Expected First Return Times for Random Walks on Bounded Grids
Nan An
TL;DR
This paper develops a general formula for the expected first return time to a fixed vertex in a finite Markov chain by augmenting the chain with a waiting-room structure, yielding a block form $M=QR0T$ with a sub-stochastic $Q$ and a finite fundamental matrix $N=(I-Q)^{-1}$. The key result expresses the first-return time as $E = U_{o,o} + s N^2 (s')^T$, and the approach enables explicit, closed-form calculations on bounded rectangular grids under different boundary conditions. The authors derive concrete formulas for grid return times under periodic, absorbing-stay, and reflecting boundaries, illustrating how boundary effects alter $E$ and enabling efficient computation.
Abstract
We derive a general formula for computing the expected first return time of a random walk on a finite graph. Using this framework, we calculate the expected first return time in various settings over bounded rectangular grids with different boundary conditions.
