计算机实验指借助数学模型及计算机程序来模拟真实物理、化学过程的仿真实验。为了在有限的实验次数内获得更加充分的信息，需要进行实验设计。本项目研究适合计算机实验模型预测及统计推断的实验设计方法，主要研究问题包括基于格子点的最大最小距离拉丁超方设计生成方法，兼具优良分离距离性质与多维投影均匀性质的设计生成方法，适用于混合了连续、定序、定类变量的计算机实验的设计生成方法，以及适用于高斯过程模型的计算机实验自适应设计方法。有别于当前文献中通常使用的最优化方法，本项目拟采用构造为主辅以优化步骤的思路，利用格子点的距离性质生成设计。

Computer experiments are simulations that are based on mathematical models and computer codes. Experimental designs are necessary for computer experiments to obtain sufficient information from limited experimental trials. This project aims at developing experimental designs that are desirable for prediction and inference of computer experiments. Major research questions include construction of maximin distance Latin hypercube designs, construction of designs with both excellent separation distance and multivariate projective uniformity, designs for computer experiments with mixed continuous, ordinal and categorical variables, and adaptive designs of Gaussian process emulation of computer experiments. Unlike most existing methods which are numerical optimization algorithms, in this project we plan to develop constructional methods that exploit distance properties of lattices.