🚀 go-pugleaf

This discussion seems to have gotten embedded in sci.physics, although I
suspect there must be a more appropriate venue elsewhere on the net. The
proceedings of Intelligent Vehicles '92 (sponsored by IEEE Industrial
Electronics Society, co-sponsored by SAE and some other folks) contains a
representative sampling of what is going on. The book VISION-BASED VEHICLE
GUIDANCE (edited by Ichiro Masaki) also contains a good selection of papers.

The reliability requirements for autonomous road following are daunting --
the mean time between fatal accidents is approx. 1 million vehicle-hours.
For a vision based system running at 10 Hz, that translates into an average
of 36,000,000,000 images processed without making a sufficiently serious
error to cause a fatality.  The California PATH people have been looking at
plopping ceramic permanent magnets into the pavement to mark the lane center
and tracking those.  

Which of the competing approaches "wins" from the point of view of
non-technical considerations is unclear. The Feb '91 IEEE Transactions on
Vehicular Technology contains a number of good papers discussing policy
issues, and pointing off to references evaluating palletization vs.
entrainment vs. platooning vs. uniform spacing at moderate headways.