Surrogate Optimization

Surrogate optimization is an optimization technique suitable for objective functions given by simulations, ODEs, PDEs, or other time-consuming computations, including black box models. It attempts to find a global optimum while using few objective function evaluations.

The surrogateopt function in Global Optimization Toolbox™ builds and optimizes a surrogate model in place of the expensive function. This derivative-free method can be applied to nonconvex problems. 

The problem solved in the video is to determine the optimal position and angle of a cannon to fire a projectile as far as possible over a wall. Evaluating this objective function requires solving an ODE. A nonlinear constraint is implemented as part of the objective with a penalty function. Alternatively, nonlinear constraints can be specified as separate nonlinear inequality constraints. The example is available here.