Optimal Multimodal Parameter Identification in the State Space Model of the Human Operator.

摘要

A technique is developed which can be used to identify the j-dimensional hypersurface of a multimodal human operator model. The j is equivalent to the number of system parameters plus one for the parameters'performance. The technique uses a bounded random search to select the parameters which are used to calculate an output from the model. Parameters which produce an output that meets the performance criterion are stored;then,they are used as an input to a clustering algorithm. The clustering algorithm produces clusters or groupings of parameters which identify the model's hypersurface from which local maximums can be determined using existing techniques such as Newton-Raphson or gradient search. The local maximum with the best performance is considered the global maximum and the parameters associated with the global maximum are referred to as the optimal set of system parameters. One use of this technique is parameter tracking such as is required in human operator modeling over long periods of time or under changing tasks.