[shabble]

I've seen one or two pedestrian crossings which included fast-reacting automatic bollards, similar to ones in [1]. They took maybe 4-5s to raise and lower, and afaik were highly effective at protecting the users. The initial capital costs, maintenance, and throughput disruption if they fail-up make them impractical as a general solution, at least for the time being.

[1] http://www.edsuk.com/item.php?id=260