My interpretation of this is that the car is looking for opportunities to do maneuvers that require zero accommodation by other drivers or pedestrians and also leave a small margin of error. But what this video shows is that in busy scenarios, those opportunities sometimes do not exist. It is expected in these circumstances for vehicles to make maneuvers that actually could result in collisions if the other agents were not paying attention and accelerated at the wrong time, or sometimes just failed to brake slightly to let them in. The challenge which I am not sure people are realizing is that for the Waymo vehicle to be more aggressive means that it will be relying on the accommodation of the other drivers and pedestrians. This means that minor errors by the other party could result in collisions. So to drive "normally" and effectively in these busy scenarios actually requires a significant risk to be taken by Waymo. And if they do that continually, there is a statistical guarantee that even if the Waymo vehicle drives perfectly (but human-like, still allowing for some risk in terms of other drivers), one of these times the other driver will make an error that results in a collision. And the hard part is, that some people will then blame Waymo for not being more cautious. So they are kind of "damned if they do, damned if they don't" here. Which is not fair and points to human ignorance in my opinion.
The software used to expect accommodations. During testing a while back, it pulled in front of a bus which unsurprisingly didn’t accommodate.... I think they may have over compensated...
By not being more aggressive the Waymo vehicle is going to cause accidents by impeding the flow of traffic as well as by makung peoole frustrated. All driving relies on accomodation of other drivers, as you put it. You assume the other driver will respect the red light, for example. Self-driving cars are not ready for full roll-out until they can judge those situations well. they need to accurately judge what is a reasonable expectation of what the other driver will do. another possible mistake i noticed in the first video posted on this channel is that the waymo car apparently braked when a truck heading towards the side of the car didn't stop immediately at the intersection. it was hard to see, but it looked like the proper thing to do was to accelerate, not stop, to avoid potential danger.
@@RichardSpindler put that car in a busy city and it would go nowhere. in fact, there must be places in Phoenix where it can't turn out of a place for 15 minutes. it would never get on the southern state parkway in ny. it would just sit there and cause a traffic jam.
With each year this is gonna get better and better and slowly expand to more locations and then ten years from now we'll take self-driving taxis for granted lol
i dont think so. tesla and google have spent the last decade to get this far,level 2 automation and still there are many situations the car cannot handle,construction work,fog,snow,heavy rain,red signs that look like traffic lights etc...and what if the radar or cameras pack up? the car will just sit there,in the middle of a highway.
@@Heath-GallagherYes but the pace of AI advancement has accelerated in the last two years and shows no sign of slowing down. Most experts agree that the level of AI we’ll have in 10 years is literally unfathomable to us now. The problem with full self driving cars is entirely an AI problem right now, the hardware is there, it’s just waiting for the right “brain” to make it all work. That artificial general intelligence brain is a lot closer to becoming a reality than it seems right now.
I am impressed by the amount of autonomous driving that has been achieved. But in, for example, the Amsterdam city centre with bikes and trams or any other really busy place this car would go nowhere and politely wait all day.
Always interesting to watch how the autonomous car handles different situations in the real world. I've learned that I would not want to use a robo ambulance if I were injured.
I think there is a future for self driving vehicles. Just not ones that have been adapted or designed from existing ones. I believe the roads will be reclaimed for non traffic use, residential, recreational or industrial purposes. SDV's will have their own separate paths thus speeding up journey times and increasing public safety.
Thanks for the video. It handled all the situations, albeit a bit slowly. It's very defensive and patient with its maneuvers, which is great, but not as aggressive as a human driver.
Though it has a few glitches, this is fantastic! I would gladly, happily, and willingly sit through these glitches rather than drive myself through the chaos of a city as big as Phoenix. Go Waymo!
Good one. With regard to those parents with kids the car needs the robotic equivalent of a smile and a hand wave to let them know that the car sees them and they should go ahead and cross the street. Also lesson learned, don't ask to be picked up in the crosswalk in front of a grocery store.
Ah, see that crosswalk was the whole point. I wanted to see how it would react :D Yea and I think that's what these new screens going into the cars are, the digital equivalent of a hand wave
As a customer, for sure. As a developer, and knowing this this still under development, this is the *ideal* place to ask it for a pick up, to test it and see how it’d react (assuming it’s far enough along, which it is, best to stall than to run someone over). This reminds me of an old QA engineer joke: “A software QA engineer walks into a bar. He orders a beer. Orders 0 beers. Orders 99999999999 beers. Orders a lizard. Orders -1 beers. Orders a ueicbksjdhd. First real customer walks in and asks where the bathroom is. The bar bursts into flames, killing everyone.”
Wow, an action packed video so far. I just made it to 9:30 so far. Lots of real world scenarios! The insanely busy parking lot! The busy traffic which is unusual compared to many of your other videos. Trying to merge after the stop light blocking the bus for a bit. Waiting for the kids with the grownups. It was extremely cautious with the pedestrians which is a good thing. The whole reason for going into the neighborhood was to change directions. Are U-turns legal on these streets? I assume most drivers make U-turns all the time. I wonder when/if Waymo will start making U-turns!
Most of the traffic issues it had won't exist when all cars are self driving as self driving cars will let each other out. I wonder if there could be a trial whereby everyone in a town is given a self driving car for a year and see how they perform together (difficult as this would seriously dent the male ego in that town for a while)
Love to see these videos. Seems like we're still a long way from this becoming mainstream. I can't be calling a ride and sitting in the parking lot for 5 minutes while it figures out what to do.
Thanks for sharing. Definitely nifty to see Waymo's ongoing approach to self-driving. The visualizations are great, but man is the vehicle way too cautious/hesitant in situations. Hopefully, they manage to improve this, because a hesitant driver is not a safe driver, and makes other people around them act dangerously.
@jshowa o If you think that someone who's hesitant and therefore less confident and predictable behind the wheel of a 2-ton death machine is a safe driver, I don't know what you're talking about, and I'm quite sure you don't know what you're talking about either.
@@bearcubdaycare Yes, depending entirely on everyone and everything else to be totally predictable isn't safe or reasonable, but that wasn't what I was saying either. There's a baseline expectation of rule-following (stay in your lane, signal when changing lanes, stop at stop signs and red lights, etc.) and defensive driving can bridge the gap to a degree, but no driver on Earth, no matter how talented, could safely and reliably make their destination if all driver behavior was 100% random. Someone who's driving hesitantly is more erratic and unpredictable than a confident driver, and is therefore more dangerous. So Waymo's driver needed to be less hesitant to increase safety. Hopefully in the year that has passed, they've done that.
