Yep, but silly that the presenter is not aware of this. Tesla purposely removed the other systems and feel that the software can do all the needed things with 'just' camera's.
As far as I understand, XPeng approach is more akin to Waymo or GM Cruise. They are using HD maps and lidar. It’s really faster and a short cut to get to localized L4 autonomy. However, it is more expensive and geographic constraint. The real winner will be the one getting to generalized FSD L5 autonomy. That is a lot more difficult and as far as I know, Tesla is the only one focusing directly on anytime anywhere autonomy. Time will tell who will win.
This comment! Poor choice of title by one of my go to channels. I enjoy the Chinese EV content normally, but this just seems poorly researched. Tesla has a fleet of millions collecting data. I really can't imagine anyone beating them to it. 2024/25 will be interesting times for FSD!
Yeah first thing I noticed when I saw the screen. I think both can work side by side but Tesla up until now has focused on FSD driving as if it has no idea about the layout of the road ahead and has to learn it on the fly. This system has all the roads and layouts preprogrammed so it knows exactly whats coming up and wont get caught out by bad markings. I think Tesla is starting to do this too though so they can eventually have both going at once. It's a shame that this video and channel doesn't mention this stuff, its a very base level knowledge of how this and FSD works, essencially just point out that this has lidar and Teslas dont, but thats not the biggest difference like you say.
I am very very skeptical. Its like that Boston Dynamics jumping robot. it looks great (on the surface) but it is as dumb and a sack of spuds. We are far away from self-driving being safer. oh, and i makes car soooo boring.
These technological leaps will continue to exacerbate the meaning crisis humans are already facing. Driving is very enjoyable for many and this technology will inevitably lead to human participation behind the wheel becoming illegal. We are literally innovating our own obsolescence. Think I'm crazy? A.I. could be doing almost everything including its own maintenance within three or four generations. The horse and buggy analogy doesn't hold water any longer.
I very much enjoy and appreciate the work of the fully charged show. That being said it is surprising, maybe even slightly unforgivable, that the host can be so wrong and so repetitively wrong about Tesla and its radar. This should’ve been picked up way before the video was shot, and if too late for the shoot, disclaimers should’ve been inserted all over the screen in post. Keep up the good work, I know you can do better. And if you have been, thanks for reading…
My thoughts exactly. I’m a fan of Fully Charged in general and consider them a good source of information so the repeated errors about Tesla and FSD really stood out and are very disappointing. Elliot is my least favorite host on Fully Charged and this isn’t the first time he’s come across as uninformed in a video.
@@gormauslander Agreed. Also, Pretty sure its a geo-fenced city. Not useful in all those other cities where the other 6 billion people live. He should take it to a random city and test it there. Try it on a random country road. Then tell us who's winning.
Gonna be amazing with any amount of Tesla robotaxis on the road when weather hit. How many Teslas does it take to create total gridlock of a metropole? If they are ever allowed to operate at level 5, we'll find out the first day of special weather. Just when we need the grid most.
@@Cloxxki not had any issues with FSD Beta during weather recently. During heavy rain it slows down and turns wipers on fast. During fog it slows down but sees much better than I can in the fog. Don't know how it would do with smoke or heavy snow.
@@Cloxxki Tesla’s FSD works extremely well in rain and fog. The neural nets are able to compensate for the conditions and the car can see better than a human in those conditions. There are plenty videos out there. You should try watching one.
I have never seen a not positive fully charged video. All cars are good everything is nice all well here. I wonder if it has anything todo with pleasing the manufacturers. It would be nice to see an opionen sometime, any time
I imagine Robert will address the comments we've made about Tesla's not using Radar for a little while. And they stopped using Radar on Full-Self-Driving Beta last year around this time actually. It's been vision only for a reasonable amount of time. Even Robert's car should have gotten that update and told him by now. I'm not that offended that it was wrong but like others have said it is reasonable to expect Fully Charged to know this about the Tesla cars seeing as they are by far the most popular Electric car brand and have been doing autopilot for so long.
@@audiovideophile5317 no they can build precision maps from the cameras but it doenst require maps at all to actually drive, latest build on roundabouts not even on the maps works fine
In an interview, the former head of Tesla AI told that the sensors creates additional noise to the system and doing more harm than good. Also, the advantage of having sensors is tiny but it costs a lot more compared to just having cameras. That’s why they removed them.
The evolution of these vehicles is continuing. These self-driving cars will be of great benefit to elderly drivers. They will allow older drivers who would normally have to stop driving as they age to still maintain their freedom and mobility. 👍👍
@@teofilol2666 I think good competition would benefit even Tesla fans, and maybe if it proves itself, people would consider shifting from Tesla. The reality is though, this test was done in completely optimal scenarios and was only done 1 time. You never test something like this once and say it's perfect We would need to see a lot more tests before determining how good it is, but I'm optimistic
Bothers me when you get a fact like Tesla only using vision wrong. On a channel dedicated to EVs. In saying that, love what Xpeng is doing. Hope it comes to Australia!
Interesting that he had to drive out manually to the main road before engaging. Tesla FSD beta can be initiated in many cases from parking lots and from peoples driveways and the car can drive to the main road. I did notice however that this Xpeng did make a u-turn which Tesla currently don't do but this is a function of the the digital rails that are pre-mapped within the geofenced area. Tesla will likely do U-turns everywhere (w/0 pre-mapping) due to their FSD approach.
A few examples of U-Turns have been shown by FSD testers, but it isn't baked in yet. Those were just flukes, until they focus on them for a week and then U-turns are magically solved
It's likely because of when this was filmed but its worth nothing that Tesla has dropped radar going forward. Tesla Vision only was certainly the long and hard path, but it seems it may have been the right call.
@@jeffmocha9478 Tesla says it adds noise, which creates errors. Also it's just cost, weight etc. And certainty don't underestimate how much it costs to keep developing/improve ON TOP of vision. Essentially the same can be said for LIDAR except amplify it. On the other hand, many believe Teslas approach cannot be successful.
@@jeffmocha9478 my tesla drives only with cameras and it‘s totally fine… it can keep the distance to the car in front of me, slow down smoothly when there are cars in front of me or when the car in front of me slows down , and it even stays on the right side of the road when there are no lanemarkings /the lanemarkings dissapear.
@@jeffmocha9478 It's counterintuitive, but no, it would add noise (confusion) to the system. If the cameras and radar/lidar disagree, what does the system believe? The cameras are much more information rich sensors and with capable AI can become superhuman at low cost.
Brilliant video, as always. Impressive FSD. Tesla vision does not use radar. But jury is out for who can drive 10 times better than humans, on every city road, first. Eamon
Bristling with fricking lasers Elliot! - just like the latest NIO I looked over in Berlin a few days ago. Video of that up on my LinkedIn feed. Another lush and lovely vid my friend! 👏
According to Sandy Munro, Forward Looking Infrared(FLAIR) is used by the military for guiding its missiles. He says it is more accurate and detects things better than LIDAR in all types of weather. It is also more expensive than LIDAR. I am aware of 6 different LIDAR companies, most of which I own some stock. LIDAR manufacturer, Aeva's , has announced its LIDAR system will be used by NASA to map the lunar surface. You might want to do a video on the different types of LIDAR systems.
I wonder how long it would take to fall apart on a roundabout in Spain. Especially in Valencia where you can have a few lanes entering the roundabout, a traffic light almost immediately for the next entrance where cars just line up side by side with no lanes at all and then a mad dash when the lights turn green to get into one of the few lanes leaving.
