Use of neural networks to achieve dynamic task allocation: a flexible manufacturing system example

To attain optimum performance of the automated system, task allocation betw een human and computer becomes very important. However, a critical problem existing in the technology of dynamic task allocation is how to develop an implicit human-computer communication interface. Two models of 'neural netw ork' and 'predictive method' are proposed in this study to allocate the tas k between the human and the computer. The first phase in this study was to find some important and sensitive indexes to measure the mental workload in supervisory task through the multiple regression equation. The second phas e of this study was to construct a programming system in an FMS to evaluate the workload index and allocate the task dynamically through the applicati on of the back propagation network (BPN) and the predictive values of the m ultiple regression equation. Twenty-two subjects attended the experiment an d were divided into two groups, one was the dynamic group and the other was the static group. The result showed that the workload of the dynamic group was significantly lower than the static group (p-value = 0.0426 < alpha = 0.05). The neural network proved to be an effective method for decreasing t he mental workload through dynamic task allocation.