Options
Local search heuristics for the job shop
Author
Lim, Sharon Kah Yee
Supervisor
Shutler, Paul
Abstract
In the job-shop scheduling problem, a vast majority of schedules are infeasible, thus local search in the space of schedules requires more effort to stay inside the feasible region. A new representation of feasible schedules are priority lists, where a priority list is defined as an ordered list in which each job appears as many times as the number of its operations. Local search in the space of priority lists turn out to be simpler and faster as there is no infeasibility problem.
The huge space degeneracy of priority lists suggests that local search in the space of priority lists is ineffective. However, we find that the neighbourhood degeneracy is very small. We begin with the simplest iterative descent on priority lists, and find when coupled to small random sampling in the neighbourhood the results are comparable with the current leading schedule-based simulated annealing.
We then extend to tabu search on priority lists, but the results are disappointing because of the degeneracy. Finally by altering the sampling ratio in the course of the heuristic, we obtain a conceptually simple priority list-based pseudo-simulated annealing heuristic which outperforms the current leading schedule-based simulated annealing and tabu search heuristics.
The huge space degeneracy of priority lists suggests that local search in the space of priority lists is ineffective. However, we find that the neighbourhood degeneracy is very small. We begin with the simplest iterative descent on priority lists, and find when coupled to small random sampling in the neighbourhood the results are comparable with the current leading schedule-based simulated annealing.
We then extend to tabu search on priority lists, but the results are disappointing because of the degeneracy. Finally by altering the sampling ratio in the course of the heuristic, we obtain a conceptually simple priority list-based pseudo-simulated annealing heuristic which outperforms the current leading schedule-based simulated annealing and tabu search heuristics.
Date Issued
2002
Call Number
TS157.5 Lim
Date Submitted
2002