Once the self driving car can communicate with other vehicles and stop lights, the journey will be much smoother. Other sd cars will slow to allow lane access. Shame pedestrians can't be fitted with self driving technology!
19:55 Good maneuver to brake for that truck and interesting visualizations with the yellow circle and outline around the truck. I had to check again to see if your car could move over slightly but it was right next to the curb on the right. If I’m in my Tesla and if it brakes hard, I’m worried about getting rear ended.
Yeah, but there's a reason that's it's just in one part of Phoenix even after years. The roads in this video, and the conditions, don't look anything like those in my town. Or most other towns that I've been to. Apparently even a handful of orange cones for construction flummox it, to the point that it just stops in traffic for minutes. Then randomly starts and stops.
Drive-throughs are honestly a fascinating use case, because you have to wait for an interaction partially inside the car to finish to move forward. I have no idea what if anything Waymo is planning but I wouldn't be surprised if, were they to eventually support using a drive-through, it had an "I'm done with this step, go to the next step" button you needed to push after every window (order, get food, pay). But I doubt that will come until after they add support for intermediate stops, because you presumably need to go somewhere *after* the drive-through.
My boy friend wrote the code (part of it), I am so proud and in tears watching this. He works so hard everyday and I take care of his lunch during the lockdown. I am part of this success too!
There will be no issues if everyone follows the rule or almost no one follows the rules. In either situation decisioning is easy. With 50-50 it gets difficult to run logic even for humans. As most mobility becomes autonomous the electronics will be able to signal better and provide faster movement.
Being an engineer, it is cool to see the things we can achieve with the help of technology but do we really need to go driverless? What's wrong If a taxi driver makes some money while taking somebody from point A to B. Waymo has started. Tesla, Uber and Lyft will follow. As much as we teach machines to make decisions they will get better in time and will soon replace us. AI is coming soon. I am not sitting in taxi unless a human is driving it.
Pretty cool, but like I've been saying for the past two years, this wont be ready for prime time for at least a decade. Its fine for open roads, but in real New York traffic this would be a nightmare.
11:27. The car didn’t seem to react to another car’s blinkers on. We had discussed if Waymo sees blockers or not. It seems like for a human driver that we rely on blinkers to predict what other cars will probably do. But many (bad) drivers don’t use blinkers so you can’t rely on it.
I know Teslas don't know what a blinker is. They're just trained on if the car changes lanes or not. I assume Waymo does the same. That way the neural net can consider the blinker, driver behavior, and road layout all together to make a decision.
@@keco185 this makes sense, but I wonder if in the future self driving cars will recognize and react to blinkers. On a side note, I wonder if human drivers will take advantage of robotic drivers by cutting them off knowing that the computer WILL NOT HIT THEM. Computer driven cars will slam on the brakes and not hit the rude driver. I can imagine human drivers knowing this and taking advantage of the situation.
I think it actually does notice it. Look how the car gets a tiny yellowish frame around it as soon as the car goes past Waymo. But, the car never eases into the lane (enough space in front to do so), so Waymo did not have to do anything but instead was prepared and did as we human does, wait to see if it a false signaling or not. You can see in the behavior if someone want's to merge in.
@@Luke_Starkenburg Waymo is professional. And if you are professional you MUST 'see' when someone is indicating. That is like the whole point of an indicator, to help other drivers. Sadly, some people (often in expensive BMW, Audi or similar) have something wrong with their indicator. For BMW, i think it is just the blinker fluid that is bad or out but for Audi and similar, i guess it was an option when they bought the car, and they did not have the money to buy it so the car just lacks that add on.
This is perfect. Future is now. How will you say this compares to tesla autopilot, particularly latest version? And how limited is waymo in its destinations?
Self driving cars is no where near being ready for even testing like this on normal chaotic public roads in city traffic (or nowhere else where unexpended thigs can happen), this is complete insanity!
- These calls when you check in are creepy. They really need to stop that. The only thing I can think off why they'd do this: Checking out if there's the correct person in the car? - 2:00 if it keeps blinking right, everyone else will start overtaking, obviously. It should really stop blinking, so other cars know it's going to continue - not even sure if that pickup spot is very clever? Is it allowed to stop there for some time? Nice video for testing the limits!
I assume the reason is that most people have nothing to do when they get in except put on a seat belt, and he has to take a short time to set up the camera. They probably have it tuned to detect any delay in start ride as a potential issue.
@@runvnc208 I think this is it. Cut in video was probably some fluffing around, so they just check up. I imagine some people would easily forget to push the go button.
Poor thing doesn’t know how to preplan, and completely screwed up that merger at 6:40 ... I was surprised it performed so poorly. I was impressed with the smoothness of the controls modulation. It was very precise in the turns which is something with which Tesla is still struggling ...
Just wondering. After waymo launch their driverless service in Phoenix for a few weeks, why is it that I practically didn’t see any other waymo videos from anyone else? Is it that you need some special permission to make a video on waymo rides ?
Not sure if it matters, but at 9:01 the car's screen does not show the trash can at the side of the road. That is a limitation of pre-mapped routes, it cannot take into account something that is not permanently located. Trash cans don't suddenly start moving on their own (unless the wind pushed it), but what if the object was capable of moving, but was never mapped?
I know 100% that it sees the trash cans, since they show up on the LiDAR point cloud if they're off to the side of the road, but not on the road. It just doesn't display them, I promise it goes around the parked cars, and it goes around trash cans. 👍 I submitted that as feedback to Waymo, hope they add that to the visualization
Another question is if the Waymo car see's brake lights or not. I don't think it does since it measures the distance between it and the car in front of it many times a second. There is no need to see the brake light because it can see the car's movements so precisely.
Very interesting vid. They drive quite differently to the current Tesla FSD. Smoother and more consistent steering especially. How is it with things appearing on the road? People, leaves, road cones etc?
I'm guessing this is Phoenix? I've lived all over the southwest but I can literally never tell the difference between Phoenix and Vegas in street level videos. They look nearly identical without landmarks.
You sure are testing the limits (your choice of starting location with painted stripes, people on phones crossing, cars going around you and at an intersection within the parking lot - all real-world clearly) and agree with your assessment that the algo could be a little more aggressive there. Too bad there is no horn button for the passenger!
