Applying Systematic Local Search to Job Shop Scheduling Problems: Basics, Concepts, and Methods

There has been increasing interest in developing hybrid search methods and in applying them to real-world applications. Systematic Local Search is one of such methods that synthesize desirable aspects of constructive search and local search. The former guarantees to find an optimal solution and to prove it while the latter scales excellently with the problem size. In this work, an instance of Systematic Local Search is defined for the job shop scheduling problem. In particular, a set of activity assignments precluded from an optimal schedule is induced on the precedence relations of the critical path, thus integrating the completeness of constructive search into the hybrid. The effectiveness of this instance is evaluated on benchmark job shop scheduling problems. Experimental results show that Systematic Local Search outperforms heuristic search methods such as simulated annealing and compares favourably with methods designed specifically for job shop scheduling problems. This work should be useful to researchers interested in extending it for other optimization problems as well as to practitioners interested in applying Systematic Local Search to scheduling applications.

Lei Duan, M.Sc.: Computing Science at Simon Fraser University. Ph.D. Candidate at University of Toronto.Dr. William S. Havens, Associate Professor of Computing Science at Simon Fraser University. Chief Scientist at Actenum Corp.Bistra Dilkina, B.Sc.: Computing Science at Simon Fraser University. Ph.D. Candidate at Cornell University.