Description:
For most computationally intractable problems there exists no heuristic that is equally effective on all instances. Rather, any given heuristic may do well on some instances but will do worse on others. Indeed, even the 'best' heuristics will be dominated by others on at least some subclasses of instances. It thus seems worthwhile to identify - for each instance anew - a heuristic that is appropriate for the instance at hand, instead of applying always the same algorithm, regardless of its suitability. Adaptive control schemes attempt to do this with iterative algorithms by dynamically exploiting the experience gained in earlier iterations to guide the underlying algorithm in later iterations. In a recent study, several adaptive control schemes have been evaluated on one of the most fundamental scheduling problems, the resource-constrained project scheduling problem (RCPSP). With this contribution, we intend to complement that research by results on a substantially more general problem, viz. the RCPSP under partially renewable resources. This new resource concept allows to formulate a variety of temporal and logical relations between scheduling objects, and is thus of interest in many practical settings pertaining to time-tabling and manpower scheduling.