how do we know 2018 VINS weren't produced in 2017? Other car companies do this.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
- Status
Video going with this article - big Tesla smile and "a disappointing lack of things to complain about!"
JRP3
Hyperactive Member
There kind of is, heat signature. I've always assumed some sort of infrared/thermal imaging camera would find its way into vehicles. Even more effective when there are no more ICE's on the road "clouding" the picture.
Should complement radar (metal detection vs. flesh) nicely.
Already posted, way back on the first page of this thread!
Well I guess now we know why they weren't using the heat, they wanted every bit of range they could get: Tesla Model 3 Sets New EV Cannonball Run Record of 50 Hours, 16 Minutes
JRP3
Hyperactive Member
I will happily admit that I am wrong if Tesla is producing M3s at 4000/wk at the end of Q1... of course I would... If Tesla produced 50k M3s for 2017... I'd have happily admitted that I was wrong also. My stated expectations for 2017 were less than 20k. I've been saying this for a year... completely against the grain... called out for FUD... and they aren't even coming close to even my expectations!
Mod: 4 posts moved to snippiness, one deleted. --ggr.
Ulmo
Active Member
mutle
Member
Re. Model 3 VINs and production, I recently came across this video on YouTube which was posted on December 6th. At 4:34 it shows the VIN (1898) and a production date of 11/2017:
Have there been other reports on the production dates listed for VIN numbers?
Have there been other reports on the production dates listed for VIN numbers?
On 12/29 when I picked up my Model 3, the guy who was doing my paper work said they were moving over 100 that day.
DurandalAI
Member
What do you do today? You slow down. Full autonomy wasn’t promised at normal daylight speeds in all conditions. The cars will have to drive slower. When forward facing LIDAR is $100 a sensor, then they can add that as a new sensor input, but until then the cars can simply go slower.So, I only threw a glance at that discussion as it pertains to LIDAR. And I spent today on the road, the last hour or so through thick banks of fog in moose country. Anyone trusting their lives and those of their loved ones to light wavelength tech alone is reckless and suicidal and should be confined. Radar also will not dependably detect animals. One always has to adapt to current conditions. I wish there was a law mandating reflectors on moose (let's start with humans)
Reciprocity
Active Member
Guys.. Vision is light based just like LIDAR. You dont need to know what a moose is to know you dont wan to hit the unidentifiable object in the middle of the road. Using Binocular cameras, the car can see the Object exists and has physical dimensions. Radar can confirm that something is physically there even if the signature is not detailed. I think Elon once said that if a UFO landed on the road, the car would stop for it, though I could be mistaken on the exact quote. I think this is kinda what he meant. You dont have to know what every object is and how it acts, you only need to know that the object exists and it something you dont want to hit. Shadows for example could look like an object in the road, but the vision system and radar determines that it has no depth and reflects no radar, so its ignored.
Edit: Just another quick note. More sensors is not necessarily better. Less sensors that are better is much more desirable. If you have more sensors, which is right when there is a conflict? Vision + Radar can be a reasonable substitution for LIDAR and can allow for enough accuracy and detail. The advantages are many, including cost and the fact that LIDAR will bounce off anything, including fog. There are LIDARs and solutions that work in Fog, but they are more expensive then a more simple system.
Last edited:
Actually he said, "we are very busy, we are delivering 100 today". I took that to mean 100 Model 3. But your point is well taken. There were a number of Model S's that I saw delivered during our 2 hours there.
ValueAnalyst
Closed
It's not that simple. LIDAR is active, cameras are passive. With binocular vision you have to do photogrammetry - matching up objects in one scene with those in another. It's an inherently error-prone process, as anyone who's ever done, say, panoramic photo stitching can attest, because some points in pairs of images will look similar, even though they're not the same object. Our minds don't screw it up because they can reason out what's the most logical explanation. But determining what makes sense or not is an AI-hard problem.
Radar is a nice piece to the puzzle, but radars see the world strangely. A moose is like a translucent ghost to a radar. I'm not even sure if thin things like its legs would even show up at all. A piece of plywood or fiberglass wouldn't. Yet a scrap of aluminum foil is like a glaring beacon to radar. Radar also has lots of reflection-noise problems, particularly in cities. The way LIDAR sees the world is more similar to how we see the world, and as a general rule, if it's solid and a possible threat to your car, LIDAR will see it, while otherwise, it won't. Not that LIDAR is without its problems; for example, an exception to the above is precipitation, which they have to try to filter out of the datastream algorithmically. LIDAR can also suffer reflection problems, but not as badly as radar. But in general, LIDAR is a superb-quality datastream. This is offset by its awkward form, high price, and the fact that you still need a vision processing system regardless; the LIDAR just helps you figure out what is where in a much better manner than photogrammetry does.
IMHO, I think there's a lot of potential to time-of-flight LIDAR. It's basically just very time-sensitive cameras, and should be very cheap in mass production - and they double as the cameras for your vision system. Broad (not narrow-beam) light pulses are sent out and the cameras determine when they're detected at each point on the sensor. Tesla should be able to readily incorporate it into future vehicles should they decide to, just replacing their existing cameras. By replacing photogrammetry with a true depth-measuring system, their existing vision system should return much more reliable results.
I wish there was a law mandating reflectors on moose (let's start with humans)
Aside: my wife doesn't like hunting for sport. Her compromise idea is to allow hunting with paintball guns. Win-win.
- Status
Similar threads
- Locked
- Article
