Driving requires careful orchestration of attentional resources between multiple tasks. In this paper the authors present the building blocks of a computational framework for drivers' attention management and task scheduling and indicate how they interact. The authors adopt a three layer hierarchical structure in which each layer consists of a satisficing decision maker and a set of dynamic mental models that communicate with the decision makers. The three layers characterise decision making from a strategic, tactical and operational point of view. The mental models provide the predictive and evaluatory assessments used by the satisficing decision maker to derive an adequate solution (i.e., strategy, decision, action). Driving tasks are categorised into several domains each associated with a mental model. At the lowest level, where interaction with the external world takes place, the mental models are tied to a set of low level perception/action processes from which one or more depending on the situation are activated to accomplish the task. Particular focus is directed to the middle layer where attention management and task scheduling take place. The authors provide a framework for how the information provided by mental models is used to direct and schedule attention to the various task domains. (A)
Abstract