This paper presents a Genetic Algorithm approach to two-dimensional shape optimization. Shapes are represented as arrays of boolean pixels (material/void), or bit-arrays. The inadequacy of the (one-dimensional) bitstring representation is emphasized, both a priori and experimentally. This leads to the design of crossover operators adapted to the two-dimensional representation. Similarly, some non standard mutation operators are introduced and studied. A strategy involving evolutionary choice among these different operators is finally proposed. All experiments are performed on a simple test-problem of Optimum Design, as the computational cost of real-world problems forbids extensive experimental tests.
CITATION STYLE
Kane, C., & Schoenauer, M. (1996). Genetic operators for two-dimentional shape optimization. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1063, pp. 355–369). Springer Verlag. https://doi.org/10.1007/3-540-61108-8_50
Mendeley helps you to discover research relevant for your work.