No more radar means the car cannot see through fog?

[Muzzman1]

Remember one of the "amazing" selling features of the radar was the car could see through fog, and could even see the car in front of the car that was in front of you. So "If a UFO landed on the road, the car in front would hit it and the Model S will stop because the radar can see through the car in front"
SO since Tesla is going to "pure" vision, does that mean fog and sandstorms will not be as safe anymore?

 

[Silicon Desert]

Thankfully, I don't live in Area 51

 

[DerbyDave]

Tesla can't even see an 18 wheeler across the road in front of it now, because Tesla does not use its radar much now anyway. The two or more collisions with HUGE objects across the road mean the radar now is either not used, or not working on any of its cars. If they can't figure out how to distinguish between a shadow on the road, the road going under a bridge, or a huge truck lying in the road now, it won't get any worse when they stop using the radar completely. A camera may be able to see better at night, in snow, and in rain -- but it sure seems like radar could add good input into driving decisions if it were used properly, and may help in foggy conditions..

 

[CaliSurfer]

When it gets foggy where I live, my car says the radar can't see and auto pilot is currently unavailable so I don't think much will be affected by losing the radar.

 

[SSonnentag]

When it gets foggy where I live, my car says the radar can't see and auto pilot is currently unavailable so I don't think much will be affected by losing the radar.

Agree, other than perhaps AEB not being triggered while on ACC without the radar to see an object through the fog. But then I'd argue that ACC shouldn't be used in the fog to begin with, so it's really a moot point IMO.

 

[kavyboy]

Last time I used AP1 in the fog/rain, it worked fine. I hope that doesn't go away.
AP is actually very useful in bad weather conditions, like slower highway travel in torrential rain. Did the invisible grey car without headlights that just passed cut in front of me? Did the car (that I can't see) in front of the car in front of me (that I can see) stop suddenly? AP detects and reacts, which help without me having to take my eyes off the road.
(There are plenty of traffic pile-up videos out there showing why stopping by the side of the road isn't necessarily safer, BTW.)

 

[serendipitous]

Tesla can't even see an 18 wheeler across the road in front of it now, because Tesla does not use its radar much now anyway. The two or more collisions with HUGE objects across the road mean the radar now is either not used, or not working on any of its cars. If they can't figure out how to distinguish between a shadow on the road, the road going under a bridge, or a huge truck lying in the road now, it won't get any worse when they stop using the radar completely. A camera may be able to see better at night, in snow, and in rain -- but it sure seems like radar could add good input into driving decisions if it were used properly, and may help in foggy conditions..

You're mixed up. It was the reliance on radar that CAUSED those incidents, because radar doesn't do a good job of detecting fixed, stationary objects. The car doesn't expect stationary obstructions to simply block highways, so when the radar stops detecting fellow moving vehicles in front of you, it figures it's probably all clear, absent very deterministic visual evidence to the contrary. Then, to counteract that phenomenon and the risk of bad press, they started being overly cautious with things like shadows, causing an increase in phantom braking - I'm so used to relying on radar that even though I don't sense anything, I do see a suspicious shadow so let me slam on the brakes just in case it's the shadow of a sideways semi, etc. Take it with whatever size grain of salt you like, but there's merit in Elon's claim that sensor fusion is risky - if you're going to say that you can't trust radar sometimes, then why use it at all? You still have to pick your winner among your sensors.

 

[CyberGus]

Infrared can cut through fog and dust, although I do not know which wavelengths are being imaged by the cameras.

Agreed that pulling over when encountering zero visibility is only useful if everyone behind you does the same.

 

[beatle]

if you're going to say that you can't trust radar sometimes, then why use it at all? You still have to pick your winner among your sensors.

There are times when you can't trust cameras either. Having more information from more sources at your disposal can be useful in avoiding pitfalls of each other, at least if you have enough of them to break a tie. Having cameras and radar (and lidar) would be useful in having the best information to decide how to proceed given the conditions.

 

[RickParker]

As a pilot for 45 years, I don't have cameras or LIDAR but I have a very expensive RADAR (and TAS) system. I trust but know it's not perfect in it's interpretations of a weather system in a dynamic weather system. RADAR can't see the other side of a target. I have never flown into a Cumulus Granite cloud as those clouds are not at all dynamic. Autopilot can't be absolute and they fail in planes too. #justsayin. Airplane radars can't be turned on while on the ground, they can be harmful to people.

