The Virtual Driver: Integrating physical and cognitive human models to simulate driving with a secondary in-vehicle task.

摘要

Models of human behavior provide insight into people's choices and actions and form the basis of engineering tools for predicting performance and improving interface design. Most human models are either cognitive, focusing on the information processing underlying the decisions made when performing a task, or physical, representing postures and motions used to perform the task. In general, cognitive models contain a highly simplified representation of the physical aspects of a task and are best suited for analysis of tasks with only minor motor components. Physical models require a person experienced with the task and the software to enter detailed information about how and when movements should be made, a process that can be costly, time consuming, and inaccurate. Many tasks have both cognitive and physical components, which may interact in ways that could not be predicted using a cognitive or physical model alone.;This research proposes a solution by combining a cognitive model, the Queuing Network-Model Human Processor, and a physical model, the Human Motion Simulation (HUMOSIM) Framework, to produce an integrated cognitive-physical human model that makes it possible to study complex human-machine interactions. The physical task environment is defined using the HUMOSIM Framework, which communicates relevant information such as movement times and difficulty to the QN-MHP. Action choice and movement sequencing are performed in the QN-MHP. The integrated model's more natural movements, generated by motor commands from the QN-MHP, and more realistic cognitive decisions, made using physical information from the HUMOSIM Framework, make it useful for evaluating different designs for tasks, spaces, systems, and jobs.;The Virtual Driver is the application of the integrated model to driving with an in-vehicle task. A driving simulator experiment was used to tune and evaluate the integrated model. Increasing the visual and physical difficulty of the in-vehicle task affected the resource-sharing strategies drivers used and resulted in deterioration in driving and in-vehicle task performance, especially for shorter drivers.;The Virtual Driver replicates basic driving, in-vehicle task, and resource-sharing behaviors and provides a new way to study driver distraction. The model has applicability to interface design and predictions about staffing requirements and performance.