Hacker News Reader

Comment by rwmj | Hacker News Reader

This is hardly city driving. Try driving in a European city. (I love this project and can't wait to buy or rent a self-driving car)

View on HN

Southern Europe in particular, Sicily was good fun; I hear Naples is even more exciting.

The interesting thing was that the drivers were quite skilled but the way they drove and the traffic moves they had to be and even with those skills there were far more dented cars around than in the UK where systems and organisation allow a lower skill level.

Harder yet, Southeast Asia. Hundreds of pedestrians, tons of vehicle variety (some entirely makeshift), uncontrolled intersections, very little regard to any regulations.

> very little regard to any regulations.

That's an important point, I never expected lanes count and lane direction to be interpreted as suggestions until I saw people going through a roundabout on both sides at the same time.

Your other comment in this thread got killed, probably because of the url shortener.

Edit: I think comments killed for some reasons can't be recovered by editing them.

Thanks, I replaced it with the huge ugly full URL, does it work now?

edit: well, removed the link entirely, better?

edit 2: welp, I can't delete it either so whatever. Thanks for the notice and help.

Yeah I was just biking in Vietnam and Cambodia and it's quite a different system with stuff all over the place. There is a kind of logic to it - you project their trajectory and yours and if there's no impact you keep going, otherwise you modify speed and direction in a minimal way to avoid impact regardless of going the wrong way on the wrong side of the road etc. It takes quite a bit of getting used to as a western driver though from a computational software point of view the above is probably a fairly simple algorithm.

The basics of flock behavior.

I think swarm-like logic might work better at that level.

In my mind structured road conditions are a prerequisite for driverless cars.

I've been in many situations where conservative driving according to the rules of the road is not possible. In the West this is more limited to dense cities and exceptional situations. In India, though, this is the rule.

If you waited for pedestrians to finish crossing or for cars to clear intersections, maintained a safe following distance, signaled before turning, waited for other parties to yield, etc, you would never leave your driveway.

It's difficult for me to imagine a computer deciding to gently nudge pedestrians with its bumper in order to proceed through a green light, which I've had to do several times while driving there.

Crime of the future: change the road environment at the right time to get the target's car to crash themselves.

@zimbatm (because there isn't a reply option at this level)

I'm also sure there are screenwriters already working on incorporating such scenes into movies. Likely we'll see it on the big screen before we see it in real life.

Which is why Self Driving Cars™ will never be legal without strong AI.

There's no such thing as "Structured Road Conditions" anywhere in the world. Whether it's the ice storms of the Northeast US, radar interference, one lane two-way roads, potholes that can devour a wheel or roads that turn into mud -- the analog world is a harsh and unrealistic place for a technology that has never made its way out of cherry-picked municipalities in ideal conditions.

The last chapter is not city driving. It's the first chapter. So far Google has tackled only the lowest hanging fruit.

I will give you a more harder test environment. Try driving in any major Indian city, especially in the city core.

Super narrow roads. Zero adherence to traffic rules. Kids playing on the roads, sudden reactions from people on the road when they have to cross. Unsignalled sudden breaking, turns, U-turns, people popping up in front of the vehicle from no where etc etc.

What's more how should you expect the car to react when a dog comes chasing. Or when a traffic police guy asks you to stop or take a diversion(You won't believe how common this is here). How will you negotiate frequently occurring speed breakers and pot holes. The list is endless.

I believe starting to test some where in the streets of a city like Bangalore/Mumbai would help.

I think self-driving cars will be introduced very gradually. First only highways where it's really easy for a computer, then maybe American cities. The harder the street layout, the longer we'll have to wait for this technology. So maybe American cities will get the technology a few years before European.

I don't understand why the detractors don't understand this. Many people (including some commenters here) seem to think that the technology is useless unless it can handle all possible situations.

Auto-drive on long highway sections would be an amazing improvement by itself and prevent all kinds of fatigue related accidents. The cars could just announce ("Driver, take control") as they exit the highway system.

Hell, trucking companies could just have drivers waiting at highway exits to do the last-mile driving, the long-haul sections of trucking wouldn't require a person on-board.

I don't think it would be unfeasible to install beacons designed to aid driverless vehicles either, especially on highways where things don't change very quickly. Simple posts with machine-readable signalling (visual/radio/whatever). Most signage is designed for human recognition, but I bet it wouldn't be impossible/terribly expensive to install signage designed for automated driving.

I wonder if far more cars surrounding the driverless car will make it even more robust in some ways, for the short-medium term. If there are swarms of cars surrounding a driverless car, tricky things like hand signals and sign interpretation will be less important, as other human cars will hint at a correct traffic flow. It's surely far easier to detect the predict the motion of a gigantic car, as opposed to relying on interpreting roadwork signs.

Exactly, I live in Rome and you'd really be surprised by the number of bikers who drive on the sidewalk to avoid morning traffic. I don't think Google cars are prepared to respond to people merging from the sidewalk, they're just unpredictable and nothing will ever change that.

Human drivers can't predict that either, and the Google car should have a much faster reaction time than a human driver.

Reaction time isn't the problem. Cars can't make moral decisions.

It's not uncommon that a pedestrian will jump in the road (sometimes intentionally to commit suicide). This happens to me personally on a daily basis in San Francisco. Sometimes the driver has a moral choice to ether run the car off the road killing people on the sidewalk or killing the pedestrian.

What do you expect a Self-Driving car to do in that situation?

"Cars can't make moral decisions."

There are moral problems that I'd be uncomfortable putting to a computer.

This isn't one of them. Am I missing something, is there something wrong with simply modeling likely impacts and trying to cause the least harm to the fewest number of people? That's what I would do, only with worse reaction time and less information.

Wait .. what ..? People try to kill themselves under your car daily? Are you confusing reality with GTA?

I don't see why not. If I were writing a self-driving car I would care that

1. My car stays in lane and obeys the rules of the road

2. It doesn't hit anything

Fulfilling 2) is "easy", you just need to make sure you have sensor coverage within a safe stopping distance.

Besides, a biker is just a faster cyclist, and it's common and often legal for cyclists to merge from the sidewalk, if the sidewalk is marked for that purpose

Reply on news.ycombinator.com