Parameterized Complexity of Scheduling Chains of Jobs with Delays

TL;DR

This paper investigates the parameterized complexity of scheduling chains of jobs with delays, revealing hardness results and algorithms for various parameters and delay representations, advancing understanding of computational limits in scheduling problems.

Contribution

It provides a comprehensive complexity classification for scheduling chains with delays, including hardness results and dynamic programming algorithms for different parameters and delay encodings.

Findings

Scheduling with exact unary delays is W[t]-hard for all t when parameterized by thickness.

Scheduling with a small number of chains is W[1]-complete for a single or constant number of machines.

Scheduling with minimum delays is W[1]-hard for fixed machines and W[2]-hard for variable machines.

Abstract

In this paper, we consider the parameterized complexity of the following scheduling problem. We must schedule a number of jobs on $m$ machines, where each job has unit length, and the graph of precedence constraints consists of a set of chains. Each precedence constraint is labelled with an integer that denotes the exact (or minimum) delay between the jobs. We study different cases; delays can be given in unary and in binary, and the case that we have a single machine is discussed separately. We consider the complexity of this problem parameterized by the number of chains, and by the thickness of the instance, which is the maximum number of chains whose intervals between release date and deadline overlap. We show that this scheduling problem with exact delays in unary is $W[t]$-hard for all $t$, when parameterized by the thickness, even when we have a single machine ($m = 1$). When parameterized by the number of chains, this problem is $W[1]$-complete when we have a single or a constant number of machines, and $W[2]$-complete when the number of machines is a variable. The problem with minimum delays, given in unary, parameterized by the number of chains (and as a simple corollary, also when parameterized by the thickness) is $W[1]$-hard for a single or a constant number of machines, and $W[2]$-hard when the number of machines is variable. With a dynamic programming algorithm, one can show membership in XP for exact and minimum delays in unary, for any number of machines, when parameterized by thickness or number of chains. For a single machine, with exact delays in binary, parameterized by the number of chains, membership in XP can be shown with branching and solving a system of difference constraints. For all other cases for delays in binary, membership in XP is open.