View Single Post
Old 06-20-12 | 10:17 AM
  #221  
pallen's Avatar
pallen
Descends like a rock
 
Joined: Oct 2010
Posts: 4,034
Likes: 16
From: Fort Worth, TX

Bikes: Scott Foil, Surly Pacer

Originally Posted by wphamilton
What I'm thinking, from map data you could find things like stoplights or corners at the bottom of a hill, uncontrolled intersections, volume of cross-traffic, business and residential driveways, correlated with gradient. Possibly detect blind curves, narrow shoulders and on-street parking, lane width.

By analyzing cyclist speed changes and correlating with the above you could maybe detect map points where cyclists tend slow or stop, which could point to traffic tendencies from causes external to the specific route. If you could get traffic data from google (such as the real-time traffic conditions we see on some websites), enough of that data could identify map points with tendencies for high speed, high volume, accidents etc.

From elevation data you could probably flag specific street areas as places where a cyclist could easy generate unsafe speeds. It all depends on the quality and sufficiency of data, how well you could match physical characteristics to patterns, and whether you could mash it all into a meaningful metric for potential danger. I don't know if any of that is feasible, but it seems possible in theory.
1. I dont think that level of map detail is easily processed in a formulaic way to asses "danger". The deeper you try to go with analysis, the more you open yourself to lawsuits, IMO because now you claim to know what is dangerous and what isn't.
2. That kind of spatial processing likely takes some serious horsepower and expense. I'm shocked at how fast Strava and process an uploaded ride the way it is. Add this kind of complexity and you probably lose that instant upload and segment matching.

I think they are better off keeping the judgement call on the individuals, but blocking the more obvious dangers - like where its possible to exceed the speed limit. You might throw in certain downhill gradients over a certain distance, but some of those can be relatively safe with no traffic and other favorable conditions.

I think allowing users to flag dangerous segments is sufficient. It only takes one person to pull the segment. Unfortunately, that means I can no longer see my progress on my favorite hill climb. It was flagged a while back, I assume because of the blind corner at the bottom. Its actually one of the safer areas on my commute because there's almost no traffic through there, you just have to not be an idiot when you approach the blind corner.
pallen is offline  
Reply