Sensory Integration in Simulated and Remote Piloting of Vehicle

Principal Investigator
R. Brent Gillespie (U. of Michigan)

Industry
Micah Steele (John Deere Corp.)

Government
Harry Zywiol (TARDEC)
Kaleb McDowell (U.S. Army Research Lab.)

Student
Kevin Rider (U. of Michigan)

Our objective is to determine the features in a vehicle’s driver interface that are most critical to the development of “overlearned” driving skill, or sensory-motor behavior that has become so automatic that secondary tasks can be undertaken without degrading driving performance. We will determine the relative contribution of visual, haptic, and ride-motion cues to driving skill using a dual task experimental paradigm. We will validate models of sensory integration and open and closed-loop motor behavior operating under limited cognitive resources. Applications for the model include remote piloting of vehicles in addition to traditional driving during conditions of supplemental cognitive and decision-making loads.

A new driver model that incorporates sensory integration will quantify the relative value of multiple sensory channels to driving performance under single and dual motor/cognitive task demands. This will have direct application to remote piloting of unmanned vehicles using visual feedback from on-board cameras without haptic or ride motion feedback. The impact of the missing information display channels will be assessed and sensory substitution will be explored. The results of the proposed work will couple with development and testing of the ARC-supported Virtual Driver that integrates cognitive and physical modeling.

Current understanding holds that, once sufficient practice has been accumulated, driving becomes automatic or “overlearned,” and secondary tasks can be taken up without undue effect on driving performance. Driving, however, is a complex process involving the integration of visual, haptic (from the pedals and steering wheel,) and motion (vestibular and limb proprioception) cues, and the coordinated production of vehicle control commands under conditions that are often novel. While several research programs have investigated dual driving/cognitive task performance, the relative importance of the visual, haptic, and motion sensory inputs to the motor program is not well understood. The degree to which one stimulus can substitute for another during temporary inattention or absence of a stimulus remains unexplored. A control-theoretic modeling approach to driving performance would also predict robustness to unexpected conditions, response times, ability to direct attention as needed, and ability to cope with biodynamic feedthrough. Applications in remote driving under diminished visual and possibly missing haptic or motion input are of prime relevance to future unmanned military systems.

The US Army-TARDEC’s human-rated Ride Motion Simulator (RMS) will be used to selectively provide visual, haptic, and ride-motion cues during driving and combined driving/cognitive tasks. The method of visual occlusion will be used to assess visual demand under conditions involving metered haptic and motion display with and without secondary tasks. In addition to standard driving scenarios, piloting remote vehicles under various display conditions will be tested. Also, the impact of various active safety systems such as biodynamic feedthrough cancellation and attention re-directing schemes will be tested.