The Route Not Taken.

I try to come up with ways to facilitate exploring the built environment, but such that we get pulled into the bits and pieces that are typically overlooked, or are simply not considered interesting enough. Everyone comes to see the sights, but what’s in-between can be as interesting, and certainly has its own story to tell.

Thanks to AI, it really has never been easier to implement experimental ideas, and so, when I wanted to try and come up with an approach to identify the streets that are overlooked when asking mapping apps to move from A to B in the city, I figured I could have a shot at exactly that, finding overlooked paths, and put something usable together.

I managed, the results are kind of pretty, as you can see in the screenshots, though I’m still ambivalent about the usefulness of the results.

The process I went with is this:

1. Define a center point.
   The screenshots are set in São Paulo, New York, London, Delft, and Brussels.
2. Define a circle around that point at a certain distance.
   In the screenshots, all circles are at 500 metres.
3. Evenly space points on the circle.
   In the screenshots, all circles have 40 points, except the experiment with 16 points.
4. Calculate the shortest route between the centre and all points on the circle, as well as between opposite points in the circle. The experiment also pairs up all points on the circle.
5. Identify the unused paths inside the circle.
6. Construct a route through some of the unused paths.
   In the screenshots, all constructed routs began and ended in the centre, and were attempted to be 8km long, though because of optimisation of covering unused/used paths, the eventual lengths differed quite a bit. These routes are visible in blue.

The resulting images are pretty, but, knowing the cities I tested this for, the paths not part of the identified routes are not really that unused, defying the purpose of the exercise. I might know these places too well for this process to be useful, but trying the same process in cities I don’t know also kind of defies the purpose of the exercise; if I don’t know the city, every street will be new to me.

Further experimentation could be warranted. Including calculated routes from every point on the circle to every other point on the circle would likely cover the space significantly more. However, calculating routes is computationally intensive and, with 40 points, would increase the total number of routes from 78 to 858. A bit much.
A test using exactly this principle, but with only 16 points on the circle, gives a promising result.

Next, of course, is to put this in practice, in a place I’m not overly familiar with, and learn about the found unused paths.

Note I’m not sharing access to the tool. The number of routes that each attempt has to calculate is significant and, when publicly accessible, runs the risk of seeing usage quickly spiralling out of control, and thus becoming expensive.

•  discovery  mapping  walking