A COMPARISON OF COLLOCATION METHODS FOR SOLVING DYNAMIC OPTIMIZATION PROBLEMS

This paper describes and compares a global end-point collocation method with the SOCOLL method. The effects of the number of interior collocation points (N) and the collocation methods themselves are investigated using an example. A second example shows the effectiveness of the endpoint collocation method with a state constrained problem. For the first example, the overall deviation between the approximate state profile [generated via the nonlinear program (NLP)] and the true state profile (generated via simulating the original model with the optimal control policy) was smaller for the end-point collocation method. The results also showed a nonuniform improvement with respect to increasing N. For the second example, the end-point collocation method (N = 5) gave better results than those found in the literature.