This is amazing. It is already a beta, fully functional ride sharing autonomous service. Ready to be expanded. TSLA still requires the driver to keep their hands on the wheel.
@@Luke_Starkenburg as far as I know they just give it instructions as opposed to taking direct control. It asks a question or tells the controllers what's confusing it and they answer
I think for the really busy parking lot situation, the car should make small, slow movements in the direction it wants to go in order to indicate to other drivers (and pedestrians) what its intentions are, especially in scenarios where if it is moving forward, it has the right of way (e.g. parking lot rows meeting up with the main drive in front of the store). At least, that's what I do. Really, really slow inching forward to assertively, but safely, get through the mess. I'm not sure what the driving laws are in AZ, but some states stopping like that on a lane merge isn't legal. I'm assuming Waymo has taken that into account though. Even if not, that can be even more dangerous than being more assertive in merging. Perhaps it should have tried to get into the other lane before the light. Or sped up faster to match the speed of the cars in the adjacent lane so it could merge more smoothly. I think this was an excellent video that 1) illustrates how hard this problem is, 2) how much continued work is needed to refine the driver and 3) the trade offs between safety and "functionality" when there are unpredictable humans in the loop (environment). Erring on the side of caution is laudable, understandable, and IMO the right choice. But it also means that creating a scaled and reliable service will be a much slower process.
Their models are designed for a world of unpredictable human drivers. Even when the world goes autonomous, their systems will get walked all over by other systems.
The difference between this and Tesla "Full Suicide Driving" is astounding. This is clearly the way forward. I'm sure there will be challenges in mapping, and making the sensor suite affordable. But it makes all other systems look like a joke. An incredibly dangerous, irresponsible and pointless joke.
My only question is whether the cops can pull them over for a moving violation or for a defective brake light etc and if they can who do they talk to as well as give the ticket to especially if there's no passenger in the car.
This is very interesting. If other drivers were less aggressive, the Waymo could continue uneventfully. The impatience of other drivers and aggressive behaviour is hampering the Waymo. For example, once the Waymo signaled its intent to merge left with the lane running out, other drivers should have yielded. Instead, other drivers closed the gap making it impossible to merge. If everyone drove like Waymo, there would be less stress, less accidents, and everyone would still get to their destinations safely.
14:22 That car on the left with no lights is not parked! Autonomous cars should *BRIEFLY!* honk and flash their high beams at people who forgot to turn on their lights. It doesn't need to be aggressive. Honking can be informative or advisory if you do it briefly enough, and it might help to keep someone from harm. In fact, that's what the car horn's intended purpose is. 15:33 There's another one! Do you guys not warn each other when you see someone driving at night with no lights? 17:37 And another one! People, the lights aren't just for you to see where you're driving, they are also for you to be seen by people who need to yield to you. Don't wait until after it's completely dark to turn them on.
When you’re a passenger and you have to wave “sorry” at motorists and roll down your windows and say “sorry” to pedestrians, something is VERY wrong. Bottom line is, these things are an absolutely terrible idea, they clearly do not work, and they’re gonna get a LOT of innocent people killed if their presence in the roads increases (which I doubt they will increase). There are just too many unknown variables to account for in population dense areas, and these things are dangerous. All these things do is confuse and/or frustrate the people around them.
Excellent video, it shows that Waymo needs to read the traffic a lot better. It needs to see opportunities and predict where cars and pedestrians will be in a second. Tesla FSD also has its issues with traffic and it wouldn't probably handle it any better.
I guess there is still no substitute for intuition and experience in traffic. Also, in a lot of cases (if it is not dark) I am looking at the face of a driver that is coming from the right or a pedestrian, just to note if they see you. Many times you can predict a manoeuvre somebody is going to make before they actually make it. This is something that an autonomous system lacks.
@@JJRicks Tesla AP feels very human like in terms of merging between cars, downside it will make a mistake with the same solid feel.. Waymo seems to be taking a different approach here..
@@teslascoop2177 the problem with this is that a bunch of people who only watch this video just see this super hard situation I threw at it, and not the other perfect videos. So the they immediately pass judgement
@@JJRicks Well this is how you judge what good is 1,000,00 miles of easy straight line driving right? BTW I'm referring to merging not accel/brake etc. Just noting the difference between systems..
For next time when having the frame problems just stretch one of the videos till they are the same lenght in total! Then after go edit the video so everything is in sync! Just at tip! Im loving the vids just can´t imagine this happening in Amsterdam here with all the bikes haha!
This is AMAZING. The lives it will save. Out drinking. Just call your car from your garage to pick you up. No more Uber/Lyft drivers. More safety. Less truck accidents on the highway. Wonder how they will handle BLACK ICE.
Am I the only one who thinks this system is beta? Threading into the track took forever. The bus driver behind it has already honked. With pedestrians nearby, the system doesn't seem to know what to do and just stops. That would frustrate me as an inmate. Especially when I'm in a hurry.
I wonder if a police car can stop this car from behind just with visual and acustic signals. I would also like to see the officer's face when he wants talk with the driver. :-)
So many people just walking in front of the Waymo, preventing it from starting the drive! 😂 Did the car fail to indicate its intention to depart? Or is it more that most folk now know the car can't (hopefully 🤞) run them over? I feel they're going to have to up the aggressiveness algorithms. Maybe the solution is for the AI to ignore every 50th person walking in front?... Just to keep everyone on their toes! 😉
Hahahahahaha yea, I'm mostly not annoyed, really the fact that it made it out at all is a huge testament to progress. Though people walking in front is definitely annoying
People are impatient. In a grand scheme of things, minute or two of delay, is not big deal. The car handled everything well. I am not sure why it had emergency lights on , maybe it was related to the call.
The mapping is excellent. But consider how much it cost, then the cost of mapping the land area of the planet! Then keeping it updated on a daily basis. Not viable, IMO.
You don't need to map the planet for this to be useful technology, they can just launch in specific cities. A taxi service that can take you anywhere within new york city, for example, would be incredibly valuable. In terms of the cost, keep in mind the cost of mapping the area is offset by the savings from not having to pay any drivers. Let's say a company had a fleet of 1000 human-operated taxis operating in new york city. Each driver makes an average of $15/hour and works 40 hours per week. The weekly cost of labor would be $15/hour * 40 hours * 1000 drivers = $600,000 per week. Per year it would be $32 million. As long as Waymo is able to maintain their mapping of NYC for less than $32 million, they would be able to come out ahead of the human-operated taxi company.
