The maritime pilot at work: Evaluation and use of a time-to-boundary model of mental workload in human-machine systems

People have proven to be flexible and reliable in many control tasks, such as car driving and ship navigation. Much effort has been invested into automating these tasks but the benefits have so far been limited and the problems enormous. Other tasks, such as plant control, where complicated systems are tightly coupled to obtain large volumes of high quality products with very strict production demands show a much higher level of automation. This automation makes control of these complicated systems possible, relieving the controller of many tasks, improving the production quality, and reducing the operators workload. In both cases, in order to fit the work to the operator and to ensure that his capacity is used to the maximum, that system safety is optimal, and that working conditions meet human long term needs, extensive knowledge of operator abilities and limitations is required. In this study the relationship between process characteristics and monitoring behaviour was studied in order to learn more about operator control behaviour. This was done in two situations: a relatively complex process simulator in which the operator had to perform a rather complex and realistic control task, and in the real situation of maritime pilots on board sea ships. The operator's monitoring activity was measured using mental workload measures. The hypothesis was that workload is a linear function of the time-to-contact or time-to-boundary of each of the process variables. An assumption was that the mental orkload is the result of sampling and decision making and is proportional to the frequency of this cycle (sampling and decision making). The technique used to record mental workload was the use of heart-rate, in particular the use of heart-rate variability. The heart-rate is not a constant, but fluctuates about 10% around the mean heart-rate. It is known from literature that an increase of mental workload coincides with a decrease of the heart-rate variability (HRV), and this decrease of HRV was used as an indication of increased workload, which in turn was an indication for the operators monitoring behaviour. This hypothesis was tested using a simulated process (DURESS): a simulation of a hot-water production-plant. The subjects had to produce water of a certain temperature and quantity by carefully adjusting the controls of the simulator. During the experiment the heart-rate was recorded, together with all his control activities and his production performance. The results show that the time-to-boundary (TTB) approach is successful in explaining a large part of the operators monitoring behaviour: the TTB measure correlated well with the HRV during the control phase, which confirms the theory on monitoring behaviour. This means that there is a direct relationship between the time left for the operator to intervene and his sampling frequency. The second experiment was a similar study but now with a real task: maritime pilots doing their normal work. Four Rotterdam pilots participated in an experiment in which they were recorded on video, their heartbeat was recorded, and their voyage was logged on maps, all during their normal work. Twenty-five voyages were recorded, with a large range of types and sizes of ships and destinations. This experiment provided a series of results. The element that was considered most important was the relationship between the TTB that was the result of the fairway layout and the mental workload. The correlation functions between HRV and MTB were largely as predicted by the theory. These correlation functions had maximum at about delta x=-0.5km for all location, meaning that the minimum level of HRV was reached before the TTB had reached its minimum, i.e. the workload precedes the critical moment, or, in other words, the pilot makes a decision about 0.5km before the situation becomes critical. Taking the results of the DURESS experiment and the maritime-pilot experiment together gives very strong support to the theory on the relationship between monitoring and time-to-boundary in a complex control task. This relationship was inferred from the relationship between heart-rate variability (HRV) and the minimum time-to-boundary (MTB). In addition to these main results various conclusions are drawn with respect to the recording techniques used, pilotage, and shore-based radar support.Mechanical Maritime and Materials Engineerin