A challenging problem in symbolic execution is to solve complex mathematical constraints such as constraints that include floating-point variables and transcendental functions. The inability to solve such constraints limit the application scope of symbolic execution. In this paper, we present a new method to solve such complex math constraints. Our method combines two existing: meta-heuristic search and interval solving. Conceptually, the combination explores the synergy of the individual methods to improve constraint solving. We implemented the new method in the CORAL constraint-solving infrastructure, and evaluated its effectiveness on a set of publicly-available software from the aerospace domain. Results indicate that the new method can solve significantly more complex mathematical constraints than previous techniques.