Driving in your lane

One of the stock examples of control is driving a car and using lateral pressure on the steering wheel to maintain its relationship to the margins of your lane of the road. This has been described most often in terms of driving down a highway with disturbances from cross winds. There is a memorable description in the archives of Dag driving and Bill, to make the point experientially (‘viscerally’ as we say) about the nature of control, reaching across from the passenger seat and gently but more and more strongly pulling on the steering wheel, then relaxing the pull. Dag got the point in a memorable way.

Here I introduce evidence which you each can verify in your own experience that we do not control the relationship of the right and left edges of the front of the car to the immediately adjacent margins of the road. Instead, we control the direction of motion toward a centered position some distance ahead along the road.

I still remember vividly the experience when I was first learning to drive. Trying to align the edges of the front fenders with the edges of the road did not work because when I was steering wrong the variables change too rapidly. Aside from over-steering and veering, alarmed attention on the near environment interfered with looking ahead to important things like crossroads, oncoming traffic, and curves in the road. Now, as an experienced driver, you can experimentally shift attention to the immediate relationship of front fenders to road margins. You will find that it feels alarmingly less well controlled.

This is why we cut corners: since the target direction is some distance ahead around the curve, it is natural to aim for that location, using peripheral vision to avoid the inside edge of the curve. If we purposely maintain centeredness in the lane rather than hugging the inside, then it is necessary to focus attention on the nearer margins, where the variables are changing more rapidly, and it feels alarmingly less well controlled.

When we drive around a sharp curve, e.g. on a switchback mountain road, the distance to which we can look ahead is reduced. It feels alarmingly less well controlled. Of course, the more ‘blind’ the curve (e.g. because of the rising terrain on the inside of the curve on a mountain road), the greater the concern about possible oncoming traffic or other road hazards ahead, while your attention is demanded on relationships closer to you.

I wonder if these factors apply to the more general cases in the contentious ‘power law’ discussion.

I wonder if these factors apply to the more general cases in the contentious ‘power law’ discussion.

I don’t have a driver’s licence, so I’ve been asking friends to describe their driving experiences for exactly this reason. It seems that lane following while driving a car could be similar to line following by hand. Warren also suggested this a few years ago (maybe other people on the list?).

There are many interesting papers on the use use of visual information in driving. The earliest I found is by J.J. Gibson (the affordance guy), Gibson, J. J. , & Crooks, L. E. (1938). A theoretical field-analysis of automobile-driving. The American Journal of Psychology, 51, 453–471.

From more recent ones I really liked: van der El, Kasper, et al. “A unifying theory of driver perception and steering control on straight and winding roads.” IEEE Transactions on Human-Machine Systems 50.2 (2019): 165-175.
There is a nice diagram showing the “aim point” where drivers are aiming. There is also a projected position located some distance ahead of the car, depending on the speed of the car. It is something like “where the car would be located if it would continue moving at this speed in T seconds”, and T is something like 0.3s.

You mention the fender-lane alignment method does not feel very good, but this seems to be the method used in Thrun, Sebastian, et al. “Stanley: The robot that won the DARPA Grand Challenge.” Journal of field Robotics 23.9 (2006): 661-692.

In the literature, it seems that experiments with people often involve displaying a road on a computer screen, like in a video game, and they use a wheel to steer along the road, generally with constant speed. They also used blocking parts of the screen, to try to find controlled variables, or what they called “visual cues” or “salient stimuli”.

I’ve been working on a driving analogy model for drawing, inspired by the findings from driving experiments and it works pretty good, but it is not exactly reproducing human speeds yet. One of the controlled variables is the current distance from the path, plus the projected distance - this keeps the cursor close to the line. Using both works better than either one alone. The other cv is the average speed or maximum speed. The outputs of these two systems change the reference point of a lower-level target tracking system.

Very nice to see the literature. Even more, to hear about your modeling.

I wonder how peripheral vision and foveal vision work in parallel. We foveate the projected point down the road. I don’t know, but it seems pretty obvious that a loop from peripheral vision to limbic functions is an important part of being alert for hazards, and that those ‘snap judgement’ limbic processes would redirect foveal vision accordingly–not talking about routine driving now, but very generally. Bringing it back to the driving situation, when I attend to the proximal relation of fenders to road margins rather than the projection (however that works), I find that I’m relying more on peripheral vision taking in the ever-changing configuration of the road, and I’m not controlling the relation of the car-configuration to a foveated part of the visual field. On a curve, foveal vision turn to the inside margin (the right side for a right-hand curve), and the outside of the curve is felt to be more hazardous to approach perhaps not only because centrifugal force is a disturbance I am resisting but also because it is relegated to peripheral vision.