Shifting representation search for hybrid flexible flowline problems

This paper considers the hybrid flexible flowline scheduling problem with a set of additional restrictions and generalizations that are common in practice. These include precedence constraints, sequence dependent setup times, time lags, machine eligibility and release times. There are many potential solution representations for this problem, ranging from simple and compact, to more complex and complete. Typically, when choosing the degree of detail of the solution representation, a tradeoff can be found between efficiency of the algorithm and the size of the search space. Several adaptations of existing methods are introduced (memetic algorithm, iterated local search, iterated greedy), as well as a novel algorithm called shifting representation search (SRS). This new method starts with an iterated greedy algorithm applied to a permutation version of the problem and at a given time, switches to an iterated local search on the full search space. As far as we know, this shift of the solution representation is new in the scheduling literature. Experimental results and statistical tests clearly prove the superiority of SRS compared with classical and existing methods.