Spanish roundabouts are special, its like they are created to cause crashes, who ever decided that it was a good idea to use the outside lane to go all the way around is an idiot. A challenge for humans so a good test for autonomous driving.
I have Tesla model Y and I think relying solely on video cameras for full self-driving is short sighted. Video cameras are great for picking out detail, but don't work so well when poor weather impairs visibility. It is much better and safer to use a range of sensors such as cameras, radar, lidar, etc to provide an augmented enhanced visibility for self-driving, which improves on what humans can do unaided today.
Hmmm .. you realize lidar is light based as well and is much worse at bad weather - and there are weird lidar issues like mirroring as well…. Radar also has it‘s challenges - I‘d say Tesla has proven it can win Euro NCAP without Radar on the car …. The jury is out for USS vs Cams but I‘m rooting for Tesla as well :)
@@_baumi_ I think the main problem with visual is processing speed. Lidar space mapping take minuscule percent of processing of visual identification. I see FSD slows down and can't think fast enough in dense city traffic.
@@enigmatum I've never seen FSD slowing down. Can you mention a video where this happens? (Title/timestamp - don't post links, your comment gets removed) Remember that the on-screen visualization is independent to FSD, and runs on a separate computer system. Even if the on-screen visualization is lagging or even freezing, FSD would be unaffected.
I don’t think it’s short sighted, after all we humans only have a pair of eyes. The system only needs to be better than us to work. Also, having three different inputs (visual, LiDAR, radar) may be very difficult to work with if you start getting conflicting data (something Tesla explained as the main reason to abandon the others).
@@logitech4873 it was a 2-month old video from AllAboutTeslaY. I can’t find it on his list quick. 20 pedestrians and heavy traffic right turn in NYC overloaded FSD. It just died.
Looks very capable. But with the expensive lidar sensors this will only be reserved for the top end cars. This system won't end up on a $20K car in the future (when battery prices drop enough). And, as he said in the video, I wonder how this will work in crappy weather. Radar can see through mist, but gives too many false positives and resolution is low. In the end, vision is the only sensor required to drive, as that is what humans use. Different approaches, different results. It's going to happen, time will show us how.
I was driving at night recently on some windy mountain roads without any lighting other than my headlights. It was scary and I had to drive very slowly. Presumably with Radar and Lidar I could have managed a more normal pace.
What I really want to see is how traffic flows when many different L4 / L5 systems have to share the road. Each with its own idiosyncrasies and weaknesses. Lidar has the benefit of high precision, has issues with dust/smoke/rain/snow. Radar can cut through most all environmental issues, but has real issues with refined object understanding... though I'm given to understand newer versions are getting much better. What I understand is the real difference here is between those that use Lidar approach and the Tesla approach is precision pre-mapping. You can build in smarts to handle weird events with the Lidar system (debris on the road, construction, etc) but they are fenced within the pre-mapped regions. That's fine if your area is mapped and you don't care about manual driving between pre-mapped areas. However, the Tesla approach is AI driving. They might take hints from navigation maps and GPS, but really the system drives as if it's the first time it's seen that stretch of road. It has to see and understand, then drive. Upshot is this works anywhere with no pre-mapping. Downside, it's a way harder problem to solve. Anyway, great vid! Thanks it was really interesting. Looking forward to more.
The tesla vision only system will be the degrade mode for system metioned here. Tesla is the only brand in the industry has such wide West like high tolerance on exposing their consumer to risk. There has been few accidents caused by failure of FSD in China. But their consumers base do not have much interest to pursue legal action against tesla. If it is a domestic brand, you will see media coverage flooded every where and the consumer will have class action lawsuit to suit the company to bankrupcy.
Just wondering if multiple vehicles equipped with radar and Lidar in close proximity would they interfere with each other or create a high noise floor that they wouldn't work properly, IE they couldn't see the reflected signal through the noise of other equipment operating on the same frequencies in the RF and light spectrum.
@@h20dancing18 not quite sure myself but radar transmits a radio signal then measures the doppler shift on the reflected signal if you have many car's on the same radio frequency or close adjacent frequency will it cause problems with the receiver on the car, I don't know how many car's on the roads have these onboard systems yet I suppose time will tell as more car's adopt semi autonomous or full autonomous driving.
Self driving is not a hardware problem; it's a software problem. Vision, lidar, radar, flir; doesn't matter in most cases. It's what the computer does with the info.
@@morilot NO! So if radar can see a car in front of you but the camera can’t because of fog, you absolutely want to ignore the radar because of less conflicting data? NO, it’s stupid. You want to get ‘super vision’ for the car, not some 2007 smartphone cameras.
@@EVPaddy Those "cellphone cameras" can see way better than you in fog, especially since they don't use IR-filter before processing. Just look at the statistics. How many Teslas has collided in fog while being on autopilot?
Most impressive. I get nervous when other people are driving, can’t imagine when the car is driving itself…lol. I wonder if other car’s LiDAR beams will interfere with each other when more of these cars are on the road?
every car will use a different wave length... .if they can put millions of beams in one Optical fiber and don't interfere with each other, I think they can ''fix'' this too.
One question I have that I don’t understand is: tesla fans are arguing that a big advantage it has is the large number of cars on the road collecting a lot of data. but tesla fans are also attacking xpeng and its system for significantly using data it has collected to make maps of areas, And saying that this means Xpeng is limited to driving only in areas where it has data maps. But isn’t the majority of data being collected by Tesla cars also related to the streets, towns and cities that those cars are driving in?. Also, it seems highly probable that in the next few years that 99% of cities and towns around the world will be mapped with the type of mapping that xpeng is using in this video. It seems that either a big company like Google will do maps like that and sell them to the other car companies. Or that the car companies will sell the maps to each other and or trade the maps with each other so as to get global coverage. For example, the Chinese car companies will sell or trade their maps with American car companies. Also, most of the cars sold by the big companies are sold in North America, Europe and China… So they really only need advanced mapping in those three regions. Between all of the car companies, autonomous driving companies and tech giants like Google and Applle, I think they will fairly soon be advance maps of 99% of those three regions. At that point, even if tesla develops the ability to do autonomous driving in unmapped regions, I don’t think it will matter commercially speaking. The final point is that driving in China is pretty crazy due to the number of pedestrians and bicycles, which result in many unpredictable things taking place. I think that if Chinese companies are able to handle those things well, they will be able to handle most other unexpected things.
The big danger in self driving systems is the interim stage which is where we are now. The systems are improving and can be impressive when operating within their prescribed parameters. However people are likely to be fooled into thinking that these systems are infallible and will exceed the operating limits eg drinking, watching videos, playing games and probably even sleeping. This will be fantastic when these systems are truly autonomous but at this stage they are likely to contribute to higher accident rates than they are to lower them
I share your concern and manufacturers are well aware. That's why some also build attention monitoring systems in the car. It's not a 100% guarantee, but will make it more fool proof. In the end irresponsible drivers will always be on the road, no matter the type of car, unfortunately.
That's Tesla's weak spot right, the reason Tesla autopilot crashed into a white truck, because the camera saw it too late against the white clouded sky
Impressive, XPeng is one of the best in China if not at the top for ADAS. It seems like many are looking at LIDAR now, including NIO, Huawei, etc .. and other brands.
Still to few. In Europe almost all new crossings are roundabouts, traditional crossing exists only in pre-existing crossing (tipically in old cities) with too little space for a proper roundabout.
seriously interesting btw but why only 90% of situations? what happened to get you to reduce this from 100% or even 99%? yes it was a final beta by the sound of it but what did you cut out?
