Herramienta de Optimización Multiobjetvo Basada en pREferencias

Solving an optimisation problem means finding the value that the variables should take to make the objective function optimal while satisfying a set of constraints. But most real engineering problems require the simultaneous optimisation of multiple criteria, some of which are opposing, so that different and conflicting objectives must be simultaneously satisfied. Decision-making processes in companies are characterised by the interplay of many variables and multiple decision criteria. Multi-objective optimisation techniques have been developed to address these problems and provide analysts with useful tools for making better decisions. Desirable features in these tools are flexibility, ease of problem modelling, and the ability to add new elements or constraints if more model detail is needed.
The simplest tools used to solve multi-objective and multi-criteria problems of optimisation combine multiple objectives into a cost function. This combination of objectives is usually weighted for each of the various objectives. The choice of these weightings is clearly arbitrary and difficult to establish a priori. Therefore, the solutions produced reflect this arbitrariness and are generally sensitive to the values assigned to the weights.
HOMBRE (an acronym in Spanish for multiobjective optimisation tool based on design preferences) aims to handle multi-objective optimisation problems by incorporating information directly from the problem through a range of design preferences expressed with linguistic labels (desirable, tolerable, undesirable, and so on) for the goals. The aim is to avoid setting meaningless weightings and guide the optimiser according to the preferences expressed. The scope of this technique is extremely varied as these problems arise in every area and sector from the macro to microeconomic level: energy planning, production planning; structural design; part design; engineering (electrical, control, mechanics, etc.); telecommunications; and the petrochemical industry.
Design preferences:
Each objective is assigned a preference (minimise, maximise, include a range, take a value, and so on). Each objective is then shown in its own units and range of preferences (the designer specifies which values are good, average, poor, and bad). This information is supplied to the optimiser (if the designer wishes to calculate a solution according to his or her preferences), or the information is supplied to the decision support module (if the aim is to analysis how the solutions to the multi-objective problem fit the preferences).