@@kevinaud6461 There ARE the other costs - updating the maps on, to be effective, an almost daily basis, then software engineers, computer maintenance, cleaning (which drivers do as and when they have no fare) and so on. Not to mention having drivers available to go cross-city to extricate the vehicle from situations where it is stuck - and someone to drive the rescue car back or just find it a parking space. Autonomous cars might work on a large scale in areas designed for them but.......
@@Wol747 Keeping the map up to date doesn't involve re-mapping the area daily, it just involves re-mapping the parts that have changed. See the "Keeping our maps up to date" section of this blog post: blog.waymo.com/2020/09/the-waymo-driver-handbook-mapping.html Basically the cars cross-reference real-time data with map data and if they find a discrepancy then it gets reported to HQ at which point they re-map it. Only having to map the things that have actually changed makes it pretty scalable.
Thanks for sharing! This is interesting! The lidar sweep shown on the screen is slow occurring every 4s or so. In actuality do you know how fast it is sweeping/rotating?
The AI needs to study lots and lots of and lots of actual drivers’ behavior in parking lots to learn how to handle these scenarios. Obviously they haven’t don’t that yet.
@@bmaiceman Already made an app video, all you have to do is scroll through my videos a bit. ruclips.net/video/L8HLrxJp2n8/видео.html also JJRicks.com/faq for more questions
Feels like waymo has quiet a way to go too. I'm actually surprised that they allowed them to drive it without anyone needed to take control. Is this limited to special streets/areas or can you really run it everywhere? Considering you started on a parking lot I'd guess it can just drive whereever. I'm not sure I could say this is better than FSD though, considering some comments made by waymo officials. I like the visualization a bit more, but then again you wouldn't watch that anyway in a final product, which would probably have passengers facing each other unless there is some safety issue with sitting backwards, but I think it's not a big concern. in the future we will probably see 4-5 seaters with a nice table and you can play cards together while driving :D
They drive in many cities, but are only open to passenger service in Chandler. Rider support can't joystick the car, they can only give it pointers like "turn here" or "ignore that thing"
how many times has a line of code failed. This is a stupid idea with way to much risk and only the perceived benefit of putting taxi drivers out of a job. It's just not worth it. Sure its cool technically to be able to, but I dont think people will want this.
My interpretation of this is that the car is looking for opportunities to do maneuvers that require zero accommodation by other drivers or pedestrians and also leave a small margin of error. But what this video shows is that in busy scenarios, those opportunities sometimes do not exist. It is expected in these circumstances for vehicles to make maneuvers that actually could result in collisions if the other agents were not paying attention and accelerated at the wrong time, or sometimes just failed to brake slightly to let them in.
The challenge which I am not sure people are realizing is that for the Waymo vehicle to be more aggressive means that it will be relying on the accommodation of the other drivers and pedestrians. This means that minor errors by the other party could result in collisions. So to drive "normally" and effectively in these busy scenarios actually requires a significant risk to be taken by Waymo. And if they do that continually, there is a statistical guarantee that even if the Waymo vehicle drives perfectly (but human-like, still allowing for some risk in terms of other drivers), one of these times the other driver will make an error that results in a collision. And the hard part is, that some people will then blame Waymo for not being more cautious. So they are kind of "damned if they do, damned if they don't" here. Which is not fair and points to human ignorance in my opinion.
The software used to expect accommodations. During testing a while back, it pulled in front of a bus which unsurprisingly didn’t accommodate.... I think they may have over compensated...
By not being more aggressive the Waymo vehicle is going to cause accidents by impeding the flow of traffic as well as by makung peoole frustrated. All driving relies on accomodation of other drivers, as you put it. You assume the other driver will respect the red light, for example. Self-driving cars are not ready for full roll-out until they can judge those situations well. they need to accurately judge what is a reasonable expectation of what the other driver will do. another possible mistake i noticed in the first video posted on this channel is that the waymo car apparently braked when a truck heading towards the side of the car didn't stop immediately at the intersection. it was hard to see, but it looked like the proper thing to do was to accelerate, not stop, to avoid potential danger.
I wish all drivers would drive like that.
Also, in the end the car got it. It is just human drivers that get impatient to quickly.
@@RichardSpindler put that car in a busy city and it would go nowhere. in fact, there must be places in Phoenix where it can't turn out of a place for 15 minutes.
it would never get on the southern state parkway in ny. it would just sit there and cause a traffic jam.
"Good to go." Suddenly, nothing happens.
It should be a little more aggressive than that
ERROR 404
so it goes back completely empty, no human in side whatsoever ? OMG that's so cool
This is great. System was challenged a lot.
Does the car learn overtime
Yes
@@KTMCDO Waymo trains the computer with simulations.
Does the car get paid overtime while learning? That would explain a lot. ; )
@@afusdoifudo3563 Not at all! They’re the perfect employee. No pay and never complain.
With each year this is gonna get better and better and slowly expand to more locations and then ten years from now we'll take self-driving taxis for granted lol
i dont think so. tesla and google have spent the last decade to get this far,level 2 automation and still there are many situations the car cannot handle,construction work,fog,snow,heavy rain,red signs that look like traffic lights etc...and what if the radar or cameras pack up? the car will just sit there,in the middle of a highway.
@@Heath-GallagherYes but the pace of AI advancement has accelerated in the last two years and shows no sign of slowing down. Most experts agree that the level of AI we’ll have in 10 years is literally unfathomable to us now. The problem with full self driving cars is entirely an AI problem right now, the hardware is there, it’s just waiting for the right “brain” to make it all work. That artificial general intelligence brain is a lot closer to becoming a reality than it seems right now.
Waymo has come such a long way, its so fun to see
I am impressed by the amount of autonomous driving that has been achieved. But in, for example, the Amsterdam city centre with bikes and trams or any other really busy place this car would go nowhere and politely wait all day.
Always interesting to watch how the autonomous car handles different situations in the real world. I've learned that I would not want to use a robo ambulance if I were injured.
I think there is a future for self driving vehicles. Just not ones that have been adapted or designed from existing ones.
I believe the roads will be reclaimed for non traffic use, residential, recreational or industrial purposes. SDV's will have their own separate paths thus speeding up journey times and increasing public safety.
