You need a certain amount of torque from the engines to get a full load up an incline. Tesla have over-specified this so that their truck clearly performs better than an ICE truck.
You also need a lot of batteries to get an adequate range. Again Tesla have over-specified it. Truckers don't need 500 miles range. That's further than they can legally drive without rest breaks.
Once you have several big motors and big batteries, the acceleration performance comes for free.
Looking behind the cab, there was a lot of empty space, between the sidewalls of the cab. This is clearly an area that they've yet to design. Obviously not a design priority, so long as the space is available.
I'd expect 2 passenger seats, one left and one right of the driver, but further back due to the curved glass of the windscreen.
I was watching a video about a different company's convoying approach. It spoke about the convoy spreading out more as it was approaching an exit, and then closing up again after the on ramp. Where there's miles between exits, that could be workable.
Tesla don't have all the time in the world to hang around waiting for one product to mature before launching another.
First of all from the business perspective. Electrification and automation has given the very rare opportunity for a new leading car manufacturer to be created. Possibly even a new market leader. But that can only be managed if they become as big as possible whilst the other companies catch up to the technology.
And from the environmental perspective, the quicker Tesla pushes the industry towards total electric, the sooner greenhouse gasses can be reduced, and the less harm is done to the planet.
Musk said Roadster 2 will do >250mph. So it's in the same category as the Chiron. Until the production release we don't know how much faster than 250mph either of them will go.
If fear would help, that can be added as a variable alongside all the others. Stop imagining that there is something unique about the human brain that means that computers can't replicate their tasks. The last 50 years has been a catalogue of people thinking that and being proved wrong.
I'm not saying that computers won't be able to do this stuff eventually. I'm saying it's harder than the issues around operating in the same space as non-autonomous cars.
Sure, but all the things you list are common enough. And the car will have good vision and perfect control.
Yes, pedestrians deliberately stopping autonomous cars is certainly going to be an issue. Not just beggars, but protestors, trolls, carjackers. There's certainly going to be problems there to tackle. But it's more a law enforcement issue than an autonomy one. Stopping when a pedestrian stands in front of a car is the only reasonable response from an automated system. Same as for a human driver in nearly every case.
For example cops could travel around in bait unmarked autonomous cars, and arrest those that interfere with them.
If the computer is overloaded it will slow down and if necessary stop. Your bizarre remote control puff balls will be visible on the cameras, same as any other tactic.
AP is only level 3. We're talking about level 4 and 5 here. AP Is only doing lane following and adaptive cruise control, and on any moderately curvy or otherwise interesting road it panic frequently and hands over to the driver. So no, that's not evidence that snow is not a show stopper for Level 4 & 5 right now.
"Baloney. Computers deal with skids way better than humans."
Reality: computers can recover from skids with traction control better than people. It would also be easy to have it steer into the skid in a skidpan. However real roads in slippery conditions need actual context aware decision making. For example last winter near me, a steep downhill was icy. The downhill ended at a stop sign leading to a busy road. 3 cars in a row had encountered the ice, and used the little control they had to deliberately crash into a wall at the side, rather than continue down the road avoiding collisions and end up careering into another vehicle at the bottom. As yet, no computer is near making a decision like that.
And given that self-driving cars have cameras pointing in every direction, plus logging of every action they take, these will be the least likely of all cars to suffer from insurance fraudsters.
What utter nonsense. You start with something that works for the general case. Then you start exploring the edge cases. Writing a test for each potential issue. that's standard Test Driven Development. And standard practice (minus doing the tests first) for every other kind of coder too.
If you are trying to tell me that people deal with the hard cases first, before the general cases, I won't believe you have any experience at all.
There are interesting issues to do with interaction with other road users. And there are people working on those things. But no, it's not nearly as big as issue as snow. Snow is a showstopper right now. It interferes with sensors. It makes the expected view completely different from the ones usually trained with. And what to do in a skid is a huge issue.
Slashdot Mirror
You need a certain amount of torque from the engines to get a full load up an incline. Tesla have over-specified this so that their truck clearly performs better than an ICE truck.
You also need a lot of batteries to get an adequate range. Again Tesla have over-specified it. Truckers don't need 500 miles range. That's further than they can legally drive without rest breaks.
Once you have several big motors and big batteries, the acceleration performance comes for free.
In motor racing it might need to be inches. With big trucks you are still gaining a long way further back than that.
There's also the point that they will spread out as they approach an exit, then close up again once they are passed the on-ramp.
