A Comparative Study of Dual Active-Set and Primal-Dual Interior-Point Method

Abstract

The relative performances of two optimization algorithms – dual active-set method (DASM) and primal-dual interior-point method (PDIPM) in the implementation of model predictive control (MPC) in terms of input response, CPU time, and number of iteration have been compared with the help of three case studies including MPC of continuous stirred tank reactor (CSTR) in series. The study shows significant improvement of convergence, CPU time, and cost per iteration using DASM for small-scale problem. Both DASM and PDIPM have been observed to be efficient, as the number of iterations is found to be independent of the problem size. Our study also shows that DASM gives better performance over PDIPM.