The Mumbai method

When I think of a future where cars are self-driving, the vision my mind conjures up is of interchangeable pods that pick you up whenever you need a ride, and then drop you off at your destination. Essentially it will be like Lyft, but with robot drivers. Or in other words, an extremely granular form of public transportation.

As I watched the traffic weaving through Manhattan this weekend, I started wondering how traffic flow would work in that future world. Would we still have a form of turn-taking roughly equivalent to what is done now with traffic lights?

I remember visiting Mumbai a number of years ago, and being amazed at how traffic worked there. In most parts of the city there were no traffic lights. At intersections cars would just weave through each other, east/west traffic seamlessly flowing through north/south traffic.

Through all of this pedestrians would cross the street and cars would simply steer around them. To my Western eyes it all looked incredibly dangerous, but the entire time I was there I never saw a traffic accident.

Given the fact that self-driving cars will actually form a single cybernetic network, that is simply communicating with itself, it seems to me that the Mumbai method might work very well. If this sort of traffic pattern can be sustained by a sea of human drivers, each needing to guess what the other will do, surely a single self-communicating computer network can do the same — and probably much better.

2 thoughts on “The Mumbai method”

  1. It’d be interesting to compare the overall traffic throughput of the Manhattan (take turns w/lights) vs. the Mumbai method (slow down and weave through).

    Does one method work better than the other in heavy traffic vs. light? If so, what’s the break-even point?

  2. Exactly!! I think the best way to evaluate this might be to create a simulation, and run the two paradigms against each other under varying conditions. It would indeed be fascinating to learn that the optimal strategy varies with traffic density.

    Although that sort of adaptive behavior shouldn’t be completely surprising to anyone who has ever observed a horse changing its gait from a trot to a gallop.

Leave a Reply

Your email address will not be published. Required fields are marked *