FloridaJohn
Member
Let's not be ridiculous here. How many of those Tesla cars have unoccupied driver's seats? The Waymo testing, be it in Chandler or elsewhere, is an entirely different type of testing than Tesla is doing at the moment.
-
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
mikes_fsd
Banned
The only place where Waymo has no drivers in the seat is in 50 sq miles of a desert. So, yes let's not be ridiculous here!
As for the rest of their "testing" we just have to take their word for it, don't we. Missing the entire point of the post.
FloridaJohn
Member
You are going to have to elaborate for me what the point of the post was then, because I completely missed it. As far as I can tell, Tesla is not doing any testing with no drivers in the seat.
mikes_fsd
Banned
Context is important.
I replied to @diplomat33 who said "Waymo is testing in 25 metro areas across 11 States in the US. So it is a big and diverse testing area. "
To which I replied that Tesla is publicly and without restriction testing in more states, and we get to watch the testing in glorious detail. (there are apparently more testers that are not allowed to publicly post about their testing)...
The jab about transparency was because we are supposed to deduce Waymo's safety record based on a stupid disengagement metric devised by bureaucrats...
Hi I thought id chime in to mess up the argument some more specifically that humans are a " two camera system" any of these systems should be able to do FSD. that is partially true but these days on both these systems the differences between human and automations are really starting to diminish. Waymo and Tesla I know know this it is but I thought I'd add just a little more to the general discussion and I am happy to see different approaches to see what works, what doesn't etc. The issue as you all know is rapid, simultaneous and accurate sensor fusion/correlation and the like , and now is seems 2-3D sensor stereo integration and rate learning, and predictive actionable information presentation ( realtime and learning based) is winning the day. so lets start with the human . The human is actually is many more sensors than two and ways to sensor integrate . The big one aside from Visual Color object sensing and classification and tracking ( motorcycle, person, blue versus grey trash can, big traffic cone versus small, curb, tree) is inertial guidance ( lane change, lateral G sense for turn rate, lane change and direction memory ( some better than others, just ask my teenager) . Don't forget road " feel", Speed to wind sound registration. and yes humans operate with an incredible amount distractions ( radio/ kids interrupts) , Flywheel processing/ Idling ( day dreaming, staring as something specific) . I always found it fascinating that we adapted by folding the spatial range of behind us (rearview mirrors) in and along side ( side mirrors) back into our two perception ranges - peripheral and direct) and the gimbal in our neck . again it generally works with our learned response mechanisms , - some better than others - just as a NASCAR or INDY driver. Humans shutdown pretty well though when we are in complete darkness. I think we would do a lot better if we had active lighting ( I dunno headlights / side lights, streetlights, brake lights anyone?) , I wouldn't mind multispectral and active 3-D radar and Lidar range data if I could find a way to put it in my head). instead I am limited to stereo object rate change data and limited visual acuity range ( I wear mutlifocals). I probably rely on my learned object predictive vector movement recognition ( ball and kid running into the street) and inertial more than I should. Recent progress the 2-3D stereo work multi camera and inertial will be the biggest players I think. if either group can figure out all the permutations of a left hand turn with multiple and simultaneously reducing lanes never seeing it before or a right hand turn into ongoing traffic when all given cues all fail we'll have something! No need to refute or say I missed some thing ( I did ) . just posting for your enjoyment on the comparison... patience! this is a hard problem!. ( yes I'm a systems engineer, life is a hard problem 
)
)
DanCar
Active Member
Yeah it is ridiculous to compare a couple of Vans testing in a small area versus a large fleet of cars testing across the entire country.
diplomat33
Average guy who loves autonomous vehicles
I am not talking about the CA DMV disengagement report. I am talking about the recent Waymo Report here:
Safety – Waymo
We don't have to deduce anything. Waymo published 3 documents, a safety report, a paper on their safety methodologies and a paper on their safety performance from 6M miles. Did you read these reports? It was not "stupid disengagement metric". The 3 papers give detailed data on Waymo's safety record. We know exactly how safe Waymo's FSD is and how Waymo's FSD works.
EVNow
Well-Known Member
I understand that - but saying they don't have any perception issues from these test runs is a big jump. After all first they capture everything in 3D, analyze it, label it before testing.
FloridaJohn
Member
I agree, especially since they are not testing the same thing. Waymo is testing vehicles without a driver. Tesla is testing vehicles with a driver. Comparisons between the two are not valid at this point in time. Hopefully they will be soon.
DanCar
Active Member
I agree, one is obviously much more impressive than the other. And we can take bets in which one will we get our first truly driverless ride in.
Last edited:
powertoold
Active Member
We can see grey stripes across a lane in the Waymo visualization to indicate a lane that is non driveable because of construction:
I'm confused. Aren't one of those lanes open for opposing traffic?
diplomat33
Average guy who loves autonomous vehicles
It does look like one lane might be open. The visualization does grey out both lanes. It might just be the visualization for "this road is being worked on", not literally which lane is driveable. Just speculation on my part.
In any case, it was something new in the visualization that I thought was worth sharing in this thread.
