Concurrent model synchronisation with multiple objectives

ABSTRACTConcurrent model synchronisation, i.e. the (bidirectional) propagation of updates between two models, is an important problem in the area of model-driven engineering (MDE). Compared to other consistency management tasks, synchronising concurrent updates is especially challenging as they can be conflicting, such that restoring a consistent state is not possible when all updates must be considered. Recent approaches create a search space of possible solutions and determine the optimum solution via exact methods, such as integer linear programming (ILP), via a configurable, scalarised objective function that takes conflicting goals into account. However, the determination of suitable configuration parameters and runtime efficiency improvements are still an open issue, which is commonly addressed by using heuristics instead of exact methods. We investigate on whether it is beneficial to apply heuristics to solve concurrent model synchronisation problems. First, a multiobjective evolutionary algorithm is used for small instances for which all pareto-optimal solutions can be presented to a user to select the best one. Second, for larger models, we propose a method to determine suitable weightings for aggregating all objectives into a single function. Finally, these insights are used to recommend a strategy for determining solutions of satisfying quality within an acceptable amount of time.