Optimizing discrete event dynamic systems via the gradient surface method

The authors propose a gradient surface method (GSM) for the optimization of discrete event dynamic systems. GSM combines the advantages of response surface methodology (RSM) and efficient derivative estimation techniques like perturbation analysis (PA) or the likelihood ratio (LR) method. In GSM, the gradient estimation is obtained by PA (or LR), and the performance gradient surface is obtained from observations at various points in a fashion similar to the RSM. Zero points of the successive approximating gradient surface are taken then as the estimates of the optimal solution. GSM is characterized by several attractive features: it is a single run method and more efficient than RSM; it uses at each iteration step the information from all data points rather than just the local gradient; and it tries to capture the global features of the gradient surface and thereby quickly arrives at the vicinity of the optimal solution. A number of examples are exhibited to illustrate this method