Looking behind the cab, there was a lot of empty space, between the sidewalls of the cab. This is clearly an area that they've yet to design. Obviously not a design priority, so long as the space is available.
I'd expect 2 passenger seats, one left and one right of the driver, but further back due to the curved glass of the windscreen.
I was watching a video about a different company's convoying approach. It spoke about the convoy spreading out more as it was approaching an exit, and then closing up again after the on ramp. Where there's miles between exits, that could be workable.
Maybe. Though I'd say any car company would be foolish to sink any money into non-electric non-automated projects at this stage.
There's nowhere to go 250mph for long. It's a party piece no matter what fuels the car.
Tesla don't have all the time in the world to hang around waiting for one product to mature before launching another.
First of all from the business perspective. Electrification and automation has given the very rare opportunity for a new leading car manufacturer to be created. Possibly even a new market leader. But that can only be managed if they become as big as possible whilst the other companies catch up to the technology.
And from the environmental perspective, the quicker Tesla pushes the industry towards total electric, the sooner greenhouse gasses can be reduced, and the less harm is done to the planet.
We can load heavy boulders on the truck from mountain tops, drive it down hill and charge the grid!
And the byproduct is a flatter planet, which makes it easier to pave.
Musk said Roadster 2 will do >250mph. So it's in the same category as the Chiron. Until the production release we don't know how much faster than 250mph either of them will go.
It's annual. Start one now and it'll be ready for next years contest.
You train a neural net of course.
No. But then if I was driving the semi, I wouldn't have kept on reversing till I hit the shuttle either.
If fear would help, that can be added as a variable alongside all the others. Stop imagining that there is something unique about the human brain that means that computers can't replicate their tasks. The last 50 years has been a catalogue of people thinking that and being proved wrong.
CPU's can't anticipate? Did anyone tell Deep Blue or the Table Tennis robots?
Try studying TDD. What I've told you is exactly the way it works. And if you think you are better than that, you're not.
I'm not saying that computers won't be able to do this stuff eventually. I'm saying it's harder than the issues around operating in the same space as non-autonomous cars.
Sure, but all the things you list are common enough. And the car will have good vision and perfect control.
Yes, pedestrians deliberately stopping autonomous cars is certainly going to be an issue. Not just beggars, but protestors, trolls, carjackers. There's certainly going to be problems there to tackle. But it's more a law enforcement issue than an autonomy one. Stopping when a pedestrian stands in front of a car is the only reasonable response from an automated system. Same as for a human driver in nearly every case.
For example cops could travel around in bait unmarked autonomous cars, and arrest those that interfere with them.
If the computer is overloaded it will slow down and if necessary stop. Your bizarre remote control puff balls will be visible on the cameras, same as any other tactic.
AP is only level 3. We're talking about level 4 and 5 here. AP Is only doing lane following and adaptive cruise control, and on any moderately curvy or otherwise interesting road it panic frequently and hands over to the driver. So no, that's not evidence that snow is not a show stopper for Level 4 & 5 right now.
"Baloney. Computers deal with skids way better than humans."
Reality: computers can recover from skids with traction control better than people. It would also be easy to have it steer into the skid in a skidpan. However real roads in slippery conditions need actual context aware decision making. For example last winter near me, a steep downhill was icy. The downhill ended at a stop sign leading to a busy road. 3 cars in a row had encountered the ice, and used the little control they had to deliberately crash into a wall at the side, rather than continue down the road avoiding collisions and end up careering into another vehicle at the bottom. As yet, no computer is near making a decision like that.
Do you enjoy being wrong, always doubting everything?
Slashdot tradition I suppose. "No wireless. Less space than a Nomad. Lame."
Right. Because your roads are special snowflakes.
There already is a self-driving racing format.
https://www.youtube.com/watch?...
And given that self-driving cars have cameras pointing in every direction, plus logging of every action they take, these will be the least likely of all cars to suffer from insurance fraudsters.
What utter nonsense. You start with something that works for the general case. Then you start exploring the edge cases. Writing a test for each potential issue. that's standard Test Driven Development. And standard practice (minus doing the tests first) for every other kind of coder too.
If you are trying to tell me that people deal with the hard cases first, before the general cases, I won't believe you have any experience at all.
There are interesting issues to do with interaction with other road users. And there are people working on those things. But no, it's not nearly as big as issue as snow. Snow is a showstopper right now. It interferes with sensors. It makes the expected view completely different from the ones usually trained with. And what to do in a skid is a huge issue.