Skip to content
Back to search results

Backstepping Control

Introduced by
Miroslav Krstic
Year
1995
Type
algorithm
Family
Nonlinear Control
Font size

What does it do?

Backstepping is a systematic nonlinear control design method that decomposes a complex nonlinear system into simpler subsystems and designs a controller recursively, layer by layer, ensuring stability at each step.

Developed by Krstic, Kanellakopoulos, and Kokotovic, backstepping enables control of nonlinear systems without requiring exact model knowledge or full state linearization, combining flexibility with guaranteed stability.

The intuition behind the method

Backstepping decomposes an nth-order system into n first-order subsystems by treating intermediate states as virtual controls. For each layer, a Lyapunov function is designed, ensuring stability of that subsystem. The next layer's controller is then designed to stabilize the error between desired and actual virtual control, 'backstepping' through the hierarchy. This recursive approach guarantees global asymptotic stability if each layer is stable. Unlike feedback linearization which requires exact cancellation of nonlinearities, backstepping allows nonlinearities to remain and compensates for them, enabling robustness to parametric uncertainty.

History

Krstic and colleagues developed backstepping in the mid-1990s for stabilization and tracking of uncertain nonlinear systems. The method has been extended to time-varying systems, systems with unmodeled dynamics, and output-feedback cases. Modern variants include adaptive backstepping for unknown parameters and event-triggered backstepping for networked control.

When to use it

Use backstepping for nonlinear systems with strict-feedback structure (cascade of integrators with nonlinearities), when you need guaranteed stability with Lyapunov proof, and when system parameters are uncertain. Ideal for nonlinear aircraft control, underwater vehicles, and robot arms. Avoid backstepping if your system does not have strict-feedback structure or if you require fast computation (backstepping design is offline).

Limitations and considerations

  • Requires strict-feedback structure; does not apply to general nonlinear systems.
  • Control law can be complex, involving high-order derivatives (peaking phenomenon).
  • Backstepping gains can be aggressive, leading to large control effort and actuator saturation.
  • Design is offline; not suitable for adaptive real-time parameter adjustment.
  • Peaking: transient overshoot can occur in the first layer, driving other states into nonlinear region.

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

Strengths

  • Systematic design with guaranteed global asymptotic stability proven via Lyapunov analysis.
  • Handles nonlinear systems without linearization or exact cancellation.
  • Natural incorporation of uncertainty and robustness through Lyapunov redesign.
  • Flexibility in virtual control selection allows trade-offs between performance and robustness.
  • Naturally handles cascaded systems and systems with integrator chains.

Applications

  • Aerospace: Aircraft altitude and heading control with uncertain aerodynamics.
  • Robotics: Nonlinear robot arm control with uncertain payloads and friction.
  • Underwater vehicles: AUV depth and orientation control with hydrodynamic uncertainty.
  • Power systems: Nonlinear generator control and synchronization.
  • Biochemical processes: Bioreactor concentration control with uncertain kinetics.

How it works

How it works

1

Write system in strict-feedback form x_dot = f(x) + g(x)u; decompose into subsystems with virtual controls z_1, ..., z_{n-1}.

2

For layer i, design virtual control z_i* and controller gain k_i ensuring V_i = V_{i-1} + (1/2)(x_i − z_i*)^2 is a Lyapunov function decreasing along solutions.

3

Check that all designed gains ensure global asymptotic stability; use Lyapunov analysis through all layers.

4

Implement final control law u = k_{n-1}(x), combining all recursive feedback laws into actual control input.

Common pitfalls

  • Choosing Lyapunov functions arbitrarily; systematic design requires careful selection of performance functions.
  • Ignoring peaking phenomenon; leads to loss of stability despite correct design.
  • Not normalizing states; unbounded gains occur if state dimensions have different scales.
  • Ignoring actuator saturation; designed control law may exceed physical limits.

Frequently asked questions

What is strict-feedback form and do I need it for backstepping?

Strict-feedback form is: x_1_dot = x_2, x_2_dot = x_3, ..., x_n_dot = f(x_1,...,x_n) + g(x_1,...,x_n)u. Each state's derivative depends only on lower-indexed states plus a control term. This structure enables recursive design. Systems not in strict-feedback require preliminary transformations or partial backstepping.

What is the peaking phenomenon and how do I avoid it?

Peaking is transient overshoot in early layers that can destabilize later layers. It occurs when gains are aggressive. Mitigation: (1) reduce gain magnitudes, (2) introduce integrator backstepping with low-pass filtering, (3) saturate virtual controls, (4) use command filtering to limit virtual control derivatives.

Related methods

Sources

  1. Krstic, M., Kanellakopoulos, I., & Kokotovic, P. (1995). Nonlinear and Adaptive Control Design. John Wiley & Sons.

    Open publication reference
    Copied.