Finite-Time and Fixed-Time Convergence in Control Systems

90 concepts77 papers3 proposals

In classical control theory, the gold standard for stability is asymptotic convergence. If you perturb a system, a well-designed controller guarantees that the tracking error will eventually vanish. Mathematically, this is often modeled by a linear differential equation like $\dot{x} = -kx$, which yields an exponential decay: $x(t) = x(0)e^{-kt}$.

While elegant, exponential decay has a fundamental practical flaw: the error only reaches exactly zero as time approaches infinity ($t \to \infty$). For many real-world applications, "eventually" is not fast enough. As Xi Ma et al. observe in Attitude control of rigid body with inertia uncertainty and saturation input, mission-critical maneuvers like spacecraft docking require systems to reach their target states within a strictly defined, predictable time window. Whether coordinating a multi-robot rendezvous (Giuseppe Notarstefano, Distributed consensus on enclosing shapes and minimum time rendezvous) or satisfying strict deadlines in temporal logic specifications (Wei Xiao et al., High Order Control Lyapunov-Barrier Functions for Temporal Logic Specifications), engineers need guarantees that errors will hit absolute zero in a finite amount of time.

To achieve this, researchers have developed a suite of non-asymptotic control frameworks: finite-time, fixed-time, and prescribed-time stability.


Breaking Lipschitz Continuity for Finite-Time Convergence

How do you force a dynamical system to reach its target state exactly in finite time? The answer lies in violating a core assumption of classical calculus: Lipschitz continuity.

If a system's governing differential equation is Lipschitz continuous, its trajectories are unique, meaning a path starting outside an equilibrium point can never intersect it in finite time. To bypass this limitation, we must introduce non-smooth or discontinuous dynamics. As Jorge Cortés details in Discontinuous dynamical systems, discontinuous vector fields allow a system to reach equilibrium in a finite, measurable duration.

The most common way to build this intuition is through a power-law feedback loop. Consider a scalar system governed by the following non-linear differential equation:

$$\dot{x} = -c \cdot \text{sign}(x)|x|^\alpha$$

where $c > 0$ and $0 < \alpha < 1$.

To see why this works, compare it to a standard linear system ($\alpha = 1$). In a linear system, as the state $x$ gets closer to zero, the restoring velocity $\dot{x}$ drops proportionally. If $x = 0.01$, the speed pulling it toward zero is tiny.

However, when $\alpha$ is a fraction (e.g., $\alpha = \frac{1}{3}$), the behavior changes dramatically. If $x = 0.01$, then $|0.01|^{1/3} \approx 0.215$. The restoring velocity remains comparatively large even as the error shrinks to almost nothing. If you integrate this equation, you find that the state hits exactly zero at a specific settling time $T$:

$$T = \frac{|x(0)|^{1-\alpha}}{c(1-\alpha)}$$

This mathematical mechanism is the foundation of Terminal Sliding Mode Control (TSMC). By utilizing fractional powers, TSMC forces tracking errors to zero along a sliding surface in finite time, a strategy deployed by Pan Song et al. in A terminal sliding mode based torque distribution control for an individual-wheel-drive vehicle to handle high-speed automotive maneuvers. Similarly, Shiyu Zhao 0002 et al. demonstrate in Distributed Control of Angle-constrained Circular Formations using Bearing-only Measurements that tuning a control law's exponent below unity explicitly transitions a multi-agent system from exponential to finite-time stability.


The Initial-State Problem and Fixed-Time Stability

While finite-time stability is a massive upgrade over asymptotic stability, it suffers from a major vulnerability: the settling time $T$ depends directly on the initial state $x(0)$.

If a system experiences an unexpectedly large displacement or a massive external shock, $x(0)$ blows up, and the time required to recover grows arbitrarily large. As Bing Yi et al. note in Fixed-time adaptive fuzzy control for time-varying systems based on nominal substitution, this dependency prevents engineers from guaranteeing a strict, uniform upper bound on convergence time across arbitrary initial conditions.

