Sensorimotor basis of motor vehicle control

Abstract

Driving represents one of the most common modes of transport world-wide. It is also one of the major causes of death and injury in developed societies prompting technological innovations through which computers share or assume control of the task of driving, through a range of Advanced Driver Assisting Systems (ADAS). Given the lengthy and substantial investment in such systems it is perhaps surprising how little we understand about how humans control motor vehicles and why, on occasion, this control fails. This is actually a problem. Driver-assist systems that do not understand human drivers well can be dangerous and promote unforeseen risks. For example, anti-lock braking systems, which are designed to prevent skidding during heavy braking, do not always have the positive impact on safety one might expect (Farmer, Lund et al. 1997, Sagberg, Fosser et al. 1997, perhaps because their operation does not mesh seamlessly with the expectations of drivers or other road users.One route to gaining a better understanding of the control processes employed by drivers, is through the study of situations in which standard steering strategies fail. In this thesis the In this thesis, we investigated this effect in a series of experiments aimed at getting to the bottom of the causes of this error. The thesis begins by seeking to generalize the effect from the special case of a straight road which was the focus of earlier studies. This was deemed important because a typical steering wheel has a natural tendency to re-centre itself. It is possible that this behaviour reduces the active steering movements a driver must make during a lane change, since they do not have to actively return the steering-wheel to the neutral position. For that reason, we designed a circular road on which a non-zero steering wheel angle was required at all times. Through this study, we were able to conclude that the effect does generalize.In a second set of experiments we attempted to investigate which visual cues are essential for drivers to correct their error. Apparently, a normal road with redundant information provides sufficient visual feedback for drivers to make a lane change. However, it is of interest to know to what extent the visual feedback can be reduced and still meet the minimum requirement. In our study, we chose optic flow as a starting point. Optic flow has previously been shown to suffice for the purpose of heading perception and heading control.A number of 'steering-towards-a-target' studies also found that optic flow is tightly related to ii steering performance. Hence, we asked whether optic flow was sufficient for controlling a lane change manoeuvre and, unlike much of the previous studies on flow, we posed the question in an active steering task, revealing that flow alone is not sufficient to prompt correct lane changing behaviour.In the concluding set of experiments we took our studies out into the field. Most previous studies on lane changing have been conducted in driving simulators. Simulators are limited in that th...