Researchers propose fourth traffic signal light for self-driving car future | Hacker News Reader

Researchers propose fourth traffic signal light for self-driving car future

popsci.com

17 pointscpeterso29 comments

>To further harness these benefits, they proposed the introduction of a fourth, “white” light to traffic signals. In this scenario, the “white” phase activates whenever enough interconnected AVs approach an intersection. Once lit, the phase indicates nearby drivers should simply follow the car (AV or human) in front of them, instead of trying to anticipate something like a yellow light’s transition time to red.

This seems like it might actually end civilization.

I don't understand the rule at all. Will the white signal show in all directions simultaneously? If so, how's it going to.. you know.. control traffic? If not, why not just replace it with green? What's the actual logic here?

I suppose because the car in front of you might possibly stop at the light.

Why would they do that? To avoid a crash? What's the light even doing then? I don't need lights to rely on other drivers' desire not to crash.

Here's human drivers navigating in a similar way in a large intersection in Addis Abada (Ethiopia), I envision white light basically enabling this: https://youtu.be/VPbUpdmAfck

I'd love to see this in real-time rather than this sped up version. I suspect it's a little more reasonable to navigate when going slow enough.

Even at .25 play back speed it's still obviously way too fast in some cases.

And yes, it looks kinda normalize when the people walking are walking, not running.

Why would you want to "enable" this?

What is actually being enabled, and is it more efficient than a regular green-yellow-red intersection? If you watch on 1x, this seems like a horribly slow/inefficient intersection.

the alternative is 5km long traffic jams in all 4 directions waiting for 2-3 cars to pass each green phase.

"Once lit, the phase indicates nearby drivers should simply follow the car (AV or human) in front of them, instead of trying to anticipate something like a yellow light’s transition time to red."

I dont know what this means but it sounds similar to the green wave light that we have here in the netherlands. All it is is a lit green double wave that tells you that the very next traffic lights going straight will be guaranteed to be green. This way you dont have to anticipate breaking even if you can see the green light has been on for an entire cycle.

This is usefull because you can chain intersections so the worst case is that you have to wait two cycles on the entire chain, coming on and going off.

It's also beneficial as it'll mean everyone drives at the speed limit.

And in my town the main road has an anti-green-wave. If you‘re not excessively speeding (like three times the speed limit), you‘re guaranteed that every second or third light will be red.

Citizens asks the town administration if they have heard of green wave. Yes, they have. Having a red light every few intersections is supposed to curb speeding.

It‘s beyond stupid, especially since people living in the houses lining that road have enough noise and exhaust already.

That seems awfull. Here in the netherlands, most automatic signals dont really care about speeding. Though alot of green waves are also designed with trucks in mind. So this puts clear upper and lower limits on timings.

I’m not sure why they need a fourth signal. It seems like the main benefit is permitting faster & more dynamic transitions between go & stop, both for cars & pedestrians.

If you’re confident enough that your routing / traffic signals are going to be respected by the AVs on the road, why not just use the green -> yellow -> red lights to signal what’s happening to the humans (with appropriately shortened yellows because you know the targeted AVs are going to stop), and still reap the benefits of your interconnected AVs for better traffic flow?

Many new traffic signals have reflective yellow frames around them. They're designed to make the signals more visible to humans, but I imagine they will help autonomous vehicles' vision systems find the signals, too.

https://highways.dot.gov/safety/proven-safety-countermeasure...

How would this work if I am turning left? I might not be behind another AV. Will the AVs see that I’m there and intend to turn left and yield? What if there is a non-AV in the oncoming lane?

>There’s even a solid case to be made that completely autonomous vehicles (AVs) will never take over everyday travel.

Have these researchers seen Tesla's latest FSD version 12.3, which uses neural nets instead of hard-coded instructions? For example this video: https://twitter.com/WholeMarsBlog/status/1768714246521827643 "Tesla Full Self-Driving Beta 12.3 Drives 85 Minutes from Palo Alto to San Francisco during Rush Hour with Zero Interventions". I was also a skeptic until recently.

sf is the most heavily mapped & learned area by tech companies in the whole world

They have driving data from all around the country which is collected from anyone driving a Tesla. Waymo is much more limited to specific cities.

I was taking a friend home from the big medical complex in Boston home up 93. I haven't driven much in the city of Boston for ages and mostly not that route. I wish I had video. I had really forgotten just how horrifying driving in Boston can be at a busy time of day. Constant merges that you sort of need to know about. Often aggressive drivers. Apple Maps tried hard but I was glad I had my passenger telling me when to get left, when to get right, follow that car, etc.

SF isn't easy but Boston is next level. And it snows there. And can get heavy rain.

For years, the breakthrough has always been in the latest beta software I just hadn't seen yet. Spoiler: It wasn't.

One of these times, it will be, but I'm going to bet against it and be right most of the time.

They recently switched it to a neural net model instead of the old approach that had a ton of hard-coded C++ logic. I agree it's hard to time when it's good enough for robotaxis and such, but good progress is being made.

No, not "zero interventions. The human driver puts their hands on the wheel at 08:36.[1]

At 12:56, the vehicle drives through a crosswalk with pedestrians present, but misses them.[2] A similar situation appears near the very end. This system drives through crosswalks rather aggressively.

There are no hard cases. No heavy traffic. No construction. No emergency vehicles. No double-parked cars. Almost always, either there's plenty of open space ahead or another car to follow. When heavy traffic is encountered, the human driver changes the route. This looks like a cherry-picked try from many trial runs. It's way below Waymo level performance.

Tesla has a history of fake self-driving videos [3], so this has to be looked at suspiciously.

[1] https://youtu.be/PXy90bAPwn8?t=516

[2] https://youtu.be/PXy90bAPwn8?t=776

[3] https://www.cbsnews.com/news/tesla-autopilot-staged-engineer...

It's too easy to create an impressive demo right now, yet be very far from solving the problem. Humans set the bar on driving ability incredibly high. Humans achieve 500k miles between collisions on average. The tesla FSD tracker https://www.teslafsdtracker.com has been showing that most versions of FSD are well under 200 miles between critical intervention with one version achieving up to 300 but dropping back down in the next release.

We don't have data on 12 yet, but it seems exceedingly unlikely they've been able to close a 3 order of magnitude gap in one release to match human ability.

I say this as an owner of FSD. It has moments where it does amazing things and moments that make me shake my head. I like it for highways, but driving in a city is too stressful to be more than a toy right now.

Why don't we get more round-a-bouts in America first before we start talking about a fourth light.

Because people cant wrap their heads around roundabouts here. Theres one major one near me where accidents are fairly common, despite it having been there for many many years. Theres another tiny one that a city implemented... aaaand then put stop signs at each entry. Ive only seen one roundabout that wasnt fully being used by idiots here, and its in a semi remote area, and a small roundabout.

Prediction:

Self-driving cars and their infrastructure are installed in a couple of densely populated places.

Then the reality of the Orwellian tradeoff involved in surrendering that much personal sovereignty sets in, and the adoption falters.

Ah yes, great automation comes by defining more manual rules.

What could wrong? Specially in America where people do not even understand roundabouts.

Reply on news.ycombinator.com