Asymptotic stability

Asymptotic stability is a stronger form of Lyapunov stability in which a system not only remains near an equilibrium when perturbed, but actually returns to it over time. Formally, an equilibrium point x* is asymptotically stable if it is Lyapunov stable and if there exists a neighborhood of x* such that every trajectory starting in that neighborhood converges to x* as time approaches infinity.

The distinction between Lyapunov stability and asymptotic stability is not merely technical; it separates systems that merely 'do not fall over' from systems that 'self-correct.' A pendulum with friction is asymptotically stable: it returns to rest. A frictionless pendulum is Lyapunov stable but not asymptotically stable: it oscillates forever at the same amplitude. In engineering, asymptotic stability is almost always the desired property; Lyapunov stability alone is usually insufficient.

The rate of convergence matters. Exponential stability — a subclass of asymptotic stability in which convergence occurs at a guaranteed exponential rate — is the gold standard for control design because it provides finite-time performance bounds. Global asymptotic stability extends the convergence property to the entire state space, not just a neighborhood, and is correspondingly harder to prove. Most physical systems are only locally asymptotically stable, and the boundaries of the basin of attraction are where the real engineering lives.