Abstract
In this paper, a new asymptotic stability condition of continuous fuzzy system is proposed. The new stability condition considers the nonquadratic stability by using the P-matrix measure. Later the relationship of the suggested stability condition and the well-known stability condition is discussed and it is shown in a rigorous manner that the proposed criterion includes the conventional conditions.
Introduction
In control theory, Takagi-Sugeno (T-S) fuzzy systems are a powerful tool for representing complex nonlinear systems by decomposing them into a set of simpler linear subsystems. However, ensuring the stability of these systems remains a significant challenge. The traditional approach relies on finding a Common Quadratic Lyapunov Function (CQLF)—a single energy-like function that must decrease for every subsystem.
The limitation of this approach is its inherent conservatism; many stable systems do not possess a quadratic Lyapunov function. This "quadratic restriction" often forces engineers to reject stable controllers simply because their stability cannot be proven using standard quadratic tools. This paper addresses this bottleneck by proposing a more flexible, nonquadratic stability framework.
Methodology
The authors introduce a stability criterion based on the $p$-matrix measure (also known as the logarithmic norm). Instead of the traditional quadratic form $V(x) = x^T P x$, they propose a nonquadratic Lyapunov function candidate:
$$V(x(t)) = \|Q x(t)\|_p^2$$
where $\|\cdot\|_p$ is the $p$-norm and $Q$ is a nonsingular matrix. The core technique involves a similarity transformation using a matrix $S$ (where $Q = S^{-1}$) to transform the system's state space. Stability is then verified by checking if the $p$-matrix measure of the transformed system matrices, $\mu_p(S^{-1} A_i S)$, is negative for all subsystems.
Key Results and Generalization
The most significant contribution of this work is the rigorous proof that the proposed $p$-matrix measure condition is a true generalization of classical quadratic stability. In Theorem 4, the authors demonstrate that for $p=2$ (the Euclidean norm), the existence of a common $P$ matrix is exactly equivalent to the new $p$-matrix condition.
By utilizing different $p$-norms, researchers can define stability regions of various shapes. For example, using the $\infty$-norm allows for rectangular or "boxy" stability regions, which may better fit the dynamics of certain physical systems than ellipsoidal quadratic regions. The paper provides specific formulas for calculating these measures for $p=1, 2,$ and $\infty$, as detailed in Table 1.
Conclusion and Future Work
This research bridges the gap between modern matrix measure theory and classical Lyapunov stability, showing that the former provides a broader umbrella for stability analysis. While the theoretical framework is robust, future work is needed to develop efficient algorithms (such as LMI solvers) to automatically find the optimal similarity transform $S$. Additionally, further investigation into how fuzzy membership functions (premise variables) interact with these nonquadratic conditions will be essential for practical applications in complex control systems.
Summary updated