This document presents exploratory designs for computational experiments, focusing on maximin distance designs. The authors propose a class of designs called maximin Latin hypercube (MmLh) designs, which are constructed using a simulated annealing algorithm. The goal is to find designs that balance the entropy/maximin criterion with good projective properties in each dimension, as guaranteed by Latin hypercubes. The paper discusses the use of a criterion function, $ \phi_p $, which is used to rank designs based on their performance. The algorithm is tested with various values of p, and it is found that smaller values of p lead to greater success in finding optimal designs. The results show that for certain problems, p as small as 5 is sufficient. The paper also presents some results on the properties of the generated designs, including their distribution and geometric characteristics. The authors conclude that MmLh designs are effective for predicting the output of computer models, both when few or many inputs are important. The paper is supported by research from the Applied Mathematical Sciences Research Program of the Office of Energy Research, U.S. Department of Energy.This document presents exploratory designs for computational experiments, focusing on maximin distance designs. The authors propose a class of designs called maximin Latin hypercube (MmLh) designs, which are constructed using a simulated annealing algorithm. The goal is to find designs that balance the entropy/maximin criterion with good projective properties in each dimension, as guaranteed by Latin hypercubes. The paper discusses the use of a criterion function, $ \phi_p $, which is used to rank designs based on their performance. The algorithm is tested with various values of p, and it is found that smaller values of p lead to greater success in finding optimal designs. The results show that for certain problems, p as small as 5 is sufficient. The paper also presents some results on the properties of the generated designs, including their distribution and geometric characteristics. The authors conclude that MmLh designs are effective for predicting the output of computer models, both when few or many inputs are important. The paper is supported by research from the Applied Mathematical Sciences Research Program of the Office of Energy Research, U.S. Department of Energy.