Imagine a driver that treats pedestrians with respect and courtesy! Another excellent test, thx Joel
How about pedestrians that don't try to own the road when they don't have the right of way at all times
Thanks for the video. It handled all the situations, albeit a bit slowly. It's very defensive and patient with its maneuvers, which is great, but not as aggressive as a human driver.
I will take this one over driving any day. I am ok with a bit of waiting. Safety first, safety second, coolness third XD
This car at least cares about pedestrian.
Though it has a few glitches, this is fantastic! I would gladly, happily, and willingly sit through these glitches rather than drive myself through the chaos of a city as big as Phoenix. Go Waymo!
Good one. With regard to those parents with kids the car needs the robotic equivalent of a smile and a hand wave to let them know that the car sees them and they should go ahead and cross the street. Also lesson learned, don't ask to be picked up in the crosswalk in front of a grocery store.
Ah, see that crosswalk was the whole point. I wanted to see how it would react :D Yea and I think that's what these new screens going into the cars are, the digital equivalent of a hand wave
As a customer, for sure. As a developer, and knowing this this still under development, this is the *ideal* place to ask it for a pick up, to test it and see how it’d react (assuming it’s far enough along, which it is, best to stall than to run someone over). This reminds me of an old QA engineer joke: “A software QA engineer walks into a bar. He orders a beer. Orders 0 beers. Orders 99999999999 beers. Orders a lizard. Orders -1 beers. Orders a ueicbksjdhd. First real customer walks in and asks where the bathroom is. The bar bursts into flames, killing everyone.”
Even if I admire the incredible technical prowess, I think self-driving cars with humans driving other cars will never catch on.
They already have "caught on". What are police going to do, ticket software?
Wow, an action packed video so far. I just made it to 9:30 so far. Lots of real world scenarios! The insanely busy parking lot! The busy traffic which is unusual compared to many of your other videos. Trying to merge after the stop light blocking the bus for a bit. Waiting for the kids with the grownups. It was extremely cautious with the pedestrians which is a good thing. The whole reason for going into the neighborhood was to change directions. Are U-turns legal on these streets? I assume most drivers make U-turns all the time. I wonder when/if Waymo will start making U-turns!
As long as there's no sign saying otherwise, u-turns are allowed, yup! I think it does, but only in weird cases
Most of the traffic issues it had won't exist when all cars are self driving as self driving cars will let each other out.
I wonder if there could be a trial whereby everyone in a town is given a self driving car for a year and see how they perform together (difficult as this would seriously dent the male ego in that town for a while)
Love to see these videos. Seems like we're still a long way from this becoming mainstream. I can't be calling a ride and sitting in the parking lot for 5 minutes while it figures out what to do.
Thanks for sharing. Definitely nifty to see Waymo's ongoing approach to self-driving. The visualizations are great, but man is the vehicle way too cautious/hesitant in situations. Hopefully, they manage to improve this, because a hesitant driver is not a safe driver, and makes other people around them act dangerously.
Waymo: But, but, but, I’ve never had an accident !
Evaluators: Okay, but how many have you caused??
@@sking2173 This 100%
@jshowa o If you think that someone who's hesitant and therefore less confident and predictable behind the wheel of a 2-ton death machine is a safe driver, I don't know what you're talking about, and I'm quite sure you don't know what you're talking about either.
@@TheDavidMetcalfe If you're depending on other drivers and objects being predictable, you're not driving safely. The wildlife around here isn't.
@@bearcubdaycare Yes, depending entirely on everyone and everything else to be totally predictable isn't safe or reasonable, but that wasn't what I was saying either. There's a baseline expectation of rule-following (stay in your lane, signal when changing lanes, stop at stop signs and red lights, etc.) and defensive driving can bridge the gap to a degree, but no driver on Earth, no matter how talented, could safely and reliably make their destination if all driver behavior was 100% random. Someone who's driving hesitantly is more erratic and unpredictable than a confident driver, and is therefore more dangerous. So Waymo's driver needed to be less hesitant to increase safety. Hopefully in the year that has passed, they've done that.
Once the self driving car can communicate with other vehicles and stop lights, the journey will be much smoother. Other sd cars will slow to allow lane access. Shame pedestrians can't be fitted with self driving technology!
Wow that was so frustrating to watch.
These vids are so addictive. Tyfs
19:55 Good maneuver to brake for that truck and interesting visualizations with the yellow circle and outline around the truck. I had to check again to see if your car could move over slightly but it was right next to the curb on the right. If I’m in my Tesla and if it brakes hard, I’m worried about getting rear ended.
That lane merge 😻
so slow
@@SmartKeyboard2011 😂
Would love to see something like this where I live don't feel like driving just summon a self-driving taxi instead.
Yeah, but there's a reason that's it's just in one part of Phoenix even after years. The roads in this video, and the conditions, don't look anything like those in my town. Or most other towns that I've been to. Apparently even a handful of orange cones for construction flummox it, to the point that it just stops in traffic for minutes. Then randomly starts and stops.
I’m in that lowes like 2-3 times a week. I was actually looking for my truck in your video 😆
Nothing like a good game of where's Waldo😂😂 though I guess, where's Nathan is more like it🤔
That first parking lot.... geez!
Thanks Skynet!
I can see that the aggressive drivers are going to hate this on the road in the future but who cares
Please be sure to train the Waymo to go thru Starbucks😍
Haha eventually ;)
Drive-throughs are honestly a fascinating use case, because you have to wait for an interaction partially inside the car to finish to move forward. I have no idea what if anything Waymo is planning but I wouldn't be surprised if, were they to eventually support using a drive-through, it had an "I'm done with this step, go to the next step" button you needed to push after every window (order, get food, pay). But I doubt that will come until after they add support for intermediate stops, because you presumably need to go somewhere *after* the drive-through.
My boy friend wrote the code (part of it), I am so proud and in tears watching this. He works so hard everyday and I take care of his lunch during the lockdown. I am part of this success too!
Heyy thanks for watching! Doing good work over there
The word has it that Waymo is still trying to merge.
Haha, maybe
There will be no issues if everyone follows the rule or almost no one follows the rules. In either situation decisioning is easy. With 50-50 it gets difficult to run logic even for humans. As most mobility becomes autonomous the electronics will be able to signal better and provide faster movement.