3:35 The lidar detects there is an object there, but needs the camera to identify that it's a person. It only measures distance, which with today's AI, cameras can also do that.
Exactly. All lidar does is estimate the distance of something using pulses of light. It takes way more to take that data and extrapolate that it's a human. That's where Tesla's Vision system is superior. Two actions in one, see object and identify object.
it's great that others are working on FSD too, but the compression and ignorance of Tesla's system make me wonder about the quality of the overall research of FCS.
Good to see Elliot covering EVs for us from the inside as usual! But gosh XPeng, I wanna be a fan for competition, but that double tap stalk to activate driving assistance is a little too familiar? It's really not the best activation method and you gotta just have ways to do that better right
Melon Melon: So if interfaces have similarities, it's a problem? Why? I don't want to have to learn a completely different system every time I use a different device from a different maker. It's bad enough with computers and different OS's, for example, already.
11:08 is the most impressive to me. Like it noticed the car, two cars in front of you, that wanted to turn left. Shifted lanes for a more uninterrupted drive ie efficiency. I’d say well done 👌🏼🤩
There's plenty of footage going back several years that show Teslas predicting other cars ahead of it having an accident and avoiding them. Tesla also seem to map people and roadside objects better. Elliot was also too impressed with the car slowing down at traffic lights where the lights were obscured by the vehicle in front. Didn't it occur to him that the car can detect THE VEHICLE IN FRONT? This level of self driving is great but it's old hat for Tesla.
The Xpeng is great and self drives very well. But, I haven't seen anything in this video that a Tesla doesn't do already. Elliot is impressed with the way the Xpeng slows down and speeds up as cars intrude on it's lane while driving. I hired a Kia e-Nero a short while ago and active cruise mode does exactly that. I set my speed to 80mph and the car tracked the speed of traffic ahead of me and kept me in line with general traffic speed be it 70mph or 40mph. When someone jumped into my lane it would instantly slow down and when the lane ahead of me cleared it would instantly speed up again. I don't know how Elliot is so impressed with what are now standard features in EVs. And he really didn't explain exactly what the advantage is of lidar. In fact, he explained the disadvantage of using it at night. Also he was driving the Xpeng around a part of town adjacent to the Xpeng facilities. The car is familiar with the area and the surroundings are also well mapped on the satnav system which gives the Xpeng a home advantage as it were. Maybe the gentleman sitting behind Elliot had something to do with his over-praising of the tech?
I think you dont get it , can you choose a location on a tesla map and let it take you there. ? Thats what he did , autonomous driving, you get into the car , open map , pick your location and the car makes it happen. The others you drive to a location while sonetimes letting the car do some stuff like keeping lanes , keeping distace , mantaining ceuize speed etc . This is complete " car take me to work"
That was not a real test. How about driving on roads that don't have great road markings. Also have some tight areas with cars parked on either side of the road. Also tesla don't use radar anymore. So much FUD from Fully Charged, did you have Chinese minders from Xpeng telling you what to say? What's the price difference between that car and a Tesla Model Y?
Of COURSE it was a real test, Fanboi..! It was a test of the car's ability to self drive in a city - which appears to work ..... Stop being a pathetic, cowering slave of St. Elon - you poor, poor fool ....
There are a lot of questions I have about this: is this a swarm system? Does it only work on roadways that are already mapped and in the database? Is the system analyzing *all* the moving objects?
It is not clear from this video why it would be better than Tesla solution. The (current) title “This car is winning the race to Full Self-Driving” smells like click bait.
This is an extremely poor episode by Fully Charged. The headline is highly misleading....no, this car is not "winning" full self driving. It is using different technology compared to Tesla and if the presenter had researched properly, he would have been able to highlight the pros and cons of each. A simple search of FSD on UA-cam will highlight how Tesla's FSD has reached impressive levels compared to this very limited Xpeng "demo". Both will still have a long way to go to really become fully autonomous but to label this as "winning" is really disappointing.
I own a Tesla and although build quality is improving, the zealots in the Tesla Vision department are making the cars worse. There’s no sensor redundancy, no radar, they’re even removing ultrasonic parking sensors. Everything will rely on the cameras despite their blind spots and susceptibility to dirt and dazzle from lights and the sun
It's very good to see competition in this sector, because that will drive up the function and drive down the cost. Better than FSD? I'm saying no after watching this. You only need to see how the car reacted to the pedestrian after the U-turn to say it's not at the same level. In time, this and others will be acceptable for 99% of driving scenarios and we'll all be safer.
This is china so even if everything he said is what the company told him, it is highly probable that this is not the ole story. This is the country that paints grass green and staples leafs in trees. He said it drove 99% without interventions but the video shows 100% without interventions so something happened that didn't make to the final video.
🤣 this thing did far better than every fsd video I have seen. But keep fan boying. Tesla can't even park itself or navigate a traffic cone in the middle of a lane. I have seen the latest vids.
@9.55 in china the standard satnav has info of the traffic lights as well as the GREEN/AMBER/RED state including a countdown timer .. the satnav informs you of the GREEN light and alerts you to go . i think that info may also be in use here ..im planning to get the G9 next month 😊
Impressive. Now they just need to advance in another area: stop the genocide, and stop making threats to other countries. Then it could be a great car.
@@frangalarza not great if you have 1, but ok if you have 2. Questionable if there is not reference, which usually dont. That is why they use laser in the industry when precision measurement is required.
Okay, this is interesting but I feel there are a couple of points to be made, firstly Tesla FSD no longer uses radar and is phasing out ultrasound so I lost some confidence in this review within the first few minutes. Secondly from what I could see 99% of the driving was done on duel carriageways, which isn't 'city' driving. I would also mention that the car was full of engineers so I'm not sure you could call it an independent test, we're talking about China here obviously. I'd be interested in seeing another video with this car when you're allowed to drive it for a period of time without the minders then we'd see how it really performed. TBH all this video showed was how far Tesla is ahead of the field.
I don't think there is another company within 5 years of Tesla on the self driving front. They're only using vision and are trying to basically solve General AI, whereas the others are using high res maps, radar, lidar etc.
After watching this video I feel even more impressed by Tesla's FSD in comparison. Tesla uses only cameras while the car in this video have to use cameras, radar *and* lidar to be able to drive in a limited area.
@@林振华-t4v they're going to need to release it to the public if they want real testing done. The data they'll gain from regular users will be overwhelmingly more useful than these controlled/limited tests
@@realshnx most companies wont release alpha sofware to public and release beta sofware to Limited or selected public, some orgnization will add and Release Candidate phase to general public to sort out what you metioned, have as many people as possible to break the sofware . Tesla going out her own way asking 1.3 K for a beta (no, not me, I did not opt in the beta auto drive on my model 3), and basically having mutiple major outstanding engineering issues on pretty much every (except the Y) model at its first generation release. They pretty much try to patch issue over the whole life cycle of their first generation of the product. That means every unit sold is at a lost, because the design did not get frozen long after the mass production started. That is why they did not have a profit year till 2019 or so. Think about it, burning hundred million if not billion year by year since 2012( just for the sake of focus on the serveity of the probelm) till 2019, 6 years consecutively. That is not a sustainable model even as tesla themselves. And now look back, how far has tesla get ahead with the game? Not so much Id say, Other solutions rely on infastructure and mutiple sensors hace achived same or better result. Ditching other sensors and focus solely on visual spectrum camera will somehow reduce the difficult on system integration. Yet, the lag of data aquisition method will limit the system capability. Even us human wont solely rely on our visual for driving. We also hear, feel and even smell as well. So, what will the outcome of tesla visual only solution moving forward. There is a good chance it can only achived what other systems can achived at degrade mode. e. g. other sensors all failed while visual system still operational.
