Detailed Results for High-Multiplicity RCPSP/max
datasetposted on 2018-11-26, 00:12 authored by Steven EdwardsSteven Edwards
Extended summary of results obtained for the high-multiplicity RCPSP/max.
The instances can be obtained by taking multiples of individual projects from feasible instances of the UBO10 dataset from the Project Scheduling Problem Library (PSPLib) found at the the following website,
The first five columns are identical to that of resultsLess.csv,
- Instance: the corresponding instances from UBO10
-Multiplicity: the number of multiples of the instance that are considered
-UB: the objective function associated with the best known solutions
-LB: the best known lower bound for the instance
-Status: (*) if the UB=LB
The next two columns are as follows
-Canonical UB: The upperbound determined by the (canonical) schedule obtained by repeating the optimal schedule of a single multiple of the project for the required number of multiples.
-Resource LB: The lower bound obtained that ensures the time-horizon is at least big enough such that the resource demands can be met.
The remainder of the file report the Lower and Upper bounds obtained by the methods considered in the corresponding paper. There are five families of approaches (DDT, OOSDDT, Chuffed, Cpx, CpOpt). The approaches are testing with different methods of symmetry breaking indicated as follows;
- 0: symmetry is not removed from the model, i.e., no symmetry breaking is used,
- 1: symmetry is removed through additional precedence constraints
- 2: symmetry is removed through remodelling (only applies to the MIP models).