43  Stochastic stability

Updated

June 28, 2023

Consider a discrete-time stochastic system that starts from a constant initial condition x0 and evolves according to (1)Xt+1=f(Xt,Wt) where XtRn, {Yt}t1 is a Rd-valued stochastic process on a probabilty space (Ω,F,P). Define Ft=σ(W1:t). Then {Xt}t1 is an {Ft}t1 adapted process. In addition, if {Wt}t1 is an independent process, then {Xt}t1 is Markov.

A point x is said to be an equilibrium of (1) if f(x,y)=x for all wRd. Without loss of generality, we assume that the origin x=0 is an equilibrium.

43.1 Different notions of stability

First we recall some terminology related to almost sure convergence.

  1. A random sequence {Xt}t1 in a sample space Ω converges to a random variable X almost surely if P(ωΩ:limtXt(ω)X(ω)=0)=1.

  2. The convergence is said to be exponentially fast with rate no slower than γ1 for some γ>1 (not dependent on ω) if γtXtX converges almost surely to some Δ0.

  3. Given a set DRn, a random sequence {Xt}t1 is said to converge to D almost surely if P(ωΩ:limtdist(Xt(ω),D)=0)=1, where dist(x,D)=infx~Dxx~.

Definition 43.1 The origin of (1) is said to be:

  1. Stable in probability if limx00P(supt1Xt>ε)=0 for any ε>0.

  2. Asymptotically stable in prbability if it is stable in probability and moreover limx00P(limtXt=0)=1.

  3. Exponentially stable in prbability if for some γ>1 (not dependent on ω), limx00P(limtγtXt=0)=1.

43.2 Sufficient conditions for stochastic stability

Definition 43.2 Given a set QRn containing the origin, the origin of (1) is said to be:

  1. locally a.s. asymptotically stable in Q if starting from x0Q all the sample paths Xt stay in Q for all t1 and converge to origin almost surely.

  2. locally a.s. exponentially stable in Q if it is locally a.s. asymptotically stable and the convergence is exponentially fast.

If the above properties hold for Q=Rn, the system is said to be globally a.s. asymptotically (or exponentially) stable.

Function class K

A continuous function h:[0,a)[0,) is said to belong to class K if it is strictly increasing and h(0)=0.

Theorem 43.1 For the stochastic discrete-time system (1), let {Xt}t1 be Markov.

Let V:RnR be a continuous positive definite and radially unbounded function. Define the level set Qλ:={x:0V(x)<λ} for some λ>0.

Let φ:RnR be a continious function that satisfies φ(x)0 for all xQλ.

Suppose the following property holds: for all xRn, E[V(Xt+1)Xt=x]V(x)φ(x),t1. Then:

  1. For any initial condition x0Qλ, {Xt}t1 converges to D1:={xQλ:φ(x)=0} with probability at least 1V(x0)/λ.

  2. if moreover φ(x) is positive definite on Qλ and there exist two calss K functions h1 and h2 such that h1(x)V(x)h2(x), then x=0 is asymptotically stable in probability.

Under slightly stronger conditions, it is also possible to characterize the rate of convergence.

Theorem 43.2 For the stochastic discrete-time system (1), let {Xt}t1 be Markov.

Let V:RnR be a continous nonnegative function.

Suppose the following condition holds: there exists an α(0,1) such that E[V(Xt+1)Xt=x]V(x)αV(x),t1. Then:

  1. For any initial state x0, V(Xt) almost surely converges to 0 exponentially fast with a rate no slower than 1α.

  2. If moreover V satisfies c1xpV(x)c2xp for some c1,c2,p>0, then x=0 is globally a.s. exponentially stable.

43.3 Weaker sufficient conditions for stochastic stability

Theorem 43.3 For the stochastic discrete-time system (1), let V:RnR be a continuous nonnegative and radially unbounded function. Define the set Qλ:={x:V(x)<λ} for some λ>0.

Suppose the following conditions hold:

  1. For any t such that XtQλ, we have E[V(Xt+1)Ft]V(Xt)0.

  2. There is an integer T1 (not depent on ω) and a continous function φ:RnR that satisfies φ(x)0 for all xQλ such that for any t, E[V(Xt+T)Ft]V(Xt)φ(Xt)

Then, the implications of hold.

The sufficient conditions for exponential stability can be weakened in a similar manner.

Theorem 43.4 Suppose assumptions a) and b) of are satisfied with the inequality of b) strengthened to E[V(Xt+T)Ft]V(Xt)αV(Xt) for some α(0,1).

Then, the implications of hold.

Notes

The material in this section is adapted from Qin et al. ().

close all nutshells