Another brilliant video from Elliot and I love these little snippets of life in China, but I am not a big fan of these systems, problem because I love driving. As for those Tesla fan boys below banging on about this video not acknowledging "Tesla doesn't use Radar any more" wrong they do use Radar, because Radar exists in their proceeding models. So it would be technically incorrect to say they don't use it. Yes they may have decided not to continue with it! But then this video isn't about Tesla is it!!
They don't use it in preceding models either. It's been disabled via software. Sure, you can say it's used for the oldest cars that don't even support FSD - but that's irrelevant.
Hi Frank, absolutely correct, Elliot got that wrong. Tesla just git rid of the ultrasonic sensors too. And of any maping a few weeks back. Vision only. And of what I see Tesla works pretty much as good as this X-Peng, which would deny service once it gets to the limit of its determined area. Then the Tesla would continue to explore, analyse, decide and learn and especially: continue to drive you.
Tesla has never had LiDAR on production cars. They removed radar over a year ago (3&Y, S&X early this year) and disabled it in software. Recently they removed ultrasonic sensors on the 3&Y being produced. They let those with existing sensors have it for now (and I wouldn't be pleased when they disable it)
I'd advise you to look up the video where XP crashes at full speed into the guy and his car (killing him) that they were travelling to help after he broke down/crashed. Edit: as a side note, if you're around 6ft4 or upwards don't expect to be able to drive any Chinese electric car apart from the Nio ET7. They all have at least one of the following issues; low roof, too high seating, top of window is too low, seat doesn't go back far enough, headrest is fixed and pushes into your neck and not your head. The last one used to not be too big of an issue, but considering how some electric car companies are with modifying their cars, you might just void warranty etc by changing it..
Tesla doesn't use RADAR "Safety is at the core of our design and engineering decisions. In 2021, we began our transition to Tesla Vision by removing radar from Model 3 and Model Y, followed by Model S and Model X in 2022. Today, in most regions around the globe, these vehicles now rely on Tesla Vision, our camera-based Autopilot system."
The radar they drop is likely the parking sensor radar. Not the lidar seen here. Which for tesla, not a huge lost consider parking radar low resolution and refresh rate
@@林振华-t4v Tesla has announced that they are going full "Vision," relying only on cameras, even for parking. They even just announced they're removing the ultrasonic sensors, disabling self parking and summoning (for now).
Very nicely put together vehicle, in terms of road experience and appearance. It's been really neat getting a chance to see vehicles that, sadly, will probably never make it to the US.
Interesting, impressive and strangely compelling/suspenseful (helped by the music I think). As others in the comments have said, I think lidar and radar are the norm in these systems and Tesla are deliberately going camera only to bring a more affordable system to market. I'm not sure how cameras deal with fog and rain but as they can have access to a greater range of wavelengths than us humans, perhaps these things aren't a problem. I don't know how Teslas system would compare to this either, but I understand that they have an advantage in that they have many more miles of data to work with than others by virtue of their massive fleet of test cars - aka customers cars on the road, and it is data that makes these AI systems more competent. I can see how a company like X-Peng could catch up on this front though as they have a huge market to sell into (but so do Tesla). The winner might be the company with more data, let's see.
It wasnt really about affordability so much as the data bottleneck. By going only with vision, Teslas system doesnt have to get the data coming in from several totally different streams and converge them together. So they can focus on getting a higher bandwidth of data on just the visual channels. So they were able to upgrade to higher resolution cameras for instance because they can process that data so efficiently now. And the data they were getting from radar was essentially redundant to or of generally lower quality than what they are now able to achieve with just vision. Even in adverse weather conditions. Which when we say "vision" we have to understand that digital cameras are not the same as human vision and digital cameras are all capable of at least short infrared... in fact digital photography cameras have physical infrared filters so the images will look closer to human vision. So these cameras may be capable of seeing better than a human in certain adverse westher conditions even without fancy AI... which they are also working with their AI to improve object detection in fog. And remember that in really heavy rain both lidar and radar systems become just as useless as vision. Although vision may actually come out better in really heavy rain than lidar or radar if the AI's object recognition and persistence of memory is really really good.
@@patreekotime4578 the cameras on current tesla models are not even as good as the one you get from blink. Especially in low light condition. infared spectrum is simply not the spectrum the CCD or COMoS onboard of a Tesla can pick up. And the refresh rate on the current model is simply not there either. That is why for so many years, the sofware is still at "beta"
@@patreekotime4578 the reality is, their tests with vision only FSD has been the most success they've ever had with it. That's why they are confident they don't need anything else
There's going to be interference from other cars. I don't see how it can be avoided entirely. They are aimed at a lower angle which will minimize interference. And as lasers they aren't broadcasting over a wide area, but rather as discrete rays sweeping quickly around the area which will help. But with traffic all around you every day there will be multiple opportunities to encounter cars that can disrupt each other's lidar system. Hopefully they are designed to simply see the other source as a bright point that it can ignore instead of overwhelming the sensor.
We’ve all seen company-produced self-driving demos for years. We’re past that now. Any FSD review has to include full clips of longer drives (possibly sped up) with no cuts. Otherwise it’s a parlor trick.
Looks like hardware assisted vision based autonomous driving is the way forward, at least it's more reliable. The hardware should be able to intelligently switch on and off to adjust itself to highway drive, city drive and rural driving etc.
Did you see that our comments are constantly deleted by the Fully Charged Show? Comments like "Reaction of Tesla Fan boys is GOLDEN" are kept but if one is discussing the "facts" presented in this video it gets deleted by the channel owner. Interesting way to handle your subscribers, guys!
@@AndreasWest I haven't had any of my critical comments removed. Yours are prpbably being removed by UA-cam itself automatically, which is far more common. Don't include links, and try not to have a very hostile tone in your comments.
I see these demos, and it isn't starting to ice up, it isn't pouring, it isn't starting to be a snow storm, it isn't wanting to drive down a gravel / dirt country road / cow path. But they want to remove the steering wheel? Whenever they say they want to remove your ability to act, be suspicious, very much so. I assume there are conditions where these systems would just stop working, but no one is addressing these. What does happen in these conditions? How does one drive into a garage without a steering wheel or out of a business as in the beginning of this video? Lots of questions, no answers.
The car with no steering wheel for public roads is very very far away. And if it ever comes out somebody will have already figured that out so I wouldn't worry.
I have seen Tesla doing some amazing things. their new FSD software is handling some really complex scenarios. From my estimation, Tesla are leading the way by some distance.
@@logitech4873 Tesla fanboys obsession with HD maps is funny. What exactly do you think HD maps do? HD maps doesn't help you avoid driving into other moving vehicles. It doesnt replace cameras or sensors. It's there for planning. Also we saw what happens when a lane disappeared because of temporary construction. Which would not be in a map. Step out of your tesla bubble. If you actually interested in learning how others do this. Go watch waymos vids about how their self driving works. HD maps is only a part of puzzle.
@@daydreamer8373 No you haven't. Any complex situations get the tesla in a tizzy. The last video I saw was last week, when a tesla couldn't even navigate a traffic cone in the middle of its lane. And that's after 8 years and billions of miles. 🤣 And let's not forget the simplest of situations... i.e. parking. Tesla even get that right. Your estimates are based on fanboy bias.
