In optimisation problems involving multiple agents (stakeholders) we often want to make sure that the solution is balanced and fair. That is, we want to maximise total utility subject to an upper bound on the statistical dispersion (e.g., spread or the Gini coefficient) of the utility given to different agents, or minimise dispersion subject to some lower bounds on utility. These needs arise in, for example, balancing tardiness in scheduling, unwanted shifts in rostering, and desired resources in resource allocation, or minimising deviation from a baseline in schedule repair, to name a few. These problems are often quite challenging. To solve them efficiently we want to effectively reason about dispersion. Previous work has studied the case where the mean is fixed, but this may not be possible for many problems, e.g., scheduling where total utility depends on the final schedule. In this paper we introduce two log-linear-time dispersion propagators – (a) spread (variance, and indirectly standard deviation) and (b) the Gini coefficient – capable of explaining their propagations, thus allowing effective clause learning solvers to be applied to these problems. Propagators for (a) exist in the literature but do not explain themselves, while propagators for (b) have not been previously studied. We avoid introducing floating-point variables, which are usually not supported by learning solvers, by reasoning about scaled, integer versions of the constraints. We show through experimentation that clause learning can substantially improve the solving of problems where we want to bound dispersion and optimise total utility and vice versa.

Many optimisation problems are of an online—also called dynamic—nature, where new information is expected to arrive and the problem must be resolved in an ongoing fashion to (a) improve or revise previous decisions and (b) take new ones. Typically, building an online decision-making system requires substantial ad-hoc coding to ensure the offline version of the optimisation problem is continually adjusted and resolved. This paper defines a general framework for automatically solving online optimisation problems. This is achieved by extending a model of the offline optimisation problem, from which an online version is automatically constructed, thus requiring no further modelling effort. In doing so, it formalises many of the aspects that arise in online optimisation problems. The same framework can be applied for automatically creating sliding-window solving approaches for problems that have a large time horizon. Experiments show we can automatically create efficient online and sliding-window solutions to optimisation problems.