Fixed-time stability solves this by blending two different power laws into a single controller:

$$\dot{x} = -c_1 \cdot \text{sign}(x)|x|^\alpha - c_2 \cdot \text{sign}(x)|x|^\beta$$

where $0 < \alpha < 1$ and $\beta > 1$.

This architecture creates a two-regime defense:

  1. When the error is large ($|x| > 1$): The super-linear term ($|x|^\beta$) dominates. Because $\beta > 1$, the restoring force grows exponentially faster than the error, rapidly dragging the system back toward the origin regardless of how far away it started.
  2. When the error is small ($|x| < 1$): The sub-linear term ($|x|^\alpha$) takes over, maintaining a high restoring velocity to snap the state to exactly zero.

Because the super-linear term caps the time spent in the far-field regime, the total settling time is guaranteed to never exceed a fixed maximum threshold ($T_{\max}$), no matter how large $x(0)$ is:

$$T \le T_{\max} = \frac{1}{c_1(1-\alpha)} + \frac{1}{c_2(\beta-1)}$$

This allows designers to establish a hard temporal deadline for convergence. For instance, Jorge I. Poveda et al. apply this framework in Fixed-Time Nash Equilibrium Seeking in Non-Cooperative Games to ensure that competitive distributed agents reach an optimal steady state by a user-prescribed deadline without needing to know their opponents' mathematical cost functions.


Prescribed-Time Control and Real-World Constraints

Even with fixed-time control, calculating and tuning the parameters ($c_1, c_2, \alpha, \beta$) to hit an exact, user-specified deadline can be mathematically intensive. To bypass this tuning process, researchers introduced prescribed-time control.

Instead of relying on static power laws, prescribed-time control uses time-varying gains that act as a temporal coordinate transformation. As Farshad Khorrami demonstrates in Prescribed-Time Regulation of Nonlinear Uncertain Systems with Unknown Input Gain and Appended Dynamics, this approach maps a finite-time interval $0, T)$ to an infinite range $[0, \infty)$. By multiplying the feedback loop by a gain that blows up toward infinity as time approaches the deadline (such as $\frac{1}{T-t}$), the controller forces the system to stabilize exactly at $t = T$. This strategy provides direct control over system dynamics, allowing vehicles to maintain precise positioning within a user-defined time horizon ([Bowen Sui et al., Prescribed-Time Dynamic Positioning Control for USV with Lumped Disturbances, Thruster Saturation and Prescribed Performance Constraints).

   Convergence Profiles Compared

   State (x)
     ^
     |  \
     |   \--- Asymptotic (Never truly hits zero)
     |     \
     |      \-- Finite-Time (Hits zero, but depends on x(0))
     |       \
     |        \- Fixed-Time / Prescribed-Time (Hits zero before hard deadline T)
     +------------------------------------------> Time (t)
                                 T

However, forcing a system to zero within a guaranteed window introduces a major engineering trade-off: control effort. As the error approaches zero in a non-smooth system, or as the time approaches the deadline in a prescribed-time system, the required control input can spike dramatically.

In physical hardware, actuators have limits. As Zongyu Zuo highlights in A Bounded-Function-Based Scheme for Finite-Time Stabilization of a NWMR with Input Constraints, if a system faces input saturation, it cannot deliver the infinite slope or massive gains required by pure finite-time math. This mismatch can cause control chattering or instability at the boundaries (Zeyu Guo et al., Global Finite-Time Set Stabilization of Spacecraft Attitude with Disturbances using Second-Order Sliding Mode Control). Modern research in this field focuses heavily on designing bounded, saturation-aware functions that preserve finite-time guarantees without breaking the physical actuators that drive them.

Key Foundational Work

Finite-Time Stability of Continuous Autonomous Systems (2000) — contributed 5 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_finite_time_convergence_f47abe10,
  title = {finite-time convergence},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/finite-time-convergence/rdr_f47abe103cd749ebb746143c102215e5},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_finite_time_convergence_f47abe10_v1,
  title = {finite-time convergence},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/finite-time-convergence/rdr_f47abe103cd749ebb746143c102215e5/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake