Modeling the Use of In-vehicle Information and Active Warnings in Vehicle Convoys Using the Virtual Driver

Principal Investigators
Omer Tsimhoni, Matthew Reed (U. of Michigan)

Industry
Lenora Hardee (International Truck & Engine Corp.)

Government
Harry Zywiol (TARDEC)
John Lockett (U.S. Army Research Lab.)

Students
Helen Fuller, Siew Jin Leow (U. of Michigan)

Our objective is to apply and extend the Virtual Driver modeling approach to perform a safety analysis of cognitive and physical aspects of driving with new technologies.

In 2007, the ARC supported an empirical study of simulated convoy driving while performing in-vehicle tasks. Convoy driving at high speeds with closely spaced vehicles is a hazardous necessity in current military operating conditions. A large number of crashes with significant loss of life and equipment have occurred during convoy operations. The goal of the experiment was to examine the cognitive and physical consequences of in-vehicle display location on driving performance, task performance, and multitasking strategies. Sixteen young men and women in four groups of physical stature (ranging from short women to tall men) drove the UMTRI driving simulator while operating a simulated in-vehicle task. The task required operating a touch-screen display with a combination of fixed-location menu keys and varying-location keys, requiring additional visual demand. Task performance was collected in terms of the timing of all keys pressed. Driving performance included the simulated vehicle’s position in the lane, driver actions, and distance to the lead vehicle. Video was recorded to allow for glance analysis.

These findings from the 2007 simulator study will be used to tune and validate our integrated cognitive and physical human simulation tool - the Virtual Driver. The Virtual Driver links two leading University of Michigan human-factors engineering tools to create a new modeling paradigm for simulating vehicle operators. The Queuing Network – Model Human Processor uses the well-understood mathematics of queuing networks to simulate perceptual and cognitive information processing. This tool, which has previously been applied successfully to driving simulation, has been linked to the Human Motion Simulation Framework, an interconnected set of algorithms that can predict realistic human motions under high-level control.

The new integrated model provides the ability to evaluate the combined effects of cognitive and physical features of vehicle interior designs on driver performance and workload. The project will result in advancements in the driver model capabilities and the development of stochastic assessment protocols for information systems intended for use during convoy operations. The result can feed into efforts to understand driver workload and make predictions about the safety of new task configurations. In addition to ARC support, both the Army and industry have funded the development of the physical simulation tools that are part of the Virtual Driver. The Army has also supported the establishment of a linkage been the Queuing Network – Model Human Processor (QN-MHP - the cognitive component of the Virtual Driver) and IMPRINT, the U.S. Army’s primary human task analysis tool. These connections and leverage will increase the availability of the research results within the Army and industry.