On the Complexity of Distance-$d$ Independent Set Reconfiguration

TL;DR

This paper investigates the computational complexity of reconfiguring distance-$d$ independent sets in graphs, revealing complexity dichotomies based on graph classes and parity of $d$, extending known results for $d=2$.

Contribution

It provides a comprehensive complexity classification for the Distance-$d$ Independent Set Reconfiguration problem on various graph classes for all fixed $d \\geq 3$, including new polynomial-time and PSPACE-complete cases.

Findings

D$d$ISR under TJ is in P for even $d$ on chordal graphs.

D$d$ISR under TJ is PSPACE-complete for odd $d$ on chordal graphs.

Complexity dichotomy on split graphs: PSPACE-complete for $d=2$, P for $d=3$ under TS; P for $d=2$, PSPACE-complete for $d=3$ under TJ.

Abstract

For a fixed positive integer $d \geq 2$, a distance-$d$ independent set (D$d$IS) of a graph is a vertex subset whose distance between any two members is at least $d$. Imagine that there is a token placed on each member of a D$d$IS. Two D$d$ISs are adjacent under Token Sliding ($\mathsf{TS}$) if one can be obtained from the other by moving a token from one vertex to one of its unoccupied adjacent vertices. Under Token Jumping ($\mathsf{TJ}$), the target vertex needs not to be adjacent to the original one. The Distance-$d$ Independent Set Reconfiguration (D$d$ISR) problem under $\mathsf{TS}/\mathsf{TJ}$ asks if there is a corresponding sequence of adjacent D$d$ISs that transforms one given D$d$IS into another. The problem for $d = 2$, also known as the Independent Set Reconfiguration problem, has been well-studied in the literature and its computational complexity on several graph classes has been known. In this paper, we study the computational complexity of D$d$ISR on different graphs under $\mathsf{TS}$ and $\mathsf{TJ}$ for any fixed $d \geq 3$. On chordal graphs, we show that D$d$ISR under $\mathsf{TJ}$ is in $\mathtt{P}$ when $d$ is even and $\mathtt{PSPACE}$-complete when $d$ is odd. On split graphs, there is an interesting complexity dichotomy: D$d$ISR is $\mathtt{PSPACE}$-complete for $d = 2$ but in $\mathtt{P}$ for $d=3$ under $\mathsf{TS}$, while under $\mathsf{TJ}$ it is in $\mathtt{P}$ for $d = 2$ but $\mathtt{PSPACE}$-complete for $d = 3$. Additionally, certain well-known hardness results for $d = 2$ on perfect graphs and planar graphs of maximum degree three and bounded bandwidth can be extended for $d \geq 3$.