| Reported here are a series of five experiments that investigate the role of visual inputs, inertial inputs, and internal models in the control of steering. For these experiments, I utilized a new research tool: an electric vehicle outfitted with a portable virtual reality system. This vehicle allowed me to manipulate visual information with the virtual reality system, the inertial information by feeding steering inputs through an actuator to affect the real physical movement of the scooter, and the internal model of the vehicles dynamics by changing the steering gain (relationship between the steering wheel angle and the vehicle's turning response). Here, I show that valid inertial information (i.e. vestibular and somatosensory cues) enables accurate steering behavior when vision is absent during a lane change maneuver. However, if given ample opportunity to form an internal model of vehicle dynamics, open loop curve negotiation performance with this internal model matches performance with visual information, inertial information, or even both available. As drivers accrue experience with new steering gains, they become more accurate and fluent in steering the vehicle. Adaptation to subtle steering gain changes is rapid, complete, and without aftereffects, but adaptation to a more extreme steering gain (i.e. reverse) is slower, with aftereffects initially upon return to the normal gain. Conflicting inertial information can eventually be ignored in most cases when steering curving paths. An exception occurs when inertial information indicates physical movement that is opposite to visual movement. None of the visual/inertial conflicts administered here, however, produce significant aftereffects. |
