If you zoom out, you'll see that there's a method to the madness but a lone driving car that's not a part of a fleet doesn't have such luxuries.
I don't think the problem is solely image processing and machine learning. I just don't get self driving and how we're supposed to wow the logic. You have to be aggressive to drive in a city or people will constantly cut you off for hours. This means you will hold up traffic making the situation worse for everyone. However, you can't be too aggressive. Imagine the damage we will cause to the "optics" of self driving if we run over (at a slow 10 mph) stubborn people who cut us off. That will just look like murder.
Consider that in US the cheapest ZipCar is ~$8/hr.
Adding even the cheapest human driver working at minimum wage is ~$15/hr.
Adding a human driver in US makes the trip 3x more expensive.
Or to put it another way: a self-driving trip can be 3x less expensive.
Or to put it another way: a $30 trip from SF downtown to airport becomes $10 trip which is competitive with BART prices.
I don't know what is the minimum wage in India, but per https://en.wikipedia.org/wiki/Income_in_India per capita income is $1670 / year which is probably less than $1 per hour.
In other words: self-driving cars only make economic sense in countries where humans are expensive.
I looked up BART because of what you said and it looks pretty bad.
https://github.com/enjalot/bart/blob/master/data/bart-comp-a...
175 people make six figures in base pay. The total cost of employment is ridiculously high. I don't know anything about BART but they clearly need to lay off a lot of people. This is ridiculous.
A self driving car can have much better "perception" of the surrounding area and reaction times under 100ms, making it potentially very good at flocking safely.
It tells us that Waymo is nowhere close to "Level 5" or even to the levels Google and the latest cherry-picked "tests" (where they don't count "disengagements" as a problem) would like you to believe it is at.
Level 5 is supposed to be virtually perfect self-driving in any environment or on any type of road, for which we haven't even properly tested in simulations yet, let alone in the real-world.
I said it before, because of marketing and an incentive to be misleading about their capabilities, "Level 5" will be the "4G" wireless of 2011 (and by that I mean "3G+", in case it wasn't clear). That's actually held true since "Level 2" self-driving. For every level carmakers have been patting themselves on the back for reaching, they're at least one level behind in the real-world. They quickly discover this, too, after trying out their "Level X" cars in the real-world for a while, but they "forget" to tell us about it, after making the big announcement previously, or it doesn't get as much attention in the media.
So, I guess I'll wait for Level 6 or Level 7, before I get into one of these. Level 7 will probably be Level 6 cars that actually can't get hacked, unlike all the cars progressing through the previous levels. Carmakers won't even care that much about security and will only prioritize getting to adequate self-driving capabilities first and will "deal with security later", because we all know it's so much easier to fix security issues and redesign architectures later.
As a resident of a first-world country, it's not really my problem if India needs to wait a few decades or modernize their urban transport infrastructure to see the benefit of this technology. I haven't seen it in person, but the reputation is that Street-level activity in those environments is deadly and chaotic, regardless of whether you are a pedestrian, cyclist, or motorist.
Not exactly a situation worth preserving from anyone's perspective...
Why limit it to US And Europe. The world might soon be a better place. Imagine Starting out in Berlin Then over a couple of months go to Turkey, then India, Vietnam and beyond and then back without ever driving a single km. Just let it drive while you sleep.