An Unexpected Role for Visual Feedback in Vehicle Steering

complete the maneuver in a real driving situation. Our Some motor tasks can be completed, quite literally, results reveal that subjects do indeed make the same with our eyes shut. Most people can touch their nose mistakes in a driving simulator when no visual feedback without looking or reach for an object after only a brief is provided and that this reveals something fundamental glance at its location. This distinction leads to one of about how humans steer vehicles. the defining questions of movement control: is inforVarious models of vehicle steering control have been mation gleaned prior to starting the movement suffiadvanced, most of which assume the availability of regucient to complete the task (open loop), or is feedback lar, if not continuous, visual feedback [3–7]. Despite this, about the progress of the movement required (closed a considerable number of studies have demonstrated loop)? One task that has commanded considerable the ability of humans to carry out basic steering maneuinterest in the literature over the years is that of steervers in the absence of such feedback [1, 2]. In real driving ing a vehicle, in particular lane-correction and lanesituations, drivers have to attend to other road users changing tasks. Recent work has suggested that this and interior controls or gauges, and it therefore seems type of task can proceed in a fundamentally open loop reasonable that many common tasks will incorporate a manner [1, 2], with feedback mainly serving to correct certain degree of automation [8]. Indeed, in contrast to minor, accumulating errors. This paper reevaluates our pilot study, Godthelp [1] and Hildreth et al. [2] provide the conclusions of these studies by conducting a new evidence that some subjects can complete an entire set of experiments in a driving simulator. We demonlane correction in complete darkness, consistent with strate that, in fact, drivers rely on regular visual feedmodels of steering control that incorporate some degree back, even during the well-practiced steering task of of planning [3, 5] and revealing the limitations of models lane changing. Without feedback, drivers fail to initiate that make use of constant visual feedback [4, 6, 7]. the return phase of the maneuver, resulting in systemHildreth et al. [2] argue that their results are consistent atic errors in final heading. The results provide new with two possible models of steering control: one coninsight into the control of vehicle heading, suggesting strained by lateral position and steering wheel amplitude that drivers employ a simple policy of “turn and see,” [9], the other, by the pursuit of a virtual target at the with only limited understanding of the relationship becenter of the lane [10]. Both models require drivers to tween steering angle and vehicle heading. estimate the change in relevant, visually perceived vari-