Experimental Investigation of the Usefulness of Operator Aided Optimization in a Simulated Tactical Decision Aiding Task.

摘要

The report describes the implementation and experimental evaluation of two examples of a decision aid for Naval Task Force Commanders called 'operator aided optimization' (OAO). This type of aid enables the user to make inputs to a machine which operates on the inputs and shows problem solutions derived from the inputs. The user and the machine work iteratively until the user is satisfied with the machine solution. Problems for which OAO is appropriate tend to have the following characteristics: (1) Solution space is of high dimensionality; (2) Criterion function is nonlinear, and multi-modal, therefore the machine working alone may only find a local optimum instead of the global optimum. The problem for which the ISC aids were constructed is the selection of (a) an air strike path through a field of ten enemy sensors and (b) aircraft speeds on each leg of the path. Utility of each candidate strike path was computed according to a predetermined utility function; the utility function was nonlinear and multi-modal. One of the ISC-designed aids uses a nonlinear programming (NP) algorithm; the other uses a dynamic programming (DP) algorithm. An experiment was designed by ISC to compare decision performance for both types of aids.