Asymptotic properties of a bold random walk

TL;DR

This paper studies the long-term behavior of a non-Markovian random walk model with memory of maximum distance, focusing on the super-diffusive case where the walk is 'bold' and develops new probabilistic tools for analysis.

Contribution

It provides the first detailed asymptotic analysis of the 'bold' non-ballistic regime using novel martingale-based probabilistic methods.

Findings

Established asymptotic properties of the bold walk in the non-ballistic region.

Developed new martingale techniques for analyzing non-Markovian random walks.

Connected the walk's behavior to a continuous parameter controlling diffusion type.

Abstract

In a recent paper we proposed a non-Markovian random walk model with memory of the maximum distance ever reached from the starting point (home). The behavior of the walker is at variance with respect to the simple symmetric random walk (SSRW) only when she is at this maximum distance, where, having the choice to move either farther or closer, she decides with different probabilities. If the probability of a forward step is higher then the probability of a backward step, the walker is bold and her behavior turns out to be super-diffusive, otherwise she is timorous and her behavior turns out to be sub-diffusive. The scaling behavior vary continuously from sub-diffusive (timorous) to super-diffusive (bold) according to a single parameter $\gamma \in R$. We investigate here the asymptotic properties of the bold case in the non ballistic region $\gamma \in [0,1/2]$, a problem which was left partially unsolved in \cite{S}. The exact results proved in this paper require new probabilistic tools which rely on the construction of appropriate martingales of the random walk and its hitting times.