 

[ucmndd]

Tesla can't even see an 18 wheeler across the road in front of it now, because Tesla does not use its radar much now anyway.

This is a fundamental misrepresentation/misunderstanding of how radar works and what it's good at.

 

[whitex]

Radar also sees under cars and able to "see" a car in front of the car you are following - a piece of information Tesla AP will no longer have. Some of the other manufacturer's EAB (no AP) even use this to brake earlier if your vehicle closure rate to the car in front of the car immediately ahead of you is too fast.

 

[DerbyDave]

This is a fundamental misrepresentation/misunderstanding of how radar works and what it's good at.

Perhaps this is not how Tesla radar works. Are you saying that Tesla radar only functions in identifying moving targets, but not the stopped car or truck or wall in front of it?

 

[ucmndd]

Perhaps this is not how Tesla radar works. Are you saying that Tesla radar only functions in identifying moving targets, but not the stopped car or truck or wall in front of it?

Radar will always be poor at identifying stationary objects. Is that a boulder on the shoulder up ahead or a stopped car? Is that an overpass or a sideways truck? Radar doesn't have a good way of discerning the difference and that's basically why the problem is so hard - success is NEVER phantom braking for an overpass but ALWAYS braking for a sideways truck with a similar stationary profile.

Radar is great at detecting relative motion - a car in front of you going 5 miles per hour is way easier for radar to detect as a hazard and respond to than one that is fully stopped.

You basically said above that because Tesla radar is poor at doing the one thing radar is particularly bad at, it means they "don't use it much now anyway". That doesn't make any sense.

 

[Tam]

Perhaps this is not how Tesla radar works. Are you saying that Tesla radar only functions in identifying moving targets, but not the stopped car or truck or wall in front of it?

Please read the article:

Why Tesla's Autopilot Can't See a Stopped Firetruck

Semiautonomous driving systems are designed to ignore unmoving obstacles because otherwise, they couldn't work at all.

www.wired.com

Radar can detect many obstacles but for the car industry, they use the trick of speed: Detected objects that have a speed are dangerous and detected objects that have no speed including a stationary obstacle like a giant nonmoving truck blocking in front.. are ignored.

A tree has a speed of zero because it is not moving at all, so it is ignored. Most of the time, that is safe to ignore because no one is crazy enough to plant a tree right in front of my lane that my car is moving forward.

However, if the lightning cause the tree to block the road, and that fallen doesn't move. It has a speed of zero. In this case, ignoring this fallen tree on the road would be disastrous which is the same scenario as when the semi-truck turning left in front of a moving Tesla. The radar detects the obstacles but since they have a speed of zero, the car industry agrees that they are to be ignored and the Tesla would not apply the brakes.

That's why it's a L2 ADAS, driver still needs to brake in those scenarios.

 

[CyberGus]

Part of the problem is that radar has very low resolution; it's basically 1 big pixel lol. OK, maybe not that bad, but it has trouble distinguishing between a truck lying across the road, and an overpass above the road. LIDAR is much better at creating a 3D map of the environment (or 2 cameras + processing).

 

[DerbyDave]

Interesting topic in that as of today, Model Y and Model 3 vehicles bound for North America are being built without radar. For the time being, these models shipped to other areas of the world will still have the radar installed.

 

[ucmndd]

Interesting topic in that as of today, Model Y and Model 3 vehicles bound for North America are being built without radar. For the time being, these models shipped to other areas of the world will still have the radar installed.

I think that's because of the differences in signage, lane markings, etc etc etc - it's gonna take Tesla a long time to do all that vision/neural net training for all the various international locales and standards.

 

[Rice&Curry]

what would be the cost of the radar installed in M3/MY? I don't think the discontinuation is related to cost saving measures- but relying solely on vision seems to make it dependent only on one type of sensor to make all decisions. Each type of sensor has its own advantages and if there is a way to harness the uniqueness of each sensor and then use sensor fusion or some other method to get a more comprehensive assessment of the situation for the software to make its decision after cancelling out situations which are not applicable or have lowest probability of occuring. The processors are fast enough to process data simultaneously from multiple sensors to arrive at the most likely situation to make an informed decision. I do not know enough about radars or lidars to comment on them.

 

[Silicon Desert]

When it gets foggy where I live, my car says the radar can't see and auto pilot is currently unavailable so I don't think much will be affected by losing the radar.

hmmmm, that's not the message it conveys to me. The cameras get affected for sure, but I doubt the radar is affected by fog. At least not any radar I have every worked on.