A Local Search Based Heuristic
- 79 Downloads
In the last chapter we have studied an algorithm which combines edge-guessing with a truncated branch-and-bound algorithm. This hybrid algorithm yields good solutions in a short amount of time which is mainly due to the quality of the edge orientations derived and less to the ability of truncated branch-and-bound to find good approximations. Thus, it suggests itself to combine edge-guessing with better heuristics in the hope of obtaining better overall results. The currently best heuristics for the JSP are based on local search approaches. We will generalize these approaches for the DSP. Since a general introduction to local search has already been given in section 2.4, we will only discuss the missing details. Section 7.1 introduces neighbourhood structures that are based on the concept of blocks. In section 7.2, we derive formulas for approximating the makespan which will be used for evaluating solutions in a neighbourhood. These components are embedded in a tabu search environment which is described in section 7.3. Section 7.4 discusses how to combine edge-guessing with local search. Computational results for several versions of our tabu search algorithm are presented in section 7.5.
KeywordsLocal Search Tabu Search Critical Path Neighbourhood Structure Longe Path
Unable to display preview. Download preview PDF.