We are interested in programming a swarm of molecular robots that can perform self-assembly to form a specific shapes at a specific location. Programming such robot swarms is challenging for two reasons. Firstly, the goal is optimize both the parameters and the structure of chemical reaction networks. Secondly, the search space is both high-dimensional and deceptive. In this paper, we show that MAP-Elites[5], an algorithm that searches for both highperforming and diverse solutions, outperforms previous state-of-the-art optimization methods.