Evolutionary algorithms for multiple criteria decision making in control

Many problems arising in control and systems engineering require the simultaneous optimisation of multiple, often conflicting, design criteria, such as performance, reliability, environmental impact, and cost. Unlike in single-objective optimisation, the global solution to such problems is seldom a single point, but a family of compromise solutions known as the Pareto-optimal set. These solutions are optimal in the sense that improvement in any objective can only be achieved at the expense of degradation in at least one of the remaining objectives. Here, it is shown that multiobjective genetic algorithms (MOGAs) can be a powerful decision-making aid. It is possible to search for many Pareto-optimal solutions concurrently, while concentrating on relevant regions of the Pareto set. Also, a human decision maker may interactively supply preference information to the algorithm as it runs. Applications of MOGAs being pursued include controller design, model selection in system identification, on-line scheduling, and multidisciplinary optimisation problems. Copyright (C) 1998 IFAC.