Perturbation In Local Search For Scheduling

Local search is an established meta heuristic for scheduling problems. To avoid of getting stucked in a local optimum, the achieved solution is destroyed. Usually, this is done randomly. Here, two problem specific procedures are suggested. One focuses on jobs with the highest difference between the actual processing time and the net processing time of the job and the other one tries to remove jobs which causes the highest idle time for themselves or other jobs. Comprehensive simulations of test scheduling problems as well as a real world application show that both problem specific procedures outperform a random procedure.