A Comparison of Parametric Approximation Techniques to Continuous-Time Stochastic Dynamic Programming Problems

We compare three parametric techniques to approximate Hamilton-Jacobi-Bellman equations via unidimensional and multidimensional problems. The linear programming technique is very efficient for unidimensional problems and offers a balance of speed and accuracy for multidimensional problems. A comparable projection technique is shown to be slow, but has stable accuracy, whereas a perturbation technique has the least accuracy although its speed suffers least from the curse of dimensionality. The linear programming technique is also shown to be suitable for problems in resource management, including applications to biosecurity and marine reserve design.