Todd Burch
14-Year Member
Didn’t say the cars break down. I said the system breaks down.
If they have zero reliance on mapping data, then why do they gather mapping data and limit the regions in which they operate? Educate me.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Waymo Says Its Self-Driving Tech Is More Advanced Than Tesla FSD, Others
- Thread starter TMC Staff
- Start date
If they have zero reliance on mapping data, then why do they gather mapping data and limit the regions in which they operate? Educate me.
diplomat33
Average guy who loves autonomous vehicles
I do believe that the real test is not doing FSD but how reliable your FSD is. The fact is that there are plenty of companies now that have achieved FSD, more or less. The real test is whether the public accepts your FSD as "good enough". That's really what the FSD race is about. That's why companies with FSD are working so hard to reduce disengagements, make the driving policy more natural etc... They are trying to get to that point where their FSD is "good enough". The tricky part is that nobody really knows what "good enough" is. So everybody is working to improve their FSD to reach what they think is "good enough".
diplomat33
Average guy who loves autonomous vehicles
They gather map data because it makes the system more reliable and Waymo wants the system to be as reliable as possible. So they naturally use tools that add more reliability.
Think of this way: let's say your car can navigate intersections correctly 99% of the time with camera-only but it can navigate intersections correctly 99.9999% of the time if you include maps. Maybe 99% is "good" but 99.9999% is better. Why would you not use maps?
Basically, Tesla is going with the approach that is 99% accurate now and working to improve it to 99.9999%. Waymo is going with the approach that gives them 99.9999% now.
To your second point, it is easier to test your autonomous car is safe enough in a smaller area than a bigger area. In a smaller area, you can test all the different conditions and make sure that your car can safely handle everything. In a bigger area, it will take longer to test all the conditions. Waymo wanted to deploy a commercial service with real passengers because it will allow the public to taste what autonomous driving is like and therefore help the public accept autonomous driving. But the rides have to be safe for the public. So they picked a small area that they had already tested that the autonomous driving was safe enough for the public to ride in. But Waymo is testing in over 25 metro areas in the US. So they are testing in larger areas.
Todd Burch
14-Year Member
HookBill
Member
I don't understand why people think using HD maps and lidar are a bad idea. If tesla said they were putting in lidar tomorrow everyone would be cheering it as another great tool for tesla to use. Just because they use HD maps doesn't mean the system fails without them and just because tesla doesn't use maps or lidar doesn't make their system better. It just means it's got less inputs to interpret.
Lidar is not going to be big, bulky and, expensive for much longer.
Velodyne Unveils a New Solid State Lidar Sensor for Driver Assistance Systems & Autonomous Driving
Lidar is not going to be big, bulky and, expensive for much longer.
Velodyne Unveils a New Solid State Lidar Sensor for Driver Assistance Systems & Autonomous Driving
Todd Burch
14-Year Member
Mainly because:
My opinions:
1. You need to solve vision anyway.
2. Lidar + Camera will always cost more than just Camera, no matter how cheap lidar is.
3. There are times where lidar doesn’t work and you need vision anyway.
4. Once your camera perception system is good, lidar is redundant.
5. Lidar needs 360 degree field of view (siren on top of car= funny looking, aerodynamics penalty), or you need multiple lidars (cost problem).
6. Maps need maintenance. Increases cost to operate system. If there is a different between map & reality, who do you trust? You need cameras to reconcile the difference. See 1 & 4.
7. If humans can drive with 2 cameras, it’s proof that no other sensors are needed.
8. Road system is already built for a 2-camera system.
There are probably more reasons.
Last edited:
Redundant subsystems can be used to improve overall system reliability. We humans use multiple sensory inputs when driving, hence the "rumble strips" commonly placed atop lane markings. If we hear and/or feel our car drifting out of its lane, we are prompted to take corrective action.
Tesla's vision system itself has redundancy in that there are eight cameras. But does LiDAR potentially provide useful data that cannot reliably be discerned by vision alone? Even with a very good vision system, I think the answer is likely "yes", particularly at night. Clearly, HD maps can also be very useful. Therefore, it makes little sense, in my opinion, for anyone to be categorically opposed to integrating LiDAR and/or HD maps when and if they're available, economical, and can be packaged efficiently.
My overall feeling is that Waymo has adopted a "kitchen sink" approach to solving autonomy - spare no expense on hardware, whatever it takes to achieve maximum reliability as early as possible. I think they've been very successful relative to their own goals, and if I were in a Waymo service area, I'd feel quite comfortable riding in a Waymo vehicle (except for the fact that they use ICE Pacificas - yuck!).
Tesla, on the other hand, has a "hardware lite" approach - keep the hardware as simple and unobtrusive as possible, and work wonders in software. This approach involves compromises, but it seems like the only reasonable approach for a mass-market, relatively affordable product in the not-so-distant future. It may not be perfect, but we hope and expect that it will be more than good enough.
Todd Burch
14-Year Member
I don’t disagree in general. But in a Waymo, if the lidar or camera system goes down, do you think the car would keep driving? I bet it immediately goes into “I need to park” failure mode.
Tesla has 8 cameras primarily to be able to see all angles, however with 2 forward looking cameras (with different focal lengths) and a forward radar, that’s enough redundancy to go into a pull over and park” mode if a camera or radar fails.
Rumble strips only exist because of human inattention and sleepiness, a uniquely human problem that doesn’t affect FSD.
IF it was economical and could be packaged efficiently, sure. However, in the longer term, these systems will be competing for customers based on cost. A lidar + vision system will always cost more. Cameras are cheap. Even if lidar gets to be as cheap as a camera, it’s still twice the cost of a camera alone.
More systems require more computing power. This requires more powerful hardware, higher cooling and energy requirements. More cost.
