Software release planning under soft resource and dependency constraints

The goal of incremental software release planning is to assign planning items to future releases so that value is maximized while various resource, technical and other constraints are satisfied. We propose a new approach to this problem called soft-EVOLVE II, in which a project is first modeled using a highly-general model with a nonlinear objective function, and then a set of high-value release plans is generated using a hybrid genetic algorithm. Our formalization naturally extends the established EVOLVE II model by Ruhe et al., but is capable of expressing a much larger variety of constraints between features, in which traditional, Boolean feasibility constraints are replaced entirely by soft constraints. Such a formulation allows for highly-detailed modeling while taking uncertainty into account. We discuss the primary hypotheses and goals of this research, as well as potential validity threats and their control. Keywords—release planning; genetic algorithms; searchbased software engineering