SIMOPTIMIZATION RESEARCH PHASE I.
Annual rept. 1 May 64-1 May 65,
CALIFORNIA ANALYSIS CENTER INC SANTA MONICA
Pagination or Media Count:
The objective of the SimOptimization research is to develop efficient, economical techniques for locating improved but not necessarily optimum solutions to simulation models where analytical optimization techniques cannot be realistically applied. The three techniques developed are the decentralized gradient approach DGA which is based on dividing the system into decentralized units and, assuming certain independence, simultaneously adjusts all free parameters the linear response surface LRS approach which consists of a sequential variation of each parameter in the vicinity of the final DGA solution and the quadratic response surface QRS approach which makes a quadratic approximation to the cost surface and then locates the optimum solution on that fitted surface. The techniques are designed to be applied sequentially--each being more sensitive to parameter interactions than its predecessor, but also more expensive in terms of computing time. The DGA and LRS techniques were tested using a simple task resource model. The QRS approach remains to be tested. The three techniques, the model, and the test procedures used are described in detail. The report presents complete test results, along with an analysis of the convergence and efficiency of the DGA and LRS techniques based on the empirical data generated. Both DGA and LRS gave very good results in the tests conducted. Author
- Operations Research