Comparison of Brownian jump and Brownian bridge resetting in search for Gaussian target on the line and in space

TL;DR

This paper compares the efficiency of three stochastic search processes with resetting mechanisms in finding a Gaussian-distributed target in different dimensions, analyzing their expected search times and optimal parameters.

Contribution

It provides a detailed comparison of Brownian motion, Brownian bridge, and jump processes with resetting, including explicit calculations of optimal parameters and scaling behaviors in various dimensions.

Findings

Optimal parameters scale as σ²/D in 1D and 2D, but as σ³/D in 3D.

Brownian bridge and jump processes have different efficiencies depending on dimension.

The study reveals anomalous scaling in 3D for the expected search time.

Abstract

For $d\ge1$ and $r>0$, let $X^{(d;r)}(\cdot)$ be a $d$-dimensional Brownian motion with diffusion coefficient $D$, equipped with an exponential clock with rate $r$. When the clock rings, the process jumps to the origin and begins anew. For a parameter $T>0$, let $X^{\text{bb},d;T}(\cdot)$ be the process that performs a $d$-dimensional Brownian bridge with diffusion coefficient $D$ and bridge interval $T$, and then at time $T$ starts anew from the origin, and let $X^{d;T}$ be the process that performs a $d$-dimensional Brownian motion with diffusion coefficient $D$ up until time $T$, at which time it jumps to the origin and begins anew. Denote expectations by $E_0^{d;r},E_0^{\text{bb},d;T}$ and $E_0^{d;T}$. These Markov processes with resetting search for a random target $a\in\mathbb{R}^d$ with centered Gaussian distribution of variance $\sigma^2$, denoted by $\mu_{\sigma^2}^{\text{Gauss},d}$. Fix $\epsilon_0>0$. Let $\tau_a$ be the hitting time of $a$, for $d=1$, and the hitting time of the $\epsilon_0$-ball around $a$, for $d\ge2$. The expected time to locate the target for each of the processes is $\int_{\mathbb{R}^d}\big(E_0^*\tau_a\big)\mu_{\sigma^2}^{\text{Gauss},d}(da)$, where $E_0^*$ stands for $E_0^{d;r}, E_0^{\text{bb},d;T}$ or $E_0^{d;T}$. For $d=1$ and $d=3$, we calculate the infimum of each of the above expressions over $r>0$ or $T>0$ as appropriate, in order to compare the relative efficiencies of the three search processes. In terms of the parameters $D$ and $\sigma$, in the 1-dimensional case these infima scale as $\frac{\sigma^2}D$, which is a natural scaling, but in the 3-dimensional case, they scale anomalously as $\frac{\sigma^3}D$. We also show that in the 2-dimensional case, the infimum over $r>0$ for the first of the three search processes scales as $\frac{\sigma^2}D$ as in the 1-dimensional case.