Skip to content
Back to search results

Feedback Linearization

Introduced by
Alberto Isidori
Year
1983
Type
algorithm
Family
Nonlinear Control
Font size

What does it do?

Feedback Linearization is a nonlinear control technique that uses a nonlinear state-feedback transformation to convert a nonlinear system into a linear one, enabling the use of standard linear control methods.

Developed by Isidori, Sontag, and others in the 1980s, feedback linearization is conceptually elegant and powerful: if the system satisfies certain structural conditions (relative degree, decoupling matrix rank), the nonlinearities can be exactly cancelled through feedback, reducing the problem to linear design.

The intuition behind the method

The core idea is that many nonlinear systems can be transformed into linear systems through a carefully designed feedback control law that depends on the state. The key insight is to compute what control input u would be needed to make the system linear with respect to a new input v, then implement u using this state-dependent formula. For example, if a system has the form x_double_dot = f(x) + g(x)·u, feedback linearization finds u such that x_double_dot = v (a linear equation). Once linearized, classical linear control (pole placement, LQR, PID) can be applied to v, and the overall nonlinear control law is the combination of linearization and linear controller.

History

Feedback linearization emerged from differential geometry applied to control systems. Isidori formalized the concept in the 1980s with his foundational work on nonlinear control, proving when and how exact linearization is possible. Sontag and Nijmeijer independently developed input-output linearization approaches, and Hirschorn contributed to the understanding of zero-dynamics. The field matured with textbooks by Isidori (1995) and Slotine & Li (1991), which became standards. Extensions include partial feedback linearization, approximate feedback linearization, and robust feedback linearization for uncertain systems.

When to use it

Use feedback linearization when the system is nonlinear but satisfies the relative degree and decoupling conditions, and when you want to leverage the power of linear control design for a nonlinear system. It is ideal for systems with known, smooth nonlinearities (e.g., electrical machines, aeronautical vehicles). Avoid feedback linearization if the system has uncertain or unmodeled nonlinearities (will fail to linearize), if the relative degree is not well-defined (e.g., algebraic loops), or if the zero-dynamics are unstable (inverted pendulum on a cart is a classical example).

Limitations and considerations

  • Requires perfect knowledge of the system model; uncertain nonlinearities prevent exact linearization and degrade robustness.
  • Zero-dynamics stability is not guaranteed; some systems (e.g., nonminimum-phase systems) have unstable internal dynamics.
  • Only handles systems with a specific structure (relative degree, full decoupling matrix); not all nonlinear systems qualify.
  • Can lead to unbounded control signals if the system approaches singular configurations or constraints.
  • Requires state measurement or accurate estimation; feedback linearization is sensitive to measurement noise.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Exact cancellation of nonlinearities (for systems satisfying relative degree conditions); not an approximation.
  • Transforms the design problem to linear control, enabling use of mature linear methods (pole placement, LQR, H∞).
  • Provides clear conditions (relative degree, full rank) for when exact linearization is possible.
  • Can handle multivariable nonlinear systems elegantly through MIMO input-output decoupling.
  • Theoretical framework is well-established and supported by strong algebraic geometry foundations.

Applications

  • Robotics: Nonlinear manipulator arm control via feedback linearization of nonlinear dynamics (inertia matrix, Coriolis, friction).
  • Aerospace: Aircraft nonlinear flight control, attitude control systems with coupled multi-axis dynamics.
  • Power electronics: DC-DC converter control, synchronous motor speed control exploiting nonlinear electromagnetic dynamics.
  • Chemical processes: Nonlinear reactor control with complex kinetics, distillation column composition control.
  • Marine systems: Ship heading control, autonomous underwater vehicle (AUV) trajectory tracking with hydrodynamic nonlinearities.

How it works

How it works

1

Verify that the system has well-defined relative degree and that the decoupling matrix is invertible; these conditions ensure exact feedback linearization is possible.

2

Compute the inverse dynamics: find the nonlinear state-feedback law u = α(x) + β(x)·v that renders the system linear and controllable.

3

Design a linear controller (e.g., pole placement, LQR) for the linearized system in terms of the new input v.

4

Check that the zero-dynamics (internal dynamics on the constraint manifold) are stable; otherwise the closed-loop system may be unstable.

Common pitfalls

  • Checking relative degree and decoupling conditions incorrectly, leading to incomplete linearization and residual nonlinearities.
  • Ignoring zero-dynamics stability; designing a perfect outer controller for a system with unstable internal dynamics yields an unstable closed-loop.
  • Assuming perfect model knowledge; in practice, parametric uncertainty or unmodeled dynamics cause deviations from the linearized design.
  • Over-relying on linearization without verifying closed-loop robustness to model mismatch.

Frequently asked questions

What is relative degree and why is it important?

Relative degree is the number of times you must differentiate an output to make the input appear explicitly. For example, if y = h(x) and you need to differentiate r times before u appears in y^(r), the relative degree is r. Relative degree must equal the system order for exact feedback linearization to be possible. If relative degree < system order, you have internal (zero) dynamics that must be checked for stability.

What are zero-dynamics and why must they be stable?

Zero-dynamics are the internal dynamics that remain hidden when the input-output map is linearized. They represent the system behavior when the output is constrained to be zero. For the closed-loop system to be stable, zero-dynamics must be stable; if not, the linearized design may stabilize the output but destabilize internal states (e.g., energy building up in unobserved modes).

How does feedback linearization differ from linearization around an equilibrium?

Local linearization (Taylor expansion) only approximates the system near an equilibrium and is valid for small perturbations. Feedback linearization is global (on a large domain) and exact—it cancels nonlinearities through feedback. This is far more powerful, but it requires the system to satisfy strong structural conditions (relative degree, non-singular decoupling matrix).

What happens if the decoupling matrix is singular?

A singular decoupling matrix means the nonlinear system does not decouple into independent SISO (single-input single-output) loops via feedback. This indicates the system has coupling between inputs and outputs that cannot be overcome through feedback alone. You must either redesign the control structure or use alternative methods like sliding mode or MPC.

Related methods

Sources

  1. Isidori, A. (1995). Nonlinear Control Systems (3rd ed.). Springer-Verlag.

    Open DOI reference
    Copied.
  2. Sontag, E. D. (1983). A concept of input-output linearization. Proceedings of the 22nd IEEE Conference on Decision and Control.

    Open publication reference
    Copied.
  3. Nijmeijer, H., & Van der Schaft, A. J. (1990). Nonlinear Dynamical Control Systems. Springer-Verlag.

    Open DOI reference
    Copied.