The estimation of the lower confidence limit of a benchmark dose (BMD) is a common objective in environmental risk analysis. In a BMD estimation study, measurements are taken at different dose levels for a pollutant of interest. These dose levels need to be selected in a controlled experiment, which is a design problem. However, few design strategies have been developed for BMD estimation, especially with consideration of model uncertainty and misspecification. We propose a weighted c-efficiency criterion to obtain a design, which is robust with respect to various risk functions simultaneously for BMD estimation. In simulation studies using particle swarm optimization (PSO) and 8 equally weighted possible models, efficient 3-point designs are found. The proposed method for identifying robust designs is also demonstrated through the mammalian carcinogenic of cumene (C9H12) example.