Being an engineer, it is cool to see the things we can achieve with the help of technology but do we really need to go driverless? What's wrong If a taxi driver makes some money while taking somebody from point A to B. Waymo has started. Tesla, Uber and Lyft will follow. As much as we teach machines to make decisions they will get better in time and will soon replace us. AI is coming soon. I am not sitting in taxi unless a human is driving it.
When AI learns how to use the heat humans generate we will all be doomed.
If this thing self drives in New Delhi, Waymo will be WayNoMo.
LOL omg that merge... folks... wow.
Pretty cool, but like I've been saying for the past two years, this wont be ready for prime time for at least a decade. Its fine for open roads, but in real New York traffic this would be a nightmare.
I can see bicycle messengers being the ultimate test.
It already drives way better than a lot of people I know.
Watch the Tesla fsd videos they are driving all over the country not just in fenced areas.
@@grandelDR Yeah, no it's not. It's awesome and has been improving in bounds since it was released in October.
Fsd is so much better at driving than this.
11:27. The car didn’t seem to react to another car’s blinkers on. We had discussed if Waymo sees blockers or not. It seems like for a human driver that we rely on blinkers to predict what other cars will probably do. But many (bad) drivers don’t use blinkers so you can’t rely on it.
I know Teslas don't know what a blinker is. They're just trained on if the car changes lanes or not. I assume Waymo does the same. That way the neural net can consider the blinker, driver behavior, and road layout all together to make a decision.
@@keco185 this makes sense, but I wonder if in the future self driving cars will recognize and react to blinkers. On a side note, I wonder if human drivers will take advantage of robotic drivers by cutting them off knowing that the computer WILL NOT HIT THEM. Computer driven cars will slam on the brakes and not hit the rude driver. I can imagine human drivers knowing this and taking advantage of the situation.
I think it actually does notice it. Look how the car gets a tiny yellowish frame around it as soon as the car goes past Waymo. But, the car never eases into the lane (enough space in front to do so), so Waymo did not have to do anything but instead was prepared and did as we human does, wait to see if it a false signaling or not. You can see in the behavior if someone want's to merge in.
@@Xanthopteryx Oh, I didn't notice that until you pointed it out to me.
@@Luke_Starkenburg Waymo is professional. And if you are professional you MUST 'see' when someone is indicating. That is like the whole point of an indicator, to help other drivers. Sadly, some people (often in expensive BMW, Audi or similar) have something wrong with their indicator. For BMW, i think it is just the blinker fluid that is bad or out but for Audi and similar, i guess it was an option when they bought the car, and they did not have the money to buy it so the car just lacks that add on.
This is perfect. Future is now. How will you say this compares to tesla autopilot, particularly latest version? And how limited is waymo in its destinations?
Self driving cars is no where near being ready for even testing like this on normal chaotic public roads in city traffic (or nowhere else where unexpended thigs can happen), this is complete insanity!
- These calls when you check in are creepy. They really need to stop that. The only thing I can think off why they'd do this: Checking out if there's the correct person in the car?
- 2:00 if it keeps blinking right, everyone else will start overtaking, obviously. It should really stop blinking, so other cars know it's going to continue
- not even sure if that pickup spot is very clever? Is it allowed to stop there for some time?
Nice video for testing the limits!
Yea, for me they're just less creepy and more annoying. If I need help, I'll push the button :P
I assume the reason is that most people have nothing to do when they get in except put on a seat belt, and he has to take a short time to set up the camera. They probably have it tuned to detect any delay in start ride as a potential issue.
@@runvnc208 I think this is it. Cut in video was probably some fluffing around, so they just check up. I imagine some people would easily forget to push the go button.
Yup, exactly! Wasn't all that long, but yea pretty much
Poor thing doesn’t know how to preplan, and completely screwed up that merger at 6:40 ... I was surprised it performed so poorly.
I was impressed with the smoothness of the controls modulation. It was very precise in the turns which is something with which Tesla is still struggling ...
That thing is a menace.
so many here prefer agresiveness n dangerous driving lol
that thing is the only thing this car doesnt have
I hate that parking lot with a passion lolol. Especially by the new Starbucks and ATT that was put in. Got hit twice there.
Is getting real next will be fly cars by year 2100 or before.. 👍🏼
Definitely before.
"your car in is good to go" three hours later when the parking lot empties at closing time the car moves lol
Oh hey lol, check my recent videos
Just wondering. After waymo launch their driverless service in Phoenix for a few weeks, why is it that I practically didn’t see any other waymo videos from anyone else? Is it that you need some special permission to make a video on waymo rides ?
Yea I'm in the public program, and I guess everyone else in the public program doesn't care ¯\_(ツ)_/¯
@@JJRicks cool. I’m still waiting to see how the system react in different weather conditions 🙂
Me too
I have seen Waymo in comment sections of other people asking for permission to use self-recorded video clips a couple times already.
Haha they asked me too, but over twitter
Not sure if it matters, but at 9:01 the car's screen does not show the trash can at the side of the road. That is a limitation of pre-mapped routes, it cannot take into account something that is not permanently located. Trash cans don't suddenly start moving on their own (unless the wind pushed it), but what if the object was capable of moving, but was never mapped?
I know 100% that it sees the trash cans, since they show up on the LiDAR point cloud if they're off to the side of the road, but not on the road. It just doesn't display them, I promise it goes around the parked cars, and it goes around trash cans. 👍 I submitted that as feedback to Waymo, hope they add that to the visualization
Another question is if the Waymo car see's brake lights or not. I don't think it does since it measures the distance between it and the car in front of it many times a second. There is no need to see the brake light because it can see the car's movements so precisely.
¯\_(ツ)_/¯
What would happen if another car backed up
It would stop (source: seen it happen)
@@JJRicks your both stopped at a red light but the guy in front of you backs up
Not much it could do except sit there
The sad, but good thing is IT WILL get old haha
Car has mind on its own lol 😆😆😆
Very interesting vid. They drive quite differently to the current Tesla FSD. Smoother and more consistent steering especially. How is it with things appearing on the road? People, leaves, road cones etc?
I would say, excellently! photos.app.goo.gl/kwcEyFovMuTxmUNDA
I'm guessing this is Phoenix? I've lived all over the southwest but I can literally never tell the difference between Phoenix and Vegas in street level videos. They look nearly identical without landmarks.
Chandler, so yes
The occasional Point Cloud Projections are interesting. I still don't understand the point of SDV
You sure are testing the limits (your choice of starting location with painted stripes, people on phones crossing, cars going around you and at an intersection within the parking lot - all real-world clearly) and agree with your assessment that the algo could be a little more aggressive there.
