“It is, at it’s core, a big data problem”
Its an entropy problem. You can’t collect ‘big data’, unless you know what constitutes data...and the real world offers an infinite set of variables...you can never know all of it.
Does it have to be perfect? No. But it has to be close to it...and it can’t be.
You seem to have moved the goal post around like its on a 3 dimensional chess board. So lets recap and define some terms.
“there will be a mass produced luxury driverless vehicle on the market within 5 years”
“THEY’RE ALREADY OUT THERE. Millions of miles driven on America’s roads by self driving cars already”
“TESLA IS ON THE MARKET”
“The only reason there’s a driver in them is the law. They mostly don’t do anything, they’re ballast.”
“Tesla, Mercedes, BMW have shipped”
“That [Tesla] IS a self driving car.”
First, the NHTSA classifications:
Level 0: The driver completely controls the vehicle at all times.
Level 1: Individual vehicle controls are automated, such as electronic stability control or automatic braking.
Level 2: At least two controls can be automated in unison, such as adaptive cruise control in combination with lane keeping.
Level 3: The driver can fully cede control of all safety-critical functions in certain conditions. The car senses when conditions require the driver to retake control and provides a “sufficiently comfortable transition time” for the driver to do so.
Level 4: The vehicle performs all safety-critical functions for the entire trip, with the driver not expected to control the vehicle at any time. As this vehicle would control all functions from start to stop, including all parking functions, it could include unoccupied cars.
Tesla, BMW, Mercedes...none of them have ‘shipped’ a car that does better than level 2. They are not ‘driverless’ or ‘self driving’ in any way, shape, or form.
You have predicted level 4 will be attained in 5 years. To me, its obvious that’s impossible. We shall see.
We know what constitutes the data: the surroundings of the vehicle and the characteristics of the vehicle. The real problem comes in because any way there is to get the surrounding (visually, LIDAR, whatever else you got) is going to bring in a LOT of data, and 90% of it is worthless. The human brain has a lot of built in stuff to make it so we instinctively ignore that 90%. That’s the challenge of the self driving car, in some way shape or form it’s going to get probably 1GB of data a second, and maybe 100MB of that data is useful, it’s gotta find, understand it, and use it, all fast enough to not crash the car AND get the next second’s GB. It’s a pure data processing problem.
For driving the real world doesn’t offer an infinite set of variables. Because most of it doesn’t matter. Those cars on the other side of the street obeying the law and presenting no threat don’t matter, the buildings off the street not falling down don’t matter, the stoplight that hasn’t changed color since first observed doesn’t matter. The trick really is in ignoring all the right data. Which is ALWAYS the trick in big data. The fact is we only commit a couple of dozen actual acts a minute while driving (adjusting speed, steering, etc), it really isn’t that tough, once you’re correctly filtering the info.
The collision evasion stuff that’s in Mercedes today comes pretty close to level 3.
But really what it boils down to is this:
Once upon a time landing on the moon was impossible
Once upon a time leaving the earth’s atmosphere was impossible
Once upon a time breaking the sound barrier was impossible
Once upon a time flight was impossible
Once upon a time self driving cars were impossible
The course of human history is people like you declaring things impossible and being wrong.