To reduce traffic, ditch yellow lights and form platoons of self-driving cars | Hacker News Reader

To reduce traffic, ditch yellow lights and form platoons of self-driving cars

arstechnica.com

18 pointsbpierre11 comments

I read the original article discussed by Ars here this morning, and the authors note that they didn't consider the scenario in which multiple intersections exist in a connected roadway network. Also missing was any discussion of pedestrians, bicycles, or non-self-driving cars. This is interesting work, and may have applications in some limited settings, but I wouldn't bank on seeing it too soon even after self-driving vehicles are commonplace.

Also, as a planner, I feel obliged to point out that the title "To reduce traffic" can never be true. In a best case scenario this would allow more vehicles to use the same amount of roadway resource--i.e. increase traffic. As for congestion, that's widely accepted to be an equilibrium state. So if these intersections were successful at improving roadway efficiency, we'd expect to see the additional capacity eaten up by increased demand.

I used to study traffic a lot, and it's definitely true that eating additional capacity leads to more gridlock assuming the same type of driver.

However, consider the freeway case as a starting point. Freeways lock up because of congestion resulting from too many people merging on, causing people to plan poorly/brake suddenly, which leads to a recursive accordion effect, which leads to 5-10 mph stop and go for hours.

However, the actual throughput of cars past any one point on the freeway is actually higher at around 40-50 mph (you should get 1.5 - 2x throughput that you get at 10mph), assuming everyone is organized enough[1].

Now imagine pace cars whose sole job it is is to buffer and reduce the negative impacts of poor human decision on the freeway, with a total car increase of only 1-5%.

[1] Sorry, I don't remember the source, but I swear I had one a few years ago when I was heavy into traffic.

I haven't seen the specific study you mentioned, but it is referred to in the book, "Traffic: Why We Drive the Way We Do (and What It Says About Us)" by Tom Vanderbilt. If you are at all interested in traffic, this is a very good book to start with.

Mind you, despite being very convincing, I'm not yet a late-merger. I do find myself trying to drive more smoothly, though.

Lol, US is so slow, used in Ethiopia for ages https://www.youtube.com/watch?v=oKGeuAoGW_E

Panic inducing to anyone unfamiliar! On Monday there were some HN posts about people having peaceful, serene SlowTV video streams in the background while they work. A SlowTV channel of this would be the opposite!

Wow, that is different. I've heard not too far from the US (Venezuela to be exact) some places don't care for traffic laws.

I'm sure it would work for just cars, but what about pedestrian crossings?

Disney Springs (in Orlando, FL - Note: used to be called "Downtown Disney") got rid of crossings and made tunnels / bridges instead to keep traffic flowing. Probably not as ideal outside of Disney though. They're expanding / rebuilding the whole area. I no longer drive through there worrying about pedestrians altogether.

Or just put in roundabouts instead of lights.

http://safety.fhwa.dot.gov/intersection/innovative/roundabou...

Which, in a single lane scenario, have the additional benefit of decreasing collisions and injuries for pretty much all involved. They are also successfully proven in a roadway network (the proposed scheme here doesn't consider that case). Roundabouts are a great analog tech.

Roundabouts get less compelling when they are multi-lane (can start to suck for pedestrians and bikes in particular). They can also take up lots of space. But, on the whole, roundabouts be good.

This blog post is based on a paper[1] that explores using "slots" to arrange traffic and move the vehicles smoothly like a system of cogs. It brings up the fact that our current transportation infrastructure needs to be updated for true autonomous driving.

[1] http://journals.plos.org/plosone/article?id=10.1371/journal....

Reply on news.ycombinator.com