Computational Intelligence Multiobjective Heuristic Search

A large number of problems require the optimization of multiple criteria. Classical multicriteria optimization techniques rely on single criteria optimization algorithms, and hence we are either required to optimize one criterion at a time (under constraints on the others), or we are asked for a single scalar combined optimization function. Solutions to most real-world optimization problems involve a trade-off between multiple conflicting and non-commensurate objectives. Some of the most challenging ones are area-delay trade-off in VLSI synthesis and design space exploration, time-space trade-off in computation, and multi-strategy games. Conventional search techniques are not equipped to handle the partial order state spaces of multiobjective problems since they inherently assume a single scalar objective function. Multiobjective heuristic search techniques have been developed to specifically address multicriteria combinatorial optimization problems. This text describes the multiobjective search model and develops the theoretical foundations of the subject, including complexity results . The fundamental algorithms for three major problem formulation schemes, namely state-space formulations, problem-reduction formulations, and game-tree formulations are developed with the support of illustrative examples. Applications of multiobjective search techniques to synthesis problems in VLSI, and operations research are considered. This text provides a complete picture on contemporary research on multiobjective search, most of which is the contribution of the authors.