Too bad there is no horn button for the passenger!
Johny cab: Hell of a day...
While there are still lots of bugs it's pretty clear Waymo is WAAAAAY ahead of the competition.
@@snebold You can't run a Tesla without a driver at all and with their current sensor suite you never will.
This is amazing. It is already a beta, fully functional ride sharing autonomous service. Ready to be expanded. TSLA still requires the driver to keep their hands on the wheel.
@@andrasbiro3007 Yes, but which one is going to go to market with a ride hailing service first do you bet? TSLA or Waymo?
Well, everyone else is just wasting their time then. Tesla is taking over the whole world. Tell that to the 20+ companies working on FSD.
I can't wait to see the TSLA ride hailing service. Looking forward to it! Next week though, I'll be in San Francisco and will try the Waymo service.
Increíble algún día me gustaría viajar en un coche de esos
It's like 150 year old grandma driving.
I guess in the beginning the waymo team had to instruct the car on what to do?
Probably
I wonder if they would ever consider remote controlling the car or if they strictly rely on the self driving software.
@@Luke_Starkenburg as far as I know they just give it instructions as opposed to taking direct control. It asks a question or tells the controllers what's confusing it and they answer
Yea this ^^^^ I talked to a former safety driver and they confirmed
so there's a support person always looking at ride remotely ?
When will they be available to public to buy and ride share all over United States we are eager for this 🙂❤️
Nobody knows for sure, unfortunately
I think for the really busy parking lot situation, the car should make small, slow movements in the direction it wants to go in order to indicate to other drivers (and pedestrians) what its intentions are, especially in scenarios where if it is moving forward, it has the right of way (e.g. parking lot rows meeting up with the main drive in front of the store). At least, that's what I do. Really, really slow inching forward to assertively, but safely, get through the mess.
I'm not sure what the driving laws are in AZ, but some states stopping like that on a lane merge isn't legal. I'm assuming Waymo has taken that into account though. Even if not, that can be even more dangerous than being more assertive in merging. Perhaps it should have tried to get into the other lane before the light. Or sped up faster to match the speed of the cars in the adjacent lane so it could merge more smoothly.
I think this was an excellent video that 1) illustrates how hard this problem is, 2) how much continued work is needed to refine the driver and 3) the trade offs between safety and "functionality" when there are unpredictable humans in the loop (environment). Erring on the side of caution is laudable, understandable, and IMO the right choice. But it also means that creating a scaled and reliable service will be a much slower process.
Their models are designed for a world of unpredictable human drivers. Even when the world goes autonomous, their systems will get walked all over by other systems.
The difference between this and Tesla "Full Suicide Driving" is astounding. This is clearly the way forward. I'm sure there will be challenges in mapping, and making the sensor suite affordable. But it makes all other systems look like a joke. An incredibly dangerous, irresponsible and pointless joke.
I'm won't be anywhere near a Tesla ..... as a pedestrian.
They will turn into "drive yourself taxis"
Impressive. What would it do if a car in front was driving below the speed limit and there's a lane to overtake?
It never overtakes, at least not that I've seen. Lane changes for routing only
@@JJRicks so if a car in front breaks down, or deliberately drives at 5mph? You're stuck?
Wellll not exactly. It will go around a stopped bus, but it won't go around someone doing 40 in a 45. I'm not sure exactly where the line is quite yet
@@j8rtUfh3evz remote assistance operator will engage and reroute path to go around vehicle if needed. AV’s are continually monitored
My only question is whether the cops can pull them over for a moving violation or for a defective brake light etc and if they can who do they talk to as well as give the ticket to especially if there's no passenger in the car.
Yes, I believe they talk to rider support. Ticket goes to the company
This is very interesting. If other drivers were less aggressive, the Waymo could continue uneventfully. The impatience of other drivers and aggressive behaviour is hampering the Waymo. For example, once the Waymo signaled its intent to merge left with the lane running out, other drivers should have yielded. Instead, other drivers closed the gap making it impossible to merge. If everyone drove like Waymo, there would be less stress, less accidents, and everyone would still get to their destinations safely.
Or the technology isn’t up to par yet who’s gonna let someone going 10 mph merge in front of them nobody likes to waste their time
If only every car was autonomous and interconnected, this wouldn't happen.
This must be the Tesla killer Robo Taxi
That indeed it is
Nice
14:22 That car on the left with no lights is not parked! Autonomous cars should *BRIEFLY!* honk and flash their high beams at people who forgot to turn on their lights. It doesn't need to be aggressive. Honking can be informative or advisory if you do it briefly enough, and it might help to keep someone from harm. In fact, that's what the car horn's intended purpose is.
15:33 There's another one! Do you guys not warn each other when you see someone driving at night with no lights?
17:37 And another one! People, the lights aren't just for you to see where you're driving, they are also for you to be seen by people who need to yield to you. Don't wait until after it's completely dark to turn them on.
Did I miss the cost? Electric? No human to pay.. no tip... guessing $.25 a mile?
JJRicks.com/ride-data
When you’re a passenger and you have to wave “sorry” at motorists and roll down your windows and say “sorry” to pedestrians, something is VERY wrong.
Bottom line is, these things are an absolutely terrible idea, they clearly do not work, and they’re gonna get a LOT of innocent people killed if their presence in the roads increases (which I doubt they will increase).
There are just too many unknown variables to account for in population dense areas, and these things are dangerous. All these things do is confuse and/or frustrate the people around them.
Yeah I'll believe it when they make machines that fly, am I right?
Excellent video, it shows that Waymo needs to read the traffic a lot better. It needs to see opportunities and predict where cars and pedestrians will be in a second. Tesla FSD also has its issues with traffic and it wouldn't probably handle it any better.
I guess there is still no substitute for intuition and experience in traffic. Also, in a lot of cases (if it is not dark) I am looking at the face of a driver that is coming from the right or a pedestrian, just to note if they see you.
Many times you can predict a manoeuvre somebody is going to make before they actually make it.
This is something that an autonomous system lacks.
Great video! How far did you go in miles? Interesting how it's much less aggressive merging than Tesla's AP..
Uhhhhh.... I've taken so many rides since filming this I can't find the trip :P Thanks for watching!
