[catbird]

I don't see any reason why an autonomous car couldn't be trained to move to the left lane before an on ramp, or classify neighboring cars by their observed behaviors (slow reaction time, etc).

Also, I would argue that maintaining a buffer zone in heavy traffic, though it may feel like you are breaking less/going faster, is not actually helpful. The flow rate of a section of highway (cars/hour) is the product of speed (miles per hour) and density (cars/mile). Leaving a space in front of you lowers the density, decreasing overall throughput. Ideally with self driving cars we could have them going at high speeds very close together (or even mechanically coupled like a train).

[moneytide1]

Yes, the train example would be the optimal way to use autonomy. The idea of the buffer is rooted more in energy efficiency, rather than increasing flow rate. But I understand that time may be more valuable than energy in this situation, since we are talking about the throughput of economic components (worker and resource transport).

The idea is that if the buffer were used more frequently, then the average speed could increase. Multiplied by a lower density number, it could potentially be a higher flow rate. It requires more energy and time to move from a complete stop than it does to increase speed while already moving. The buffer could eliminate jams completely, if we define "jam" as standstill traffic.

Merging would be smoother because there wouldn't be stop and go. Think of all the thousands of people in a 5 mile jam stretch that slow to a halt and wave their hand while the other person takes time to look at them and make sure they really do have clearance. Every time that happens, dozens of cars behind must also come to a complete halt. With the buffer, speeds may temporarily slow but never halt when introducing new vehicles into the line. If the buffer was a standard implemented in drivers ed, then all drivers would have a protocol for jams.