A multi-source physiological data-driven method for evaluating the mental workload of operators in remote control environment

Abstract

With the continuous development of Maritime Autonomous Surface Ships (MASS), remote control ship will become into reality in the near future. However, the mental behaviours of remote-control seafarers and their effects on navigation safety remain unclear, particularly in the context of emerging “multi-human multi-ship” operational patterns. In view of this, experimental research for the assessment of mental workload in remote control environment was conducted, addressing the critical gap in understanding the cognitive demands of remote maritime operations. A novel experimental setup is developed to closely simulate inland waterway navigation of varying degrees of difficulty, incorporating realistic challenges such as ferry crossings and buoy avoidance. The design uniquely considers the transition from traditional seafaring to remote operations, providing insights into the behavioural differences between remote operators and onboard crew members. A fusion of heart rate variability and skin conductance were employed to achieve a highly precise classification of mental workload levels through a low-invasive method, which is suitable for continuous monitoring in operational settings. Three machine learning models are employed for classifying the physiological signals into distinct workload levels, with the K-Nearest Neighbours (KNN) achieving an accuracy of 93.1%. The statistical analysis validates the effectiveness of this assessment approach across different workload conditions. The integration of physiological data and machine learning models enables real-time detection of cognitive overload and stress in working condition. These findings provide foundational reference for designing personnel suitability monitoring systems and task allocation strategies in future remote control centres, contributing to maritime safety and operational effectiveness.