@@jeffsteyn7174 I have been following self driving for years, and have watched many videos including Waymo. and I have seen them all mess up. Tesla's were particularly bad, and would struggle with the most basic scenarios. But I have seen how the system has progressed, and it is by far the most rounded of all the systems out there IMO. It makes no difference to me who wins the race, I just find it interesting. From what I am seeing Tesla have some huge advantages over the competition.
When "full self driving" can cope with a *Mumbai* rush-hour, or the "Arc De Triomphe" roundabout in Paris, then and only then will I call the technology ready for general use....
The more assistance a car gives the less attention the driver pays. I am not sure excessive self driving is required, just proper auto braking collision avoidance.
Interesting. Can you explain what advantages Xpeng has compared with Tesla FSD. You would have to say, none. Yes I agree, any tech that makes roads safer is great. Well done to the Xpeng Team. For people who follow the world of self-driving tech, I don't know anyone who would dispute that Tesla is 5 years ahead of the competition. I have followed The Fully Charged Channel from the start and I was even a Patreon supporter for 2 years. I have seen the channel go from big supporters of Tesla to the channel's current stance that other EVs are just the same. I realise that the Fully Charged Channel needs to show and talk about all EVs available, they should give credit where credit is due. If I only got my information from The Fully Charged Channel, I would be driving a VW iD4 Instead I have owned a Model S, a Model 3 and a Model Y. Software amazing Charging infrastructure amazing Car amazing Looking forward to FSD in the UK I would not get any of this from another brand of car at the moment. Colin
Tesla pursues a much harder, universally working way. No, xpeng is not solving fsd by any means. That thing has to get to a mapped main road to work somehow. The author has no clue whatsoever.
Completely agree..we've seen many perfect and many imperfect runs with Tesla FSD. We would need to see a lot more, diversified tests from this car before we can determine how good it actually is
@@eclecticcyclist I'd say updating a map once a day is comparatively easier than having a car predict a map with 100% accuracy in real time. But I may be wrong 🤷🏻♂ I don't really care what tech is used as long as the car can drive itself :D
I’m wondering if the roundabout in the video is actually a traffic circle. They’re different things and the traffic circle is quite horrible compared to a roundabout.
Doesn’t their solution require maps? So if roads change; the car needs to be made aware of it? Tesla autopilot is made to map in real time so it can drive theoretically anywhere
This video is simply wrong, and makes me think it's deliberately miseading.. it then makes me think I wonder who is paying you to make this rubbish ?? •Tesla do not use radar as you claim, currently it's only a camera based system. •Xpeng system is currently a geomapped system, and this is what is featured in your video, this means that it can only drive autonomously on areas of the planet where a geomap has been created for its system, in this case probably Guangzhou only. The amount of work required to create geomaps is massive, and the bigger problem as you probably know, is as soon as you publish a map it's out of date the next day.. the system needs to be able to deal with the real day to day road environment variables, to be fully autonomously it needs to be self-learning, this is Tesla's method. As an engineer myself, and my son working on a geo based navigation systems, we know that Tesla's method is only system at the moment to achieve self-learning autonomous vehicles.. although some labelling is still checked by humans but this is reducing quickly and the aim is to be zero human checked. Even Xpeng themselves have stated that their aim is to achieve system similar to Teslas which does not use geomapping. The geo mapping method is probably being used in the car you tested for ensuring safety, & used in tandem with their non geobased system. From your video evidence it seems Xpeng I have made great progress dealing with all the object variables in the driving space, and it seems to me that maybe this is the second best system available at the moment, and it's good to see that Tesla may have some real competition. I suggest checking Teslas recient 'AI Day' video. You promoted that having radar, lidar and cameras is better than having only cameras, because you said "Tesla only have cameras" .. with my engineers head on again.. remember what Elon says..'the best part is no part' .. he's referring to components in car.. as this reduces costs massively and probably increases reliability. I understand that in certain situations radar can produce better object data, for instance in poor visibility, and for this reason Tesla is reconsidering radar, however the data that Tesla is receiving in such driving conditions has proven that the cameras only are orders of magnitude better than human human beings at driving safely in these such conditions.. but hey let's make it even safer by radar if its worthwhile. Checkout utube channel AI DRIVR if you're interested in seeing how Tesla system deals with driving across San Francisco using cameras only.
As someone who has worked with LiDAR, it doesn’t work so well when someone is wearing black either.. but much better than radar, and certainly better than image processing like Tesla does.. LiDAR is super duper accurate for depth info too.. down to mm usually
Lidar is a fools errant, just like hydrogen. Lidar does not help you one iota to drive your cars better. Its all in the Software. Lidar just measures the obstructions in a road better to mach the route with the built in HD map more exactly. Lets drive on a unmapped road. Now lets take cars with 25 or more sensors like BYD, Nio or XPeng and then take a car with 8 sensors - a Tesla. Which one will drive the unmapped road more safe? The one with lidar or the one with the better software?
@@wolfgangpreier9160 : I'd like both personally. knowing exact depth information of pixel im looking at when evaluating image data is kinda important. Depth info is useful, it's why we have 2 eyes.
@@billybollockhead5628 I want a working self driving vehicle. I don’t care whether they use LiDAR, vision or magic. I can not imagine that XPeng comes into our little village in the middle of Europe and makes a HD map of every road. That’s why I count on Tesla. They do not use maps. And their autopilot works today.
LIDAR is great for certain things, but Tesla has a point when they said it's a crutch. It makes things _too easy_ in certain situations, but sets you back severely once you start relying more on it - for example it can struggle to work properly in adverse weather conditions (rain/snow), and you need vision anyway for all kinds of traffic-related visual context.
Tesla doesn't use Radar anymore. They're also phasing out Ultrasonic as well. It will be Vision (Camera) only in the near future and going forward, at least for time being.
I thought Tesla already removed the radar from their system, and is now also removing ultrasonics... leaving just the cameras?
Yup, Tesla is on vision only system.
Yes, you are right. :)
yes, this video has many mistakes
Tesla gets rid of all sensors except cameras. Keeps the profits.
Yep, but silly that the presenter is not aware of this. Tesla purposely removed the other systems and feel that the software can do all the needed things with 'just' camera's.
As far as I understand, XPeng approach is more akin to Waymo or GM Cruise. They are using HD maps and lidar. It’s really faster and a short cut to get to localized L4 autonomy. However, it is more expensive and geographic constraint. The real winner will be the one getting to generalized FSD L5 autonomy. That is a lot more difficult and as far as I know, Tesla is the only one focusing directly on anytime anywhere autonomy. Time will tell who will win.
Spot on
This comment! Poor choice of title by one of my go to channels. I enjoy the Chinese EV content normally, but this just seems poorly researched. Tesla has a fleet of millions collecting data. I really can't imagine anyone beating them to it. 2024/25 will be interesting times for FSD!
Elon's fanboys are in total denial here hahaha
@@teofilol2666 as expected 🤣🤣
Yeah first thing I noticed when I saw the screen. I think both can work side by side but Tesla up until now has focused on FSD driving as if it has no idea about the layout of the road ahead and has to learn it on the fly. This system has all the roads and layouts preprogrammed so it knows exactly whats coming up and wont get caught out by bad markings. I think Tesla is starting to do this too though so they can eventually have both going at once. It's a shame that this video and channel doesn't mention this stuff, its a very base level knowledge of how this and FSD works, essencially just point out that this has lidar and Teslas dont, but thats not the biggest difference like you say.