I was just checking out some other waymo videos and while impressive, some of the things that I noticed that it still struggles with such as stop signs makes me feel the progress with Tesla's FSD is truly remarkable when compared to how long Waymo is out there compared to FSD.
Case in point the video below at 1:39 and the guy sitting in comments too much traffic that did not look too much traffic to me by any measure.
Case in point the video below at 1:39 and the guy sitting in comments too much traffic that did not look too much traffic to me by any measure.
I think teleoperation and remote assistance give Waymo a crutch that we then perceive as advancement. But if you disabled teleoperation and remote assistance for a Waymo van, or if you added teleoperation and remote assistance to FSD Beta, how would they truly compare?
To me teleoperation and remote assistance is almost no different than having a safety driver or having driver assist when it comes to comparing the autonomous self driving car capability. To keep teleoperation as a necessary component of the system, they will have to solve an online global connectivity problem before they can solve a global self driving car problem. (Will they equip long-distance Waymos with Starlink?)
To me teleoperation and remote assistance is almost no different than having a safety driver or having driver assist when it comes to comparing the autonomous self driving car capability. To keep teleoperation as a necessary component of the system, they will have to solve an online global connectivity problem before they can solve a global self driving car problem. (Will they equip long-distance Waymos with Starlink?
)I was just checking out some other waymo videos and while impressive, some of the things that I noticed that it still struggles with such as stop signs makes me feel the progress with Tesla's FSD is truly remarkable when compared to how long Waymo is out there compared to FSD.
Case in point the video below at 1:39 and the guy sitting in comments too much traffic that did not look too much traffic to me by any measure.
First time I’ve seen a Waymo ride in action, so I realize how limited this observation is and any conclusions I make are constrained to that.
I think it’s remarkable that Tesla is doing this in the open with millions to watch the progress, the screw ups, and everything in between.
More importantly, having watched this single ride from Waymo (that clearly happened within the last month or so), I’m more bullish than ever in what Tesla is doing with their FSD program. I came away no more impressed with that Waymo ride than many of the Tesla FSD beta rides I’ve recently seen, and those occurred in non-geofenced areas with a rapidly developing and improving software. Simply, wow.
I’m more excited than ever for wider release of FSD beta.
diplomat33
Average guy who loves autonomous vehicles
I was just checking out some other waymo videos and while impressive, some of the things that I noticed that it still struggles with such as stop signs makes me feel the progress with Tesla's FSD is truly remarkable when compared to how long Waymo is out there compared to FSD.
Case in point the video below at 1:39 and the guy sitting in comments too much traffic that did not look too much traffic to me by any measure.
Waymo did not struggle with the stop sign in that scenario. Waymo has not trouble with stop signs at all. You can see the other car was signaling its intent to turn in front of the Waymo but did not follow through. So Waymo was waiting for the other car to turn in front of it. The hesitation of the other car is what "confused" Waymo about what to do.
Yes, Tesla's progress has been remarkable but Tesla is not able to do driverless FSD yet, like Waymo can.
One video is not enough to judge Waymo. The fact is that Waymo is doing driverless rides for months now, with real passengers and with no serious incidents. Just think, the fact what Waymo's FSD is driverless means that it can drive around on its own and does not need human intervention. That shows how reliable Waymo's FSD is. Unlike Tesla, you can summon a Waymo from across town and it can come pick you up with no driver inside. In contrast, Tesla's FSD screws up a lot. It still needs constant driver supervision.
Last edited:
I am not judging this based on only one video. I've seen several others and some other videos on youtube by the same guy also shows some other rides where the waymo takes a much longer route to avoid either traffic or unprotected left turns.
The video for that is below : At 2:09 he shows the mapview and it's not a normal route but a longer route. Then at 3:38 it takes another route through a neighborhood either to avoid traffic or unprotected left turns (even though he says he's seen it taking unprotected left turns before).
My point is of course Tesla can't do FSD yet and that's why we're required to pay attention. I am just saying from the FSD video's I've seen so far, Tesla doesn't have those same restrictions and the only one I know if is that it can't do U-Turns yet but that's fine with me. I still think Tesla seems to be catching up really fast and I think in the end will be successful.
This is what I gathered from the video. Especially when used in the context of an entirely geofenced area with remote operators monitoring.
diplomat33
Average guy who loves autonomous vehicles
Thanks the clarification.
I suspect the routing is to make things easier in order to lower potential risk, since there is no safety driver in the car to intervene. Waymo is very safety conscience since they have paying customers in their cars. But Waymo's FSD is more than capable of handling unprotected left turns.
It is easy for Tesla to test FSD everywhere since there is a driver in the seat able to take over if there is a problem. So it does not matter if the FSD would have failed. In fact, you want disengagements so you know what to work on to make your FSD better. In Waymo's case, there is no safety driver so there is nobody to take over if there is a problem. And they have paying customers in the car. So Waymo is more cautious.
To be clear, I am not suggesting that Waymo's FSD is perfect or anything. I acknowledge there are areas that Waymo still needs to work on. My perspective is that Waymo is ahead though. I've watched the Tesla FSD Beta videos and I see it make a lot of "rookie" mistakes that Waymo does not make. So yes, Tesla is making good progress but at the moment, they are behind Waymo.
Last edited:
Similar threads
