No abstract

1234Designers perform many tasks when developing new products and systems, and making decisions may be among the most important of these tasks. The trade space exploration process advocated in this work provides a visual and intuitive approach for formulating and solving single-and multiobjective optimization problems to support design decisionmaking. In this paper, we introduce an advanced sampling method to improve the performance of the visual steering commands that have been developed to explore and navigate the trade space. This method combines speciation and crowding operations used within the Differential Evolution (DE) algorithm to generate new samples near the region of interest. The accuracy and diversity of the resulting samples are compared against simple Monte Carlo sampling as well as the current implementation of the visual steering commands using a suite of test problems and an engineering application. The proposed method substantially increases the efficiency and effectiveness of the sampling process while maintaining diversity within the trade space. INTRODUCTIONDesigners perform many tasks when developing new products and systems, and making decisions may be among the most important of these tasks, given the impact that these decisions ultimately have on the product's or system's cost, performance, time-to-market, etc. These decisions typically involve tradeoffs between competing or conflicting objectives, and many designers employ optimization-based approaches and techniques to try and help them resolve these tradeoffs. Unfortunately, most designers do not really know their preferences when they start this process [1], or perhaps more importantly understand the implications of their preferences until they have been able to evaluate some preliminary design alternatives to form "realistic expectations of what is possible". In fact, Balling [1] has noted that the traditional optimizationbased design process of "1) formulate the design problem, 2) obtain/develop analysis models, and 3) execute an optimization algorithm" often leaves designers unsatisfied with their results.Consequently, we are investigating ways to help designers formulate and solve single-and multi-objective optimization problems in a more visual and intuitive manner. This process, which we refer to as trade space exploration, is an embodiment of the Design by Shopping paradigm advocated by Balling [1]: designers want to be able to "shop" for the best design, to gain intuition about trades, to see what is feasible and what is not, and to learn about their alternatives first before making a decision. Our trade space exploration process combines a multidimensional data visualization tool -the Applied Research Laboratory's Trade Space Visualizer, or ATSV [2] -along with visual steering commands [3,4] to put designers "back-in-theloop" when performing design optimization. For example, designers can now sample new designs near any point or region of interest within the trade space by placing one or more attractors directly w...