A Combination of Two Simple Decoding Strategies for the No-wait Job Shop Scheduling Problem

This paper deals with the makespan-minimization job shop scheduling problem (JSSP) with no-wait precedence constraints. The no-wait JSSP is an extension of the well-known JSSP subject to the constraint that once initiated, the operations for any job have to be processed immediately, one after another, until the job's completion. A genetic algorithm with a new decoding strategy to find good quality solutions is proposed. The proposed decoding named as Reverse Decoding (RD) is a simple variant of the standard job-permutation-based representation for this problem. We show, through numerical examples, that by using simultaneously RD with the standard decoding better solutions than when using the direct or the reverse decoding separately are achieved, even under the same computational effort as the combined version. The proposed variant is compared with state-of-the-art approaches on a set of benchmarks. Experimental results on these benchmarks show that the proposed combination also produces competitive results.