A new solution encoding for simulation-based multi-objective workforce qualification optimization

Solutions for combinatorial problems can be represented by simple encodings, e.g. vectors of binary or integer values or permutations. For such encodings, various specialized operators have been proposed and implemented. In workforce qualification optimization, qualification matrices can for example be encoded in the form of binary vectors. Though simple, this encoding is rather general and existing operators might not work too well considering the genotype is a binary vector, whereas the phenotype is a qualification matrix. Therefore, a new solution encoding that assigns a number of workers to qualification groups is implemented. By conducting experiments with NSGA-II and the newly developed encoding, we show that having an appropriate mapping between genotype and phenotype, as well as more specialized genetic operators, helps the overall multiobjective search process. Solutions found using the specialized encoding mostly dominate the ones found using a binary vector encoding.