@@JJRicks Tesla AP feels very human like in terms of merging between cars, downside it will make a mistake with the same solid feel.. Waymo seems to be taking a different approach here..
Everyone who rides along with me says that the braking, acceleration, and turning is incredibly human-like and smooth, I can say that much :P
@@teslascoop2177 the problem with this is that a bunch of people who only watch this video just see this super hard situation I threw at it, and not the other perfect videos. So the they immediately pass judgement
@@JJRicks Well this is how you judge what good is 1,000,00 miles of easy straight line driving right? BTW I'm referring to merging not accel/brake etc. Just noting the difference between systems..
RIP Waymo
For next time when having the frame problems just stretch one of the videos till they are the same lenght in total! Then after go edit the video so everything is in sync! Just at tip! Im loving the vids just can´t imagine this happening in Amsterdam here with all the bikes haha!
Ah, that would almost work except I start the cameras at different times lol
Awesome...
Looks like Waymo has way more to do b4 they go.
This is AMAZING. The lives it will save. Out drinking. Just call your car from your garage to pick you up. No more Uber/Lyft drivers. More safety. Less truck accidents on the highway. Wonder how they will handle BLACK ICE.
Am I the only one who thinks this system is beta? Threading into the track took forever. The bus driver behind it has already honked. With pedestrians nearby, the system doesn't seem to know what to do and just stops. That would frustrate me as an inmate. Especially when I'm in a hurry.
How far did u go and how much u pay?
www.jjricks.com/ride-data
I wonder if a police car can stop this car from behind just with visual and acustic signals. I would also like to see the officer's face when he wants talk with the driver. :-)
Yep! Have demonstrated in previous videos. If you want to search through and see look for "emergency vehicles" JJRicks.com/videos
@@JJRicks Thanks, I'll search!
Well, not the officer thing, but close ish!
So many people just walking in front of the Waymo, preventing it from starting the drive! 😂
Did the car fail to indicate its intention to depart? Or is it more that most folk now know the car can't (hopefully 🤞) run them over?
I feel they're going to have to up the aggressiveness algorithms. Maybe the solution is for the AI to ignore every 50th person walking in front?... Just to keep everyone on their toes! 😉
Hahahahahaha yea, I'm mostly not annoyed, really the fact that it made it out at all is a huge testament to progress. Though people walking in front is definitely annoying
People are impatient. In a grand scheme of things, minute or two of delay, is not big deal. The car handled everything well. I am not sure why it had emergency lights on , maybe it was related to the call.
I think the hazards come on during a reverse or a remote assist
Yeah but as soon as one accident happens, everyone is going make a huge deal out of it..
The mapping is excellent. But consider how much it cost, then the cost of mapping the land area of the planet! Then keeping it updated on a daily basis.
Not viable, IMO.
You don't need to map the planet for this to be useful technology, they can just launch in specific cities. A taxi service that can take you anywhere within new york city, for example, would be incredibly valuable.
In terms of the cost, keep in mind the cost of mapping the area is offset by the savings from not having to pay any drivers. Let's say a company had a fleet of 1000 human-operated taxis operating in new york city. Each driver makes an average of $15/hour and works 40 hours per week.
The weekly cost of labor would be $15/hour * 40 hours * 1000 drivers = $600,000 per week. Per year it would be $32 million.
As long as Waymo is able to maintain their mapping of NYC for less than $32 million, they would be able to come out ahead of the human-operated taxi company.
@@kevinaud6461
There ARE the other costs - updating the maps on, to be effective, an almost daily basis, then software engineers, computer maintenance, cleaning (which drivers do as and when they have no fare) and so on. Not to mention having drivers available to go cross-city to extricate the vehicle from situations where it is stuck - and someone to drive the rescue car back or just find it a parking space.
Autonomous cars might work on a large scale in areas designed for them but.......
@@Wol747 Keeping the map up to date doesn't involve re-mapping the area daily, it just involves re-mapping the parts that have changed. See the "Keeping our maps up to date" section of this blog post: blog.waymo.com/2020/09/the-waymo-driver-handbook-mapping.html
Basically the cars cross-reference real-time data with map data and if they find a discrepancy then it gets reported to HQ at which point they re-map it. Only having to map the things that have actually changed makes it pretty scalable.
Thanks for sharing! This is interesting! The lidar sweep shown on the screen is slow occurring every 4s or so. In actuality do you know how fast it is sweeping/rotating?
Am not sure, I think the sweeping is a design choice!
I don't like how it says your car is ready to go, it should say I'm ready to go or something
The AI needs to study lots and lots of and lots of actual drivers’ behavior in parking lots to learn how to handle these scenarios. Obviously they haven’t don’t that yet.
This video is ancient; check my recents
What does the non-driver do if someone waves the car in?
Nothing :P it only goes when it thinks things are clear
What if it’s fake and they’re just remotely driving the vehicle?
Nope, it's not. That would be incredibly dangerous
Nah, too much latency to pull that off
How did the car know Where you wanted to go????You did not program your destination
I did in the app before it was ever on the way
@@JJRicks can you make another video which shows the operation of the APP?? Also WHERE IS THIS???
@@bmaiceman Already made an app video, all you have to do is scroll through my videos a bit. ruclips.net/video/L8HLrxJp2n8/видео.html also JJRicks.com/faq for more questions
Feels like waymo has quiet a way to go too. I'm actually surprised that they allowed them to drive it without anyone needed to take control. Is this limited to special streets/areas or can you really run it everywhere? Considering you started on a parking lot I'd guess it can just drive whereever. I'm not sure I could say this is better than FSD though, considering some comments made by waymo officials. I like the visualization a bit more, but then again you wouldn't watch that anyway in a final product, which would probably have passengers facing each other unless there is some safety issue with sitting backwards, but I think it's not a big concern. in the future we will probably see 4-5 seaters with a nice table and you can play cards together while driving :D
They drive in many cities, but are only open to passenger service in Chandler. Rider support can't joystick the car, they can only give it pointers like "turn here" or "ignore that thing"
now they just need to make the cars fossil fuel free to elimate the noxious gases
I think these cars will end up banned in major cities, too many problems.
Bold prediction
Is it normal to have weird box at the roof of car
how many times has a line of code failed. This is a stupid idea with way to much risk and only the perceived benefit of putting taxi drivers out of a job. It's just not worth it. Sure its cool technically to be able to, but I dont think people will want this.
I’d rather be teleported. Much safer.