Precision maps are great when they’re right, but if there’s construction and an unmapped area, you now have disagreement between the map and reality. What resolves that discrepancy? How do you shuttle data for precision maps from a data center to the car?
For someone driving from LA to New York, that’s a lot of data transfer. More cost, more energy usage. I think this is another reason why Waymo operates in limited regions. To be able to drive generally from location A in one state to location B in another, you’d have to transfer many gigabytes, perhaps terabytes of data to the car.
This is why Musk believes LIDAR is a crutch. Yes, it adds redundancy and “helps”. But ultimately, to handle the edge cases, you need vision to work on its own anyway. Nice to have, but adds cost and complexity when vision is necessary anyway. Musk is successful because he’s able to boil complex problems to the simplest approach.
His mindset was “attentive humans can drive pretty well with just 2 cameras that can only look in one direction at a time, and situated inside the car behind the steering wheel”. So 8 cameras looking in all directions, plus a radar, is a superhuman perception system.
All in all, the true level 5 systems of the future will be competing on cost.
Agree with you here on all points. No system will be perfect. But I believe that a pure vision based system can still be several orders of magnitude safer than human drivers.
This sums it up quite nicely. I do feel that Tesla's approach is absolutely the right one. Thinking ten or twenty years into the future, though, I can imagine that additional sensors such as LiDAR, microphones, etc. may be much easier and cost effective to add, along with HD maps where available, even when accounting for the additional processing power and complexity. Now is not the time for Tesla to do this.
Tell me where to meet you in your Waymo car and I will bring my Tesla, then we can compare with an interstate drive.
HighZ
Member
If Elon Musk’s goal was to build the best FSD no matter the cost, I am sure Tesla could have achieved L4 by now as well.
I believe his goal instead was to create a system that could be affordably added to a consumer car. Or at least that is where he is at.
From a technology point of view, Waymo in fact is a more capable system, but at a high cost. Tesla will sell WAY more FDS systems and cars that Waymo probably ever will. From a businessperson’s perspective, Tesla has the advantage.
I believe his goal instead was to create a system that could be affordably added to a consumer car. Or at least that is where he is at.
From a technology point of view, Waymo in fact is a more capable system, but at a high cost. Tesla will sell WAY more FDS systems and cars that Waymo probably ever will. From a businessperson’s perspective, Tesla has the advantage.
Waymo isn't in it to sell cars though. It doesn't need to be cost effective to a daily driver.
Also according to tesla FSD is going to be worth over $100,000 when it works and the price will continue to go up from here. How is that working with your statement of making fsd affordably added to a customer's car?
HighZ
Member
I don’t see that changing. No matter how much the feature is ‘worth’, a consumer is only going to pay so much, so Tesla will continue to find that limit. They obviously can’t charge $100,000 as an add-on, but could they get $20,000? Maybe. But Elon can determine what will be the price point that will make sense.
My point was as yours, Wayno isn’t looking to sell cars, but Tesla is. They have to go at the problem differently in order to achieve their respective ends.
CyberGus
Not Just a Member
Back before ubiquitous mobile GPS, I rented a car with a nav system. The route I took through the mountains caused the internal maps and perceived location to be skewed, so it was showing me as driving off the road and constantly telling me "Please make a u-turn to return to your designated route."
I asked the passenger to try to shut this thing up, but instead he managed to change the language to Japanese. So now the car was screaming at me "指定されたルートに戻るには、Uターンしてください U TURN!"
Evidently there's no Japanese word for "u-turn", I guess they can't afford the shame of admitting they were doing in the wrong direction
I asked the passenger to try to shut this thing up, but instead he managed to change the language to Japanese. So now the car was screaming at me "指定されたルートに戻るには、Uターンしてください U TURN!"
Evidently there's no Japanese word for "u-turn", I guess they can't afford the shame of admitting they were doing in the wrong direction
This just shows that it's a not a good comparison because the two goals are entirely different.
It's better to compare Tesla against MobileEye because MobileEye is targeting widespread public availability of self-driving cars. As in the technology will be in cars we can actually own.
Waymo/Cruise and other fleet targeting companies have no interest in selling vehicles as they're targeting selling rides.
They're also way more safety conscious as a result as this means what they choose to use, and where they choose to operate is going to be extremely different than what Tesla goes with.
What that being said both "operate in any driving domain", and "generalized autonomy" are fantasies. Even human beings aren't capable in our best moments of achieving either of those with the level of safety we expect robots to achieve.
Humans only operating in any driving domain because we allow the safety threshold to change, and we only have generalized autonomy because we're okay with uncertainty. We're okay that we might not handle something correctly, but we'll deal with the moment when it comes.
In the next decade we likely won't see autonomous cars working in any driving domain or generalized autonomy so who cares who's first?
In the coming years I fully expect Tesla to introduce more sensors for redundancy, and to officially use HD maps (whether done by a third party or fleet learned) not because they really want to, but they'll be necessary.
The real path to autonomy isn't for a single company to magically achieve it, but for governments and corporations to work together on a solution.
DanCar
Active Member
I suppose FSD can make use of them for calibration and sanity testing. If the car is rumbling then disengage FSD and declare rumble fault.
It's not good enough for Waymo to be first. Hypothetically, if Waymo starts scaling out their product today with their current costs, I'm not sure they could even generate a profit. I estimate the current Waymo vehicles cost something like ~150k and they are not electric. It's about who is the more efficient provider of autonomy in the long run. Tesla is aiming to produce a 25k robotaxi that has a ridiculously low cost per mile due to being a BEV and only using mostly vision. Tesla could come to market 2 years after Waymo with disruptive pricing and beat them out. For Waymo, I'm not even sure they can currently underprice Uber.
Similar threads
