Two heads are better than one: An AIS- and TS-based hybrid strategy for job shop scheduling problems

Abstract

Job shop problems (JSPs) widely exist in many fields and are usually very hard to solve. Despite the fact that many scheduling algorithms have been studied, it is still challenging to find optimal solutions for certain JSPs. Some studies have shown that it is difficult to solve these JSPs by only using a single search technique, while the hybrid of different ones is usually more effective. In this paper, a hybrid algorithm combining artificial immune system (AIS) with tabu search (TS) is proposed. The AIS is based on the clonal selection principle of biological immune systems and is used to find the solution space with potential high evaluation values. The TS is used to exploit the local solution space to further improve the quality of a solution. The neighborhood of the TS is based on a disjunctive graph model, and a smaller neighborhood structure is adopted to reduce the computational cost of the neighborhood search. To reduce the solution space of JSPs and balance convergence speed and solution quality, scheduling solutions are restricted in the set of parameterized active schedules, i.e., a subset of active schedules. Forty-three benchmark instances are used to evaluate the proposed hybrid algorithm, and experimental results are compared with those of other algorithms. Results show that the hybrid algorithm is very effective for JSPs. © 2012 Springer-Verlag London Limited.