Tesla doesn't use radar anymore, they're vision-only with AI/machine learning/neural nets.
Ultrasonic is still in use; but it being phased out like Radar was.
@@audiovideophile5317 ultra sonic right now in FSD is not used. Ultra sonic is used only for parking.
I'll make sure to never ride in any new one. Dangerous
@@Mguy393 also old one are vision only right now
@@DavideMoriello right. I think they started dropping on early 2021
That appears to be a highly detailed road map/model. How does this car do outside areas that are not mapped in detail?
Or there is a change in the road that isn’t reflected in the map yet.
Its using 5g to assist. That's not available in the USA.
@@craiggosselin6356 That what the cameras are for.
@@TegraZero I'd love to see d***heads getting fined for driving in the overtaking lane all the time for no reason
@@dstr1
5G is short range, easily blocked.
(Yes, I DO have 5G)
The aim of this technology is to make the roads safer for everyone. Let's hope that it is a success.
I am very very skeptical. Its like that Boston Dynamics jumping robot. it looks great (on the surface) but it is as dumb and a sack of spuds. We are far away from self-driving being safer. oh, and i makes car soooo boring.
Perhaps equally importantly would be this technology for buses. London buses take out quite a few cyclists.
@@bluceree7312 tell me more about those spuds, what were they doing that was so dumb?
These technological leaps will continue to exacerbate the meaning crisis humans are already facing. Driving is very enjoyable for many and this technology will inevitably lead to human participation behind the wheel becoming illegal. We are literally innovating our own obsolescence. Think I'm crazy? A.I. could be doing almost everything including its own maintenance within three or four generations. The horse and buggy analogy doesn't hold water any longer.
Turns out Ted Kaczynski was on to something.
I very much enjoy and appreciate the work of the fully charged show. That being said it is surprising, maybe even slightly unforgivable, that the host can be so wrong and so repetitively wrong about Tesla and its radar. This should’ve been picked up way before the video was shot, and if too late for the shoot, disclaimers should’ve been inserted all over the screen in post. Keep up the good work, I know you can do better. And if you have been, thanks for reading…
My thoughts exactly. I’m a fan of Fully Charged in general and consider them a good source of information so the repeated errors about Tesla and FSD really stood out and are very disappointing. Elliot is my least favorite host on Fully Charged and this isn’t the first time he’s come across as uninformed in a video.
Didn't do the research and then boldly stated in the title that this one is "winning"
@@gormauslander Agreed. Also, Pretty sure its a geo-fenced city. Not useful in all those other cities where the other 6 billion people live.
He should take it to a random city and test it there.
Try it on a random country road.
Then tell us who's winning.
Don't blame the host, blame their script writers.
He’s the worst host of the channel
Tesla doesn’t use radar anymore
Gonna be amazing with any amount of Tesla robotaxis on the road when weather hit. How many Teslas does it take to create total gridlock of a metropole? If they are ever allowed to operate at level 5, we'll find out the first day of special weather. Just when we need the grid most.
@@Cloxxki not had any issues with FSD Beta during weather recently. During heavy rain it slows down and turns wipers on fast. During fog it slows down but sees much better than I can in the fog. Don't know how it would do with smoke or heavy snow.
@@Cloxxki 😆 🤣 yup
@@Cloxxki Tesla’s FSD works extremely well in rain and fog. The neural nets are able to compensate for the conditions and the car can see better than a human in those conditions. There are plenty videos out there. You should try watching one.
This video probably was recorded before the Tesla's announcement.
I have never seen a not positive fully charged video. All cars are good everything is nice all well here. I wonder if it has anything todo with pleasing the manufacturers. It would be nice to see an opionen sometime, any time
Facts are not opinions... you want a bad review go see them... there are plenty on youtube.
So many chaperones driving with Elliot. I would how many he had to do over the comments to satisfy xpeng.
@@XFDADX There are precious few facts in this piece of useless fluff.
I imagine Robert will address the comments we've made about Tesla's not using Radar for a little while. And they stopped using Radar on Full-Self-Driving Beta last year around this time actually. It's been vision only for a reasonable amount of time. Even Robert's car should have gotten that update and told him by now. I'm not that offended that it was wrong but like others have said it is reasonable to expect Fully Charged to know this about the Tesla cars seeing as they are by far the most popular Electric car brand and have been doing autopilot for so long.
More research needed.
The point of FSD is it's not city restricted.
No LiDAR
No Radar.
Now, no ultrasonic sensors.
Google precision maps, lidar and radar are things Tesla is NOT using. And now the high frequentie sensors are taken off the new models too.
Wasn't the last release using the precision maps? I thought I saw something in the notes.
@@audiovideophile5317 no they can build precision maps from the cameras but it doenst require maps at all to actually drive, latest build on roundabouts not even on the maps works fine
In an interview, the former head of Tesla AI told that the sensors creates additional noise to the system and doing more harm than good. Also, the advantage of having sensors is tiny but it costs a lot more compared to just having cameras. That’s why they removed them.
The evolution of these vehicles is continuing. These self-driving cars will be of great benefit to elderly drivers. They will allow older drivers who would normally have to stop driving as they age to still maintain their freedom and mobility. 👍👍
Wow what a nice high contrast perfectly painted road. Just one spot with worn lines
I think that's the only geo fenced area the company allows the system to activate
Elon's fanboys are in total denial
@@teofilol2666 see you at the finish line using a telescope thru the rear window 😎
@@teofilol2666 I think good competition would benefit even Tesla fans, and maybe if it proves itself, people would consider shifting from Tesla. The reality is though, this test was done in completely optimal scenarios and was only done 1 time. You never test something like this once and say it's perfect
We would need to see a lot more tests before determining how good it is, but I'm optimistic
Bothers me when you get a fact like Tesla only using vision wrong. On a channel dedicated to EVs. In saying that, love what Xpeng is doing. Hope it comes to Australia!
Hmmm…I have this lingering suspicion this was an advertisement piece. Presented to you by Xpeng.
🙄
Interesting that he had to drive out manually to the main road before engaging. Tesla FSD beta can be initiated in many cases from parking lots and from peoples driveways and the car can drive to the main road. I did notice however that this Xpeng did make a u-turn which Tesla currently don't do but this is a function of the the digital rails that are pre-mapped within the geofenced area. Tesla will likely do U-turns everywhere (w/0 pre-mapping) due to their FSD approach.
A few examples of U-Turns have been shown by FSD testers, but it isn't baked in yet. Those were just flukes, until they focus on them for a week and then U-turns are magically solved
The letter X makes a 'SHA' sound in Chinese and 'peng' is prounced 'PONG', so put to together, "SHA-PONG"
Impressive, and Elliot wasn't even shaking when he got out so the system must work very well (given the challenging traffic conditions).
I've visited Xpeng head quarter in 2019. Back then is very impressive progress for a 4 years old company.
They took advantage of Tesla's open patents... The few car companies that did that have all gotten ahead of the legacy automakers...
@@0ooTheMAXXoo0 ??
It's likely because of when this was filmed but its worth nothing that Tesla has dropped radar going forward. Tesla Vision only was certainly the long and hard path, but it seems it may have been the right call.
Why is radar worse?
@@jeffmocha9478 Tesla says it adds noise, which creates errors. Also it's just cost, weight etc. And certainty don't underestimate how much it costs to keep developing/improve ON TOP of vision. Essentially the same can be said for LIDAR except amplify it.
