Human–computer cooperative layout design method and its application

Layout design of spacecraft module belongs to scheme design problem, which has been proved to be NP hard. This problem has not only computing complexity but also engineering complexity, and it is more difficult to meet the challenge of practical application in engineering. Inspired by the ideas of “Man–Machine Synergy” and successful design patterns from the modern engineers, this paper proposed a Human–Algorithm–Knowledge-based layout Design method (HAKD) to deal with such complicated problems. Here, the engineers’ successful design patterns mainly include the following aspects: depending on human’s experience and wisdom, consulting some prior knowledge (for example, relevant layout diagrams), and utilizing the available computer technologies, etc. HAKD takes the evolutionary algorithm (e.g. genetic algorithm) as the “algorithm” foundation, in which “human” provides artificial layout schemes (artificial solutions), layout diagrams afford “prior knowledge” solutions, and the evolution algorithm produces novel algorithm solutions. All these numerical solutions are expressed by unified encoding strings, which make up of the evolution algorithm population together, and take part in corresponding evolution operations (such as reproduction, crossover and mutation, etc.). So human’s intelligence, computer’s intelligence (evolution algorithm) and prior knowledge extracted from layout diagrams are fused in the gene level of the evolution algorithm and fully play their own strong suits for the problem solving. In HAKD, the prior knowledge (in this paper, it refers to the geometry characters and position characters of the objects) is recognized and extracted by Hough Transform technology or directly given by commercial CAD software (such as Pro/Engineer) model. A series of experiments on the layout design problem of an international communication satellite module are performed by comparing HAKD with GA, GA+Prior Knowledge and GA+Human. The experiment results show that HAKD outperforms the others.