Vector-based pedestrian navigation in cities | Hacker News Reader

Vector-based pedestrian navigation in cities

nature.com

36 pointsbcaulfield10 comments

There seems to be an over-focus in the article on how cognitive cost heuristic used to evaluate trajectories explain why humans choose certain paths. In fact a lot of things go into planning a walk, including at times not planning at all, but walking as the spirit guides you as it were.

For example, on the tendency for people to take longer, non-optimal paths for longer walks, the article states...

The tendency to deviate from the shortest path increases with distance between the origin and destination, which could be due to the increasing complexity of evaluating relatively longer paths, in line with the approximate rationality principle.

What it seems to assume is that people completely pre-plan their walk, or rather attempt to. And that optimizing the path for shortness is the top priority of the walker. But if they simply have an agenda to get from point A to point B, many things can influence the path they choose.

In real life, one may vaguely plan a walk, then at key moments make making various path decisions. And those decisions aren't necessarily about optimizing the path. One might for example not take a street because of construction or some obstruction on the sidewalk, and just continue on to the next path.

And sometimes people may choose a path that goes by a park, or a particularly interesting building, or favorite store, or out of curiosity to know what is down a perhaps unfamiliar street, or even to NOT take an unfamiliar street but to take one they are familiar with.

I'd also add that OSM is not always a good reference for planning walks. I just finished walking every street in my city (done with CityStrides website, 694 streets, around 380 miles) this year, and planning optimal walk routes really wasn't always efficient via mapping data. For example, OSM and such can't take into account simple shortcuts like crossing a railroad track or park to reach another street. Early on I hand-planned each walk to optimize street coverage, but towards the end just started walking with only a general plan in mind for the day's walk.

Some streets are genuinely unpleasant to walk along. I would want to see how certain areas or streets nudge path planning one way or another, as long term mildly repulsive or attractive forces. A street with wider sidewalks or quieter traffic conditions will draw people in. Above average urban noise and hostile street design will push them elsewhere.

I don't think street construction or temporary inconveniences are going to influence the data too much. Full closures are uncommon as it's easy to maintain access for pedestrians (especially to abutting businesses) and pedestrians have infinitely more degrees of freedom than cars. No one's going to backtrack because they have to walk under scaffolding or a jersey barrier carveout in the street.

PDF at https://arxiv.org/abs/2103.07104 (https://arxiv.org/pdf/2103.07104)

> […] All paths were map-matched to the open street map network (www.openstreetmap.org).

OpenStreetMap. Weird sloppiness there.

It may be intentional. Nature journals generally discourage using technical terminology, jargon, and abbreviations, unless there are good reasons for them. In the same spirit, brand names should be avoided as well, especially if the spelling is weird.

This could be an explanation, was the brand name not immediately followed by the domain name of the project. The capitalization also implies a direct reference to OSM. The paper also references several Google services by name.

I think it might even be inaccurate to describe something as "the open street map network", because "the" implies there is only one network of street maps that are open. Alternatives like "All paths were map-matched to the street map network provided by ..." or "All paths were map-matched to an open street map, specifically ..." would probably be more correct if the goal is to avoid the brand name.

I see no reason to discourage brand names if their products are specifically directly to the research in question. Avoiding brands is a lot easier in other fields of science, but even there attributing data sources should be acceptable (or even required).

It looks to me like the this is the result of sloppiness or an overzealous editor.

I've always followed the basic principles that you shouldn't allow brand names to dictate weird capitalization and such (so IOS instead of iOS, Yahoo instead of Yahoo!, and AC/DC instead of actually putting a lightning bolt character between the two types of electrical current), but even then you would either end up with 'OpenStreetMap' or 'Openstreetmap'.

I'm finding it oddly sloppy as well.

> consistently longer than the shortest-distance path computed using the standard Dijkstra algorithm

The algorithm used to compute shortest-distance is entirely irrelevant, unless there's some doubt about the correctness of the algorithm!

> In this Article,

Does Nature CS mandate that article is capitalized? It looks very odd.

Not entirely irrelevant. Without that statement there could be doubts about the correctness of the algorithm used.

How the brain navigates cities - https://news.ycombinator.com/item?id=28917033 - Oct 2021 (35 comments)

Reply on news.ycombinator.com