Bounded tracking for non-minimum phase nonlinear systems with fast zero dynamics

Claire J. Tomlin, S. Shankar SastryInternational Journal of Control 1997 · DOI

TL;DR

This paper tackles the challenge of controlling non-minimum phase nonlinear systems that exhibit "fast zero dynamics" – a common issue in flight control. The authors develop a novel method that combines existing techniques to ensure control inputs remain bounded even as system complexities are introduced, a crucial step for practical applications.

What Problem Does This Paper Address?

Controlling nonlinear systems is already complex, but it becomes significantly harder when the system is "non-minimum phase." This means that even if the system's primary dynamics are stable, its "zero dynamics" (related to internal states not directly observed by the output) can be unstable. This instability can lead to extremely large and impractical control inputs when trying to make the system follow a desired trajectory.

Adding to this complexity, the paper focuses on systems with "fast zero dynamics." These arise when a system is a slight perturbation of a simpler, "nominal" system. The perturbation introduces these fast, potentially unstable, internal dynamics that can be problematic. Imagine trying to steer an aircraft: if its internal stabilization system is slightly off, it might react very quickly and erratically, requiring precise and bounded control to prevent instability. The core problem is to design control laws that guarantee the system tracks a desired output while keeping the control effort within reasonable limits, especially in the presence of these fast, unstable zero dynamics.

Methods

The paper's approach is a clever synthesis of existing work. It builds upon:

  • Devasia, Paden, and Chen's work on exact tracking for nonlinear systems: This provides a framework for designing controllers that can achieve precise output tracking.
  • Hunt, Meyer, and Su's approach using an "error system": Instead of directly controlling the complex perturbed system, they propose creating an "error system" by comparing the actual system to a simpler, nominal system. This simplifies the control design problem.
  • Singular Perturbation Theory: This mathematical tool is used to analyze systems with multiple time scales (fast and slow dynamics) and to derive explicit bounds on control inputs.

The core methodology involves a two-step procedure:

  1. Design a bounded control for the nominal system: This is done using the Devasia-Paden-Chen scheme, adapted for potentially hyperbolic zero dynamics.
  2. Adapt this control for the perturbed system: This involves defining an "error system" and then applying a modified Devasia-Paden-Chen algorithm to this error system. The key is to use singular perturbation techniques to analyze how the control input behaves as the perturbation parameter approaches zero, ensuring boundedness.

Datasets

Not discussed in the paper. The paper focuses on theoretical development and applies its methods to simulated aircraft models.

Key Insights and Ideas

  • The "fast zero dynamics" problem is a perturbation issue: The paper highlights that these problematic dynamics often arise from small deviations from a well-behaved nominal system. This insight allows for a more structured approach to control design.
  • An "error system" is a powerful simplification: By focusing on the difference between the actual and nominal system, the control design problem can be made more tractable, especially when dealing with complex nonlinearities.
  • Singular perturbation theory is key to boundedness: The paper demonstrates how this theory can be used to derive explicit bounds on control inputs, providing theoretical guarantees that the control effort won't become excessively large.
  • Subtle conditions are crucial for bounded tracking: The paper emphasizes that achieving bounded tracking requires careful consideration of the system's properties, particularly the hyperbolicity of its zero dynamics and the Lipschitz continuity of certain functions.

Key Results

The paper provides a theoretical framework and conditions under which bounded tracking control laws can be derived for nonlinear systems with fast zero dynamics. The main results are:

  • A general method for deriving bounded tracking control laws: This method is applicable to both SISO (single-input, single-output) and MIMO (multiple-input, multiple-output) systems.
  • Explicit bounds on control inputs: The authors show how to calculate these bounds, which are critical for practical implementation.
  • Successful application to VTOL and CTOL aircraft models: The proposed method is demonstrated on simplified planar models of these aircraft, which are known to exhibit the type of non-minimum phase behavior with fast zero dynamics that the paper addresses.

Limitations and Future Work

  • Single Time Scale Assumption: The paper acknowledges that its analysis relies on an assumption of a single time scale for the "fast" zero dynamics. Removing this assumption is a direction for future work.
  • "Reasonably Complete" Conditions: The authors state that their description of conditions for bounded tracking is "reasonably complete," implying there might be edge cases or more general scenarios not fully covered.
  • Generalization to Non-Invertible Systems: While the paper considers invertible systems, extending the results to non-invertible systems would be a valuable next step.
  • More Complex Dynamics: Applying the method to more realistic, higher-dimensional aircraft models with more complex dynamics would be a practical extension.

Summary updated

Abstract

In this paper, we derive tracking control laws for non-minimum phase nonlinear systems with both fast and slow, possibly unstable, zero dynamics. The fast zero dynamics arise from a perturbation of a nominal system. These fast zeros can be problematic in that they may be in the right half plane and may cause large magnitude tracking control inputs. In this paper, we combine the ideas from some recent work of Hunt, Meyer and Su with that of Devasia, Chen and Paden on an asymptotic tracking procedure for non-minimum phase nonlinear systems. We give (somewhat subtle) conditions under which the tracking control input is bounded as the magnitude of the perturbation of the nominal system becomes zero. Explicit bounds on the control inputs are calculated for both SISO and MIMO systems using some interesting non-standard singular perturbation techniques. The method is applied to a suite of examples, including the simplified planar dynamics of VTOL and CTOL aircraft.
BibTeX
@article{tomlin1997bounded,
  author = {Claire J. Tomlin and S. Shankar Sastry},
  title = {{Bounded tracking for non-minimum phase nonlinear systems with fast zero dynamics}},
  journal = {International Journal of Control},
  year = {1997},
  doi = {10.1080/002071797223352},
}
Figure