Discrete approximations to reflected Brownian motion

DOI10.1214/009117907000000240MaRDI QIDQ2482284zbMATH OpenOpenAlexFDO

Publication date 16 April 2008

Published in The Annals of Probability (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/math/0611114

conditioning Dirichlet form killed Brownian motion martingale random walk reflected Brownian motion Skorokhod space tightness

Mathematics Subject Classification ID

Dirichlet forms (31C25) Functional limit theorems; invariance principles (60F17) Markov chains (discrete-time Markov processes on discrete state spaces) (60J10) Diffusion processes (60J60)

Abstract: In this paper we investigate three discrete or semi-discrete approximation schemes for reflected Brownian motion on bounded Euclidean domains. For a class of bounded domains

D

in

m a t h b b R^{n}

that includes all bounded Lipschitz domains and the von Koch snowflake domain, we show that the laws of both discrete and continuous time simple random walks on

D c a p 2^{- k} m a t h b b Z^{n}

moving at the rate

2^{- 2 k}

with stationary initial distribution converge weakly in the space

m a t h b f D ([0, 1], m a t h b b R^{n})

, equipped with the Skorokhod topology, to the law of the stationary reflected Brownian motion on

D

. We further show that the following ``myopic conditioning algorithm generates, in the limit, a reflected Brownian motion on any bounded domain $D$ . For every integer $k g e q 1$ , let $X_{j 2^{- k}}^{k}, j = 0, 1, 2, ...$ be a discrete time Markov chain with one-step transition probabilities being the same as those for the Brownian motion in $D$ conditioned not to exit $D$ before time $2^{- k}$ . We prove that the laws of $X^{k}$ converge to that of the reflected Brownian motion on $D$ . These approximation schemes give not only new ways of constructing reflected Brownian motion but also implementable algorithms to simulate reflected Brownian motion.

Review text

Reflected Brownian motion (RBM) in $D\in R^n,\;n\geq 1,$ considered here is a continuous Markov process $Y$ taking values in $\bar{D}$ that behaves like a Brownian motion (BM) in $R^n$ when $Y_t\in D$ and is instantaneously pushed back along the inward normal direction when $Y_t\in \partial D.$ The authors of this paper are interested to construct RBM in a domain with the so-called nonsmooth boudaries $\partial D$ and for that consider processes approximating RBM in $D$ that are defined on the same state space $D,$ or a discrete subspace of $D.$ They investigate three discrete or semi-discrete approximation schemes for RBM which give not only new ways of constructing RBM but also implementable algorithms to simulate it. The first two approximations involve random walks on $D_k$ which is the connected component of $D\cap 2^{-k}\mathbb{Z}^n$ for bounded domain $D\in R^n,\;0\in D,$ whose boundary $\partial D$ has zero Lebesgue measure. Let $X^k$ and $Y^k$ be the discrete and continuous time simple random walks on $D_k$ moving at the rate $2^{-2k}$ with stationary initial distribution $m_k,$ respectively, where $m_k(x)=\frac{v_k(x)}{2n}2^{-kn},\;v_k(x)-$is the degree of a vertex $x\in D_k.$ It was shown in the paper that the laws of both $X=\{X^k,k\geq 1\}$ and $Y=\{Y^k,k\geq 1\}$ are tight in the Skorokhod space $\mathbf{D}([0,\infty),\mathbb{R}^n)$ of right continuous functions having left limits and if $D$ satisfies an additional condition, which is satisfied by all bounded Lipschitz or bounded uniform domains, then both $X$ and $Y$ converge weakly to the stationary RBM on $D$ in the Skorokhod space $\mathbf{D}([0,1),\mathbb{R}^n).$ The third scheme is concerned with the so-called ``myopic conditioning. One say that a Markov process is conditioned in a myopic way if it is conditioned not to hit the boundary for a very short period of rime, say, $2^{-k}$ units of time, where $k$ is large, and if this conditioning step is repeated over and over again. A more precise description of myopic conditioning of BM is the following. For every integer $k\geq 1,$ let $\{Z_{j2^{-k}},j=0,1,2...\}$ be a discrete time Markov chain with one-step transition probabilities being the same as those for the BM in $D$ conditioned not to exit $D$ before time $2^{-k}.$ The process $Z_t^k$ can be defined for $t\in [(j-1)2_{-k},j2^{-k}]$ either as the conditional BM going from $Z_{(j-1)2^{-k}}$ to $Z_{j2^{-k}}$ without leaving the domain $D$ or as a linear interpolation between $Z_{(j-1)2^{-k}}$ and $Z_{j2^{-k}}.$ In the last main statement of the paper it is proved that for any domain $D$, the laws of $Z^k$ (defined in either way) converge to that of the RBM on $D.$ One may point out also that in the introduction the authors give a brief review of RBM on nonsmooth domains, followed by a breaf description of the approach of this paper to establish discrete approximations of RBM.

Recommendations

Cites work

Cited in

(26)

This page was built for publication: Discrete approximations to reflected Brownian motion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2482284)