A novel dynamic function allocation method in human-machine systems focusing on trigger mechanism and allocation strategy

In view of human factors, increased automation may result in deterioration of situation awareness, unbalanced workload and skill degradation as a cost of human increasingly being 'out-of-the-loop'. The concept of dynamic function allocation (DFA) was suggested to resolve these disadvantages by changing the levels of human/machine's controls over system functions. Despite the importance of function allocation, there are limited methods explicitly working out the challenges of DFA in real-world applications. In order to improve system reliability and reduce 'out-of-the-loop' problems, the present study focuses on the determination of the reallocation times known as the trigger mechanism and the way of reallocating as the allocation strategy. To this end, the prediction of agent reliability is an essential step. To handle uncertainty arising from the subjectivity of analysis and lack of data, the Bayesian network is used to predict system performance. As a case study, the interaction of humans with a distributed control system used in a gas power plant is considered. By applying the proposed DFA method, human situation awareness and system reliability would be improved by approximately 30%, specifically under time pressure, and inappropriate work environment. This approach, as decision support, provides guidance for automation designers to improve an automated system.