On the other hand, many believe Teslas approach cannot be successful.
@@LZentertainments But surely LiDAR is more accurate then just using cameras. Surely a mix of both would allow for the best of both worlds?
@@jeffmocha9478 my tesla drives only with cameras and it‘s totally fine… it can keep the distance to the car in front of me, slow down smoothly when there are cars in front of me or when the car in front of me slows down , and it even stays on the right side of the road when there are no lanemarkings /the lanemarkings dissapear.
@@jeffmocha9478 It's counterintuitive, but no, it would add noise (confusion) to the system. If the cameras and radar/lidar disagree, what does the system believe? The cameras are much more information rich sensors and with capable AI can become superhuman at low cost.
Brilliant video, as always. Impressive FSD. Tesla vision does not use radar. But jury is out for who can drive 10 times better than humans, on every city road, first. Eamon
There's only one company preparing for every city road though...
Bristling with fricking lasers Elliot! - just like the latest NIO I looked over in Berlin a few days ago. Video of that up on my LinkedIn feed. Another lush and lovely vid my friend! 👏
been a huge fan of xpeng cars for well over a year now, really hope they make it to the UK.
Me too
I think they're making a flying car too
I doubt it. This is just the sort of Chinese technology that GCHQ warned about at the beginning of the week.
@@DavidGrasty good grief really?
UK just declared China as an enemy
According to Sandy Munro, Forward Looking Infrared(FLAIR) is used by the military for guiding its missiles. He says it is more accurate and detects things better than LIDAR in all types of weather. It is also more expensive than LIDAR.
I am aware of 6 different LIDAR companies, most of which I own some stock. LIDAR manufacturer, Aeva's , has announced its LIDAR system will be used by NASA to map the lunar surface.
You might want to do a video on the different types of LIDAR systems.
I wonder how long it would take to fall apart on a roundabout in Spain. Especially in Valencia where you can have a few lanes entering the roundabout, a traffic light almost immediately for the next entrance where cars just line up side by side with no lanes at all and then a mad dash when the lights turn green to get into one of the few lanes leaving.
Spanish roundabouts are special, its like they are created to cause crashes, who ever decided that it was a good idea to use the outside lane to go all the way around is an idiot. A challenge for humans so a good test for autonomous driving.
Sounds like Spain needs to get better at designing their road rules.
I have Tesla model Y and I think relying solely on video cameras for full self-driving is short sighted. Video cameras are great for picking out detail, but don't work so well when poor weather impairs visibility. It is much better and safer to use a range of sensors such as cameras, radar, lidar, etc to provide an augmented enhanced visibility for self-driving, which improves on what humans can do unaided today.
Hmmm .. you realize lidar is light based as well and is much worse at bad weather - and there are weird lidar issues like mirroring as well…. Radar also has it‘s challenges - I‘d say Tesla has proven it can win Euro NCAP without Radar on the car …. The jury is out for USS vs Cams but I‘m rooting for Tesla as well :)
@@_baumi_ I think the main problem with visual is processing speed. Lidar space mapping take minuscule percent of processing of visual identification. I see FSD slows down and can't think fast enough in dense city traffic.
@@enigmatum I've never seen FSD slowing down. Can you mention a video where this happens? (Title/timestamp - don't post links, your comment gets removed)
Remember that the on-screen visualization is independent to FSD, and runs on a separate computer system. Even if the on-screen visualization is lagging or even freezing, FSD would be unaffected.
I don’t think it’s short sighted, after all we humans only have a pair of eyes. The system only needs to be better than us to work. Also, having three different inputs (visual, LiDAR, radar) may be very difficult to work with if you start getting conflicting data (something Tesla explained as the main reason to abandon the others).
@@logitech4873 it was a 2-month old video from AllAboutTeslaY. I can’t find it on his list quick. 20 pedestrians and heavy traffic right turn in NYC overloaded FSD. It just died.
Looks very capable. But with the expensive lidar sensors this will only be reserved for the top end cars. This system won't end up on a $20K car in the future (when battery prices drop enough). And, as he said in the video, I wonder how this will work in crappy weather. Radar can see through mist, but gives too many false positives and resolution is low. In the end, vision is the only sensor required to drive, as that is what humans use. Different approaches, different results. It's going to happen, time will show us how.
It’s economics of scale, the cost of LiDAR will decrease as more people use it.
I was driving at night recently on some windy mountain roads without any lighting other than my headlights. It was scary and I had to drive very slowly. Presumably with Radar and Lidar I could have managed a more normal pace.
@@lamsmiley1944 exactly, imagine the price of a camera, which is used not just in cars, but in so many appliances already.
@@MrBra1nDeaD yep, even my iPhone has LiDAR in it.
LIDAR is already down to about 100USD for a suitable sensor, economy of scale will push it down further.
What I really want to see is how traffic flows when many different L4 / L5 systems have to share the road. Each with its own idiosyncrasies and weaknesses. Lidar has the benefit of high precision, has issues with dust/smoke/rain/snow. Radar can cut through most all environmental issues, but has real issues with refined object understanding... though I'm given to understand newer versions are getting much better.
What I understand is the real difference here is between those that use Lidar approach and the Tesla approach is precision pre-mapping. You can build in smarts to handle weird events with the Lidar system (debris on the road, construction, etc) but they are fenced within the pre-mapped regions. That's fine if your area is mapped and you don't care about manual driving between pre-mapped areas. However, the Tesla approach is AI driving. They might take hints from navigation maps and GPS, but really the system drives as if it's the first time it's seen that stretch of road. It has to see and understand, then drive. Upshot is this works anywhere with no pre-mapping. Downside, it's a way harder problem to solve.
Anyway, great vid! Thanks it was really interesting. Looking forward to more.
Perfect
The tesla vision only system will be the degrade mode for system metioned here. Tesla is the only brand in the industry has such wide West like high tolerance on exposing their consumer to risk. There has been few accidents caused by failure of FSD in China. But their consumers base do not have much interest to pursue legal action against tesla. If it is a domestic brand, you will see media coverage flooded every where and the consumer will have class action lawsuit to suit the company to bankrupcy.
Just wondering if multiple vehicles equipped with radar and Lidar in close proximity would they interfere with each other or create a high noise floor that they wouldn't work properly, IE they couldn't see the reflected signal through the noise of other equipment operating on the same frequencies in the RF and light spectrum.
I think no for radar, radar reads, but LiDAR yes, however these cars would have to pointed right at each other
Reply if I’m wrong I’m not 100% sure
@@h20dancing18 not quite sure myself but radar transmits a radio signal then measures the doppler shift on the reflected signal if you have many car's on the same radio frequency or close adjacent frequency will it cause problems with the receiver on the car, I don't know how many car's on the roads have these onboard systems yet I suppose time will tell as more car's adopt semi autonomous or full autonomous driving.
@@adam-g7crq I would assume these radars are quite advanced but radars to keep your cruise control following another cars speed is very common.
If I had to drive with my hands like that I would rather drive. See how long you can keep your hands like that.
Self driving is not a hardware problem; it's a software problem. Vision, lidar, radar, flir; doesn't matter in most cases. It's what the computer does with the info.
In most places yes but not all places
the more sensors you have the easier it’s for the software. If you’re doing it correctly.
@@EVPaddy No. You want as few sensors as possible. Less data to process and no conflicting signals.
@@morilot NO! So if radar can see a car in front of you but the camera can’t because of fog, you absolutely want to ignore the radar because of less conflicting data? NO, it’s stupid. You want to get ‘super vision’ for the car, not some 2007 smartphone cameras.
