A Self-Organizing, Ad-hoc, Peer-to-peer, Human Network for Optimal Subterranean Transportation Routing
New Yorkers who want to be somewhere they are not are always in a rush, especially when taking the subway. Observations of a New Yorker giving directions demonstrate that they frequently have more than one route to a given destination, and choosing the optimal one is a non-trivial task.
From the Essex/Delancey station, one can take either the M train or the F train uptown to almost any station in Manhattan along this particular line. However, the trains arrive on different platforms, which are separated by a long stair case. It is impossible to observe both platforms simultaneously, and there is no automated indication of which train will arrive first and therefore reach a given destination fastest.
Commuters have devised a novel solution. Between train arrivals, passengers will cluster around the stairs separating the two platforms. As soon as there is more than one passenger, they will evenly distribute themselves along the stairs, re-calculating in O(1) time whenever a new node is added. When a train arrives on either platform, the nearest passenger, or edge node, signals the next node by glancing slightly in the direction of the rest of the passengers, and the message is similarly propagated throughout the network, allowing everyone to board the first train to arrive.
The result is an optimal, low-latency, low-overhead solution, built on existing hardware, following the primary rule of the subway: never make eye contact or speak to anyone.
And it’s a magical little reminder that New Yorkers, whatever our reputation, are good neighbors. Smart too.
