameter uncertainties is a significant challenge. This paper aims to bridge this gap by providing a systematic way to design robust H∞ filters that can handle these combined complexities.

Methods

The core of the paper lies in formulating the robust H∞ filtering problem for these uncertain, sampled-data Markovian jump systems in terms of Linear Matrix Inequalities (LMIs). The authors extend previous work by:

  • Modeling: Introducing a new class of Markovian jump linear system models that incorporate polytopic parameter uncertainty, continuous disturbances, and discrete disturbances.
  • Stability Conditions: Deriving sufficient conditions for the stochastic stability of these systems, expressed as LMIs. This involves using Lyapunov functions and Itô's lemma.
  • Filter Design: Developing a method to design a robust H∞ filter. This filter is designed to ensure that the filtering error system is stochastically stable and achieves a prescribed H∞ performance level, meaning the worst-case gain from disturbances to the estimation error is bounded.
  • LMIs for Robustness: The filter design conditions are also cast as LMIs. This is crucial because LMIs can be efficiently solved using convex optimization techniques. The paper provides specific LMI formulations for cases where the transition rate matrix is uncertain but belongs to a polytope, and also for cases where all system matrices are uncertain and belong to polytopes.

Datasets

Not discussed in the paper. The paper focuses on theoretical development and uses a numerical example for demonstration.

Key Insights and Ideas

  • Unified Framework for Uncertainty: The paper successfully unifies several sources of uncertainty (parameter variations within polytopes, Markovian switching, sampled measurements) into a single, solvable filtering problem.
  • Power of LMIs: The reliance on LMIs is a key insight. It transforms a complex control design problem into a well-posed optimization problem, making the design practical.
  • Bridging Continuous and Discrete: The challenge of sampled measurements is elegantly handled by integrating discrete-time measurement updates into a continuous-time system framework, often through augmented state-space representations.
  • Robustness Guarantees: The H∞ framework provides a formal guarantee on the filter's performance under worst-case disturbances, which is critical for safety-critical applications.

Key Results

The paper provides theoretical conditions (in the form of LMIs) that are sufficient for the existence of a robust H∞ filter for the described class of systems. These conditions allow for the design of such filters. A numerical example is presented to demonstrate the effectiveness of the proposed method, showing simulation results of the system states and their estimated counterparts. The example successfully yields a filter and achieves a specified H∞ performance level (γmin = 0.9869).

Limitations and Future Work

  • Sufficient Conditions: The derived conditions are sufficient, not necessarily necessary. This means that a filter might exist even if these specific LMIs are not met.
  • Computational Complexity: While LMIs are solvable, the size and complexity of the LMIs can grow significantly with the number of system modes (N) and the number of vertices in the polytopes describing parameter uncertainty.
  • Future Work: The paper implicitly suggests future work in exploring necessary and sufficient conditions, developing more computationally efficient algorithms for larger systems, and applying the methodology to more complex real-world scenarios. The extension to other types of uncertainties or system nonlinearities could also be a direction.

Summary updated

BibTeX
@article{zhang2009robust,
  author = {Hao Zhang and Huaicheng Yan and Junhao Liu and Qijun Chen},
  title = {{Robust <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si1.gif" overflow="scroll"><mml:msub><mml:mi>H</mml:mi><mml:mo>∞</mml:mo></mml:msub></mml:math> filters for Markovian jump linear systems under sampled measurements}},
  journal = {Journal of Mathematical Analysis and Applications},
  year = {2009},
  doi = {10.1016/j.jmaa.2009.03.027},
}
Figure