@@EVPaddy Those "cellphone cameras" can see way better than you in fog, especially since they don't use IR-filter before processing.
Just look at the statistics.
How many Teslas has collided in fog while being on autopilot?
I think you'll find that Tesla no longer uses radar/lidar, they work solely with cameras and their Vision software.
Most impressive. I get nervous when other people are driving, can’t imagine when the car is driving itself…lol. I wonder if other car’s LiDAR beams will interfere with each other when more of these cars are on the road?
every car will use a different wave length... .if they can put millions of beams in one Optical fiber and don't interfere with each other, I think they can ''fix'' this too.
I don't think that bats using their radar will interfere with each other, so I imagine it's the same for cars and Lidar.
@@CATinBOOTS81 bats use sound, lidar use laser, not the best comparison 😃
@@johnsmith-cw3wo wavelength is different, but the principle is the same
One question I have that I don’t understand is: tesla fans are arguing that a big advantage it has is the large number of cars on the road collecting a lot of data. but tesla fans are also attacking xpeng and its system for significantly using data it has collected to make maps of areas, And saying that this means Xpeng is limited to driving only in areas where it has data maps. But isn’t the majority of data being collected by Tesla cars also related to the streets, towns and cities that those cars are driving in?.
Also, it seems highly probable that in the next few years that 99% of cities and towns around the world will be mapped with the type of mapping that xpeng is using in this video. It seems that either a big company like Google will do maps like that and sell them to the other car companies. Or that the car companies will sell the maps to each other and or trade the maps with each other so as to get global coverage. For example, the Chinese car companies will sell or trade their maps with American car companies.
Also, most of the cars sold by the big companies are sold in North America, Europe and China… So they really only need advanced mapping in those three regions. Between all of the car companies, autonomous driving companies and tech giants like Google and Applle, I think they will fairly soon be advance maps of 99% of those three regions. At that point, even if tesla develops the ability to do autonomous driving in unmapped regions, I don’t think it will matter commercially speaking.
The final point is that driving in China is pretty crazy due to the number of pedestrians and bicycles, which result in many unpredictable things taking place. I think that if Chinese companies are able to handle those things well, they will be able to handle most other unexpected things.
This is not comparable to Tesla's solution but great clickbait... Disappointing. Robert could you fix this?
The big danger in self driving systems is the interim stage which is where we are now. The systems are improving and can be impressive when operating within their prescribed parameters. However people are likely to be fooled into thinking that these systems are infallible and will exceed the operating limits eg drinking, watching videos, playing games and probably even sleeping. This will be fantastic when these systems are truly autonomous but at this stage they are likely to contribute to higher accident rates than they are to lower them
I share your concern and manufacturers are well aware. That's why some also build attention monitoring systems in the car. It's not a 100% guarantee, but will make it more fool proof. In the end irresponsible drivers will always be on the road, no matter the type of car, unfortunately.
Still a lot better than your average Chinese driver. That car is the only considerate driver on all the roads in China!
No, Tesla don’t use radar.
That's Tesla's weak spot right, the reason Tesla autopilot crashed into a white truck, because the camera saw it too late against the white clouded sky
sort of. a new Tesla now is camera only 2021 and younger the MS and X also used radar.
They did, it's there models that are in production at the moment won't have them.
@@a-aron2276 radar has been deactivated in the cars that have it built in. They moved to vision only
Yep my model 3 moved to vision only now. I have a radar but it's no longer used.
Impressive, XPeng is one of the best in China if not at the top for ADAS. It seems like many are looking at LIDAR now, including NIO, Huawei, etc .. and other brands.
I think main stream solution are mutiple sense system.
Actually quite a few roundabouts in the US these days, they've gotten popular
Still to few. In Europe almost all new crossings are roundabouts, traditional crossing exists only in pre-existing crossing (tipically in old cities) with too little space for a proper roundabout.
I guess it depends on the area, here in The Washington DC area new and upgraded intersections are generally roundabouts
@@Saxafruge That's great
Correction needs to be made to this video, Tesla now uses vision only for FSD.
High precision napping... Nice one 😂
Well it's a highly precise self driving system... might as well get some high precision napping while you are at it...
The problem of autonomous driving has not been solved! Claiming more sensors is better is like having 2 plugs every appliance. Wake up!
Tesla dont use radar, cameras only
Tesla's don't use radar or lidar or ultrasonic any more, only cameras. For a video talking about self driving, you sure did get EVERY detail wrong....
its a cool car, hope to see them in EU too!
seriously interesting btw but why only 90% of situations? what happened to get you to reduce this from 100% or even 99%? yes it was a final beta by the sound of it but what did you cut out?
3:35 The lidar detects there is an object there, but needs the camera to identify that it's a person. It only measures distance, which with today's AI, cameras can also do that.
Except they don't actually measure anything, it's a good guesstimate
Not entirely true. Lidar can make out rough shape which can inform what something is
It depends how many lasers it using. If you enough laser pulses quick enough you can generate a image.
Exactly. All lidar does is estimate the distance of something using pulses of light. It takes way more to take that data and extrapolate that it's a human. That's where Tesla's Vision system is superior. Two actions in one, see object and identify object.
@@indyalx so it's a very very very low resolution camera?
it's great that others are working on FSD too, but the compression and ignorance of Tesla's system make me wonder about the quality of the overall research of FCS.
Good to see Elliot covering EVs for us from the inside as usual!
But gosh XPeng, I wanna be a fan for competition, but that double tap stalk to activate driving assistance is a little too familiar?
It's really not the best activation method and you gotta just have ways to do that better right
Melon Melon: So if interfaces have similarities, it's a problem? Why?
I don't want to have to learn a completely different system every time I use a different device from a different maker.
It's bad enough with computers and different OS's, for example, already.
I think it's one of the best activation methods, right on the driving selector
@@rogergeyer9851 I think they are implying that it's not a good activation method on any car
Many people misunderstand that it's Lidar vs Camera but it's more like HD pre mapped vs Generalized vision.
11:08 is the most impressive to me.
Like it noticed the car, two cars in front of you, that wanted to turn left. Shifted lanes for a more uninterrupted drive ie efficiency.
I’d say well done 👌🏼🤩
There's plenty of footage going back several years that show Teslas predicting other cars ahead of it having an accident and avoiding them. Tesla also seem to map people and roadside objects better. Elliot was also too impressed with the car slowing down at traffic lights where the lights were obscured by the vehicle in front. Didn't it occur to him that the car can detect THE VEHICLE IN FRONT? This level of self driving is great but it's old hat for Tesla.
The Xpeng is great and self drives very well. But, I haven't seen anything in this video that a Tesla doesn't do already. Elliot is impressed with the way the Xpeng slows down and speeds up as cars intrude on it's lane while driving. I hired a Kia e-Nero a short while ago and active cruise mode does exactly that. I set my speed to 80mph and the car tracked the speed of traffic ahead of me and kept me in line with general traffic speed be it 70mph or 40mph. When someone jumped into my lane it would instantly slow down and when the lane ahead of me cleared it would instantly speed up again. I don't know how Elliot is so impressed with what are now standard features in EVs. And he really didn't explain exactly what the advantage is of lidar. In fact, he explained the disadvantage of using it at night. Also he was driving the Xpeng around a part of town adjacent to the Xpeng facilities. The car is familiar with the area and the surroundings are also well mapped on the satnav system which gives the Xpeng a home advantage as it were. Maybe the gentleman sitting behind Elliot had something