From: John De Armond
Subject: Re: Lane Departure Warning Systems
Date: Mon, 15 Dec 2003 16:35:37 -0500
On Mon, 15 Dec 2003 19:15:06 GMT, "El Alumbrado" <firstname.lastname@example.org>
>I think "Lane Hold" coupled with adaptive cruise control would be the hot
>ticket. My Q45 had "laser-guided adaptive cruise control" that worked
>great. It would slow down if you were gaining on the car in front, then
>speed up when he got out of the way; even flashed a light if you were
>gaining too rapidly and it thought you might rear-end him. Add a GPS and
>we're starting to get close to an "automated driver assistant". It would
>be like having a co-pilot peering over your shoulder drawing your
>attention to things you may have missed("Ahhh, Skipper, watch out for
>that mountain!"), but one who didn't want to stop at every outlet mall
Unfortunately we're a LONG way from a driving assistant. I spent a lot of
time and a bit of money on this project back in the 80s. I was driving about
60 miles one way to work back then and I thought surely I could design a
system that would follow the white line on the interstate. I succeeded. Most
of the time. It was that 0.005% of the time that was the killer. Modern
digital processing is fast enough to address many of the concerns, such as
trapping and neutralizing excursions caused by foreign objects in the camera's
field of view that look like white lines. But when the lines go away, the
system makes an instant transition to requiring manual input. Something that
may not be forthcoming in time if the drive is relaxing/dozing/whatever.
GPS isn't anywhere near accurate enough, even with WAAS. Part of the problem
is the inherent accuracy of the moving GPS fix. The other part is the
inaccuracy of available road maps. Example: I'm currently involved in a
project to map the physical plant of a rural power Co-op. This involves
visiting every pole, transformer, substation and other facility, logging the
GPS coordinates and then creating appropriate maps. One of the things the
utility wants is to be able to designate which poles can be reached with
bucket trucks and which require climbing. That meas we need to place the
poles on the map in close position to the road. In some cases the road is
over 200 ft off the GPS track. We're using both SA2004 and Mapsend. I've
talked to SA and they tell me the maps that come with SA2004 are the best they
In this application a gyro/accelerometer inertial navigation system that trims
what the road sensors pick up is more appropriate. I have an instrument
designed to measure with a high degree of accuracy the dynamics of a vehicle.
It contains a high stability GPS receiver (ovenized oscillator, etc), a 3 axis
ring gyro and a 3 axis accelerometer. I believe that this instrument could do
the job. The only problem is the cost - over $2k when I bought it a year ago.
It's too bad the Intelligent Highways initiative turned into just another sty
for the consultant and political hogs to wallow in. There were several good
ideas floated around that would not have cost much money. Back in the late
80s we did a concept design of a system that would have required little money
and no experimental technology to implement. (IMNSHO)Because my company was
not a Beltway Bandit, we never got funding to take it further. Oh well.
There is a major problem with any of these systems that use the vehicle's
steering and wheels. That is, driver attention. With my experimental system,
I found that my attention wandered within minutes of turning on the control
system. Even concentrating on the equipment and the experiment instead of
daydreaming or something, I found that I quickly lost the ability to instantly
respond to an emergency. I can imagine what it would be like driving more
than a mile or two at a time.
Our system automatically attached a small "spider" to the underside of each
qualified car and took control. Cars entering the intelligent highway under
machine control would as quickly as possible coalesce into "trains" of spiders
in contact with and communicating with each other. At the predesignated exit,
the train breaks apart and the exiting car proceeds down the ramp where the
spider detaches and is recycled to the entrance lane. This system, I think,
addressed most all the problems associated with driver inattention but it did
not address the rather freak problems such as trees/deer/other objects
appearing on the highway. Modern machine vision and wireless networking can
probably now handle that problem.
Oh well, just a little off-topic diversion from the normal political geezer
From: John De Armond
Subject: Re: Lane Departure Warning Systems
Date: Mon, 15 Dec 2003 23:00:00 -0500
On Mon, 15 Dec 2003 22:57:03 GMT, "El Alumbrado" <email@example.com>
>John, I wasn't suggesting "vehicle control" based on GPS, simply
>attention reminders. I've been a pilot for a long time, so I'm used to
>electronic voices telling me to pay attention ("Pull up! Pull up!" is my
That would be similar to an electronic condom that screeches "pull out! Pull
out! when there's a defect....
>I would like to have the little voices telling me "Lane warning!" if I'm
>drifting out of lane, or "Collision!" if it looks like I'm about to
>rear-end somebody. Basically an electronic copilot that's shouts "Hey,
>pay attention!" if it sees something important that I may have missed
>because I was distracted with a cell phone or radio. We already have GPS
>systems that read me directions and warn me that my next turn point is
>coming up. I think all the pieces are there. It's just a matter of
>turning them into an integrated system.
WAAS enabled GPS is sufficiently accurate (obviously, from its purpose) to do
the job. The problem is the maps, the reference the GPS must work against,
are not. A ten foot error in placement for a plane is nothing. A 10 foot
error for a car is a ditch or the other lane. The maps I have to work with
from SA and MapSend (one is supposed to be from one major geodata source and
the other from the other, though I can never keep 'em straight) are doing good
to do +- 200 feet.
GPS could sense the change in direction of a few feet or degrees without any
problem. the problem is, there is no way to know if that movement was a lane
excursion or just the normal path of the road. There has to be a local
reference. This video system that someone referred to has the best chance of
working if it can avoid false triggers.
To consider the effect of false positive alarms, consider the common car
alarm. Do you even turn your head to see where one of those obnoxious blasts
is coming from? I don't. The alarm is almost invariably false. Similarly,
it would only take a few false alarms before the driver either gets used to it
and ignores it or turns the damned thing off.
I think a valid argument can be made that if you can't maintain your lane
while wide awake then you shouldn't be driving. No alarm is likely to have
much effect in that event.
There are some interesting devices to combat drowsiness, the root cause of
most lane excursions, I think. When I was doing long haul (car, not truck)
driving I bought a couple of gadgets that really helped. One looked like a
hearing aid that beeps when the head repeatedly tilts forward at the onset of
drowsiness. The other was a tiny Peltier junction cooler that cooled a spot
between the eyes that is claimed to inhibit drowsiness. I don't know if
there's any science associated with this thing but it sure did work on me. I
ordered it from a japan oh, probably 20 years ago. Never seen it for sale in
the US. I saw the head tilting gadget in a catalog recently. Can't remember
if it was JC whitney, Rat Shack or one of the car ones. That one worked quite
well too, though I didn't need it after getting the Peltier device.
I have read about but have never seen a device that monitors the
micro-movements of the steering wheel and alarms when they slow down. The
claim is that one gets sluggish on the wheel a period of time before one
actually starts dozing. Sounds reasonable.