An explorative context-aware machine learning approach to reducing human fatigue risk of traffic control operators

Abstract

Traffic control operators are usually confronted with a high potential of human fatigue. Existing strategies to manage human fatigue in transportation are primarily by undertaking prescriptive “hours-of-work” regulations. However, these regulations lack certain flexibility and fail to consider dynamic fatigue-inducing factors in the context. To fill this gap, this study makes an explorative first step towards an improved approach for managing human fatigue. First, a fatigue causal network that can adequately represent the context factors and their dynamic interactions of human fatigue is proposed. Moreover, to overcome its problem of high dimension sparse matrix, a novel method based on the artificial immune system and extreme gradient boosting algorithm is introduced. A case study of vessel traffic management showed that the model could predict the fatigue level with high accuracy of 89%. Furthermore, to lower the risk of fatigue occurrence, a novel scheduling algorithm is also provided to adaptively arrange work for operators considering individual differences and work types. The study results showed that 27% of operators could be rearranged to reduce the possibility of human fatigue. Nevertheless, considering that more than half of operator were still fatigue in the case study, human fatigue is still a critical problem. It is hoped this research, as an explorative study, can offer insightful references to traffic management authorities in their safety management process with better operation experience.