Domain of attraction of saturated switched systems under dwell-time\n switching

Masood DehghanarXiv 2014 · DOI · arXiv

Abstract

This paper considers discrete-time switched systems under dwell-time switching and in the presence of saturation nonlinearity. Based on Multiple Lyapunov Functions and using polytopic representation of nested saturation functions, a sufficient condition for asymptotic stability of such systems is derived. It is shown that this condition is equivalent to linear matrix inequalities (LMIs) and as a result, the estimation of domain of attraction is formulated into a convex optimization problem with LMI constraints. Through numerical examples, it is shown that the proposed approach is less conservative than the others in terms of both minimal dwell-time needed for stability and the size of the obtained domain of attraction.

Introduction

In modern engineering, control systems are rarely static. They frequently undergo switching—transitions between different operational modes, such as a vehicle shifting gears or a smart grid rerouting power. Simultaneously, these systems are constrained by saturation, the physical limits of actuators like valves or motors.

The intersection of switching and saturation introduces significant complexity. Rapid switching can induce instability, while saturation limits the range of states from which a system can recover. Engineers must determine the Dwell-time (the minimum duration a system must remain in a mode) and the Domain of Attraction (DOA) (the "safe zone" of initial states guaranteed to return to equilibrium). Traditional methods for estimating these parameters are often overly conservative, leading to inefficient designs that underestimate the system's true capabilities.

Methodology: Addressing Nested Saturation

The core challenge in these systems is that saturation effects become "nested" as the system evolves over multiple time steps during a dwell-time interval. This paper introduces a novel framework to handle this complexity:

  1. Multiple Lyapunov Functions (MLFs): Instead of a single energy function, the author assigns a specific Lyapunov function to each switching mode, allowing for more flexible stability criteria.
  2. Polytopic Representation: Saturation is modeled as a convex hull of linear functions, transforming a non-linear problem into a more manageable linear differential/difference inclusion (LDI).
  3. Auxiliary Variables for Precision: The primary innovation lies in introducing unique auxiliary matrices ($H_{i,t}$) for every individual step within the dwell-time period. This "freedom equals precision" approach provides the optimization algorithm with more degrees of freedom to find a tighter, less conservative stability region. Furthermore, the method acts as a unified framework; when the dwell time is set to a single sample period, the results align with existing theories for arbitrary switching.

By formulating these conditions as Linear Matrix Inequalities (LMIs), the stability problem is converted into a convex optimization task that can be solved efficiently using standard software tools.

Key Results and Performance

The proposed method was validated using a standard numerical benchmark from the literature. The results demonstrate a significant reduction in conservatism compared to previous state-of-the-art methods (e.g., the LDI method of [28]).

Figure 4: Comparison of the proposed method (Corollary 1) against the method of [28] across different dwell times ($\tau$).

As shown in the data in Figure 4, the proposed method (Corollary 1) achieves stability at much lower dwell times ($\tau=2$) where previous methods failed. Furthermore, for a dwell time of $\tau=5$, the proposed method yields a DOA area of 7.143, nearly seven times larger than the 1.131 achieved by the previous method. While the computational cost increases with the number of LMI constraints as $\tau$ grows (reaching 532 constraints for $\tau=8$), the gain in the estimated safe operating region is substantial.

Conclusion

This research provides a powerful tool for engineers to maximize the performance of switched systems under physical constraints. By leveraging nested saturation representations and MLFs, the proposed method offers a more accurate and less conservative estimation of the Domain of Attraction. While the ellipsoidal estimates are inherently safe, future work may explore capturing the potentially non-convex true DOA or extending this framework to optimize dynamic performance metrics beyond simple stability.

Summary updated

Abstract

This paper considers discrete-time switched systems under dwell-time\nswitching and in the presence of saturation nonlinearity. Based on Multiple\nLyapunov Functions and using polytopic representation of nested saturation\nfunctions, a sufficient condition for asymptotic stability of such systems is\nderived. It is shown that this condition is equivalent to linear matrix\ninequalities (LMIs) and as a result, the estimation of domain of attraction is\nformulated into a convex optimization problem with LMI constraints. Through\nnumerical examples, it is shown that the proposed approach is less conservative\nthan the others in terms of both minimal dwell-time needed for stability and\nthe size of the obtained domain of attraction.\n
BibTeX
@misc{dehghan2014domain,
  author = {Masood Dehghan},
  title = {{Domain of attraction of saturated switched systems under dwell-time\textbackslash\{\}n switching}},
  year = {2014},
  eprint = {1411.1486},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1411.1486},
}
Figure