Autonomous Car Progress

[diplomat33]

I assume they have a variant with manual controls? I haven't seen a picture of that.

No, I don't there is a version with manual controls. The only version is driverless AFAIK.

 

[diplomat33]

Everything you ever wanted to know about lidar but were afraid to ask:

What Is Lidar? 10 Things You Need to Know About Laser Sensor Technology

Nearly all major developers of safe self-driving technology use it. Here are the top 10 things you need to know about lidar.

groundtruthautonomy.com

 

[diplomat33]

Strengths of Lidar:

Lidar’s strengths lie in producing accurate 3D images by calculating the distance to objects with the high spatial resolution inherent in sensing using light.

Strengths and weaknesses of radar:

Radar, by comparison, can accurately detect the radial velocity, as well as the range of a moving object, but lacks the resolution to accurately detect an object’s shape and exact position. On the other hand, unlike lidar, radar performs well even in fog.

Strength and weaknesses of cameras:

Object classification is the major advantage of camera technology, with the added benefit of color recognition. However, cameras have a number of limitations. These include performance in the dark and in very bright light, and the transition between the two (such as when a vehicle emerges from a dark tunnel on a bright sunny day); their performance when confronted with reflections; their inability to detect objects of a particular color against the same colored background; and crucially, they cannot measure depth directly and instead must infer it, often with much poorer range accuracy than lidar or radar.

Why AV companies do "sensor fusion":

no one sensor type alone has the capability to guide an autonomous vehicle safely everywhere people want to travel. But when several sensors are used in partnership—typically lidar, radar, and camera—the advantages of each sensor complement the limitations of the others.

What Is Lidar? 10 Things You Need to Know About Laser Sensor Technology

Nearly all major developers of safe self-driving technology use it. Here are the top 10 things you need to know about lidar.

groundtruthautonomy.com

I think the weaknesses of cameras explain perfectly why AV companies don't do "vision-only".

 

[thewishmaster]

I assume they have a variant with manual controls? I haven't seen a picture of that.

No, I don't there is a version with manual controls. The only version is driverless AFAIK.

Yeah the manual controls version is the Toyota Highlander development platform vehicle they use with the sensors/indicators arranged in the approximate configuration of the purpose-built vehicle:

 

[diplomat33]

Yeah the manual controls version is the Toyota Highlander development platform vehicle they use with the sensors/indicators arranged in the approximate configuration of the purpose-built vehicle:
View attachment 827451

Yes, I know about the Toyota Highlander that Zoox uses as a test vehicle. I thought Daniel was asking about the same Zoox driverless vehicle, just outfitted with a steering wheel and pedals.

 

[Doggydogworld]

Without manual driver controls I believe the Zoox pod must be certified as a NEV. FMVSS limits NEVs to 25 mph max speed and to roads with 35 mph or lower speed limit. Waymo's Firefly was a NEV. It was also purpose-built and electric, so I'm not sure what linguistic gymnastics Zoox is using to claim they are first.

 

[Daniel in SD]

Without manual driver controls I believe the Zoox pod must be certified as a NEV. FMVSS limits NEVs to 25 mph max speed and to roads with 35 mph or lower speed limit. Waymo's Firefly was a NEV. It was also purpose-built and electric, so I'm not sure what linguistic gymnastics Zoox is using to claim they are first.

I believe FMVSS was recently changed to allow vehicles without manual controls.
It just seemed to me that you’d want to be able to test the vehicle with a safety driver and not risk there being some unforeseen difference between this and the highlander test vehicle.

 

[Doggydogworld]

I believe FMVSS was recently changed to allow vehicles without manual controls.
It just seemed to me that you’d want to be able to test the vehicle with a safety driver and not risk there being some unforeseen difference between this and the highlander test vehicle.

You're right. I thought it was some proposed change, but NHTSA calls it a final ruling.

 

[ThomasD]

They have signal jamming devices to block Drones. It is just a matter of time before someone comes up with a jamming device to use on autonomous vehicles

 

[drtimhill]

Personally, I think L2 hands-free will be short lived because L3 and eventually L4 will take it's place. Nobody will want L2 hands-free when L3 and L4 are available on consumer cars. L2 hands-free is only a thing now because we don't have L3 or L4 on consumer cars yet.

Assuming there is no dramatic cost difference, I agree, though I think L2 to L3/4 will be a continuum rather than a clear-cut one-or-the-other as makers jostle the various definitions/capabilities to gain a marketing edge (along with the usual specmanship we are starting to see about cramming more and more sensors into the car).

 

[Daniel in SD]

Assuming there is no dramatic cost difference, I agree, though I think L2 to L3/4 will be a continuum rather than a clear-cut one-or-the-other as makers jostle the various definitions/capabilities to gain a marketing edge (along with the usual specmanship we are starting to see about cramming more and more sensors into the car).

How could there possibly be a continuum between L2 and L3/4?

 

[drtimhill]

How could there possibly be a continuum between L2 and L3/4?

Points to the operative word in my post ... "Marketing". Never under-estimate the (in)abilities of these guys.

 

[Daniel in SD]

Points to the operative word in my post ... "Marketing". Never under-estimate the (in)abilities of these guys.

I still have no idea what you mean. Are you saying they would market a system as driverless but it wouldn't be?

 

[stopcrazypp]

I still have no idea what you mean. Are you saying they would market a system as driverless but it wouldn't be?

They won't use terms like "driverless" which may get them in trouble, but they'll use the usual marketing tricks they have already used like "hands-free driving", which the general public (including many journalists and even some auto journalists) really is clueless about the distinctions. Then throw in the door-to-door L2 systems being developed and the lines will be blurred further. Then also L3 (which I'm pretty sure most of the public does not know or necessarily care about the distinctions).

 

[Daniel in SD]

They won't use terms like "driverless" which may get them in trouble, but they'll use the usual marketing tricks they have already used like "hands-free driving", which the general public (including many journalists and even some auto journalist) really is clueless about the distinctions. Then throw in the door-to-door L2 systems being developed and the lines will be blurred further. Then also L3 (which I'm pretty sure most of the public does not know or necessarily care about the distinctions).

People keep imagining that this will be a problem but if you look at Mercedes marketing it is extremely clear. In the end it doesn't really matter because if there are issues with people using L2 systems incorrectly the NHTSA will crack down on them. I actually don't think door-to-door L2 systems will ever be widely released by anyone other than Tesla.

 

[AlanSubie4Life]

I actually don't think door-to-door L2 systems will ever be widely released by anyone other than Tesla.

Yes, they are of minimal to negative utility over onramp-to-offramp L2.

Computers suck at driving. At least so far. I don’t see that changing any time soon. Though perhaps my view has been colored by certain experiences.

 

[diplomat33]

Assuming there is no dramatic cost difference, I agree, though I think L2 to L3/4 will be a continuum rather than a clear-cut one-or-the-other as makers jostle the various definitions/capabilities to gain a marketing edge (along with the usual specmanship we are starting to see about cramming more and more sensors into the car).

Sure, automakers will try to market their systems differently. So from a marketing point of view, it may look like a continuum but from an engineering point of view, the SAE levels are very distinct and separate by definition. They cannot be a continuum. In fact, the J3016 is very clear about this. A system is either L2 or L3. There is no continuum. So for example, an automaker may try to market their system as "L2+" or as "hands-free L2". But it will still be a L2.

 

[diplomat33]

Computers suck at driving. At least so far. I don’t see that changing any time soon. Though perhaps my view has been colored by certain experiences.

I think that's an over-generalization. Some computer systems suck at driving. Certainly, FSD Beta sucks at driving if you try to judge it as a L4 system. But FSD Beta is not intended to be L4. If you look at real L4 systems like Waymo, Cruise, Argo, etc, they don't suck at driving IMO which for me implies really horrible driving. They are not perfect but they don't suck. But perhaps it depends on how you define "suck". Maybe it would be more accurate to say that some computers suck less at driving than others.

By the way, this is where the SAE levels matter because if a system is L2 then it is not designed to be driving. So L2 systems will likely suck at driving since they are not designed to do all the driving. L4 is designed to do all the driving on its own, so it is fair to judge whether L4 sucks or not at driving.

 

[drtimhill]

Sure, automakers will try to market their systems differently. So from a marketing point of view, it may look like a continuum but from an engineering point of view, the SAE levels are very distinct and separate by definition. They cannot be a continuum. In fact, the J3016 is very clear about this. A system is either L2 or L3. There is no continuum. So for example, an automaker may try to market their system as "L2+" or as "hands-free L2". But it will still be a L2.

You might know that. I might know that. Marketing will NOT know that, nor will the general public

 

[drtimhill]

By the way, this is where the SAE levels matter because if a system is L2 then it is not designed to be driving. So L2 systems will likely suck at driving since they are not designed to do all the driving. L4 is designed to do all the driving on its own, so it is fair to judge whether L4 sucks or not at driving.

But this is where you get back to my continuum. Sure, it's "only" L2, but what is the disengagement rate? If it's once every 10 miles, its clearly L2, but what about every 100 miles? 1000? My suspicion is L2 vs L3/L4 will be more about liability than actual driving ability (at least in the US, where liability is such an issue). Sure, a car. maker might be technically confident that it can achieve L3/4, but I suspect many will still brand it L2 to avoid legal issues, and then boast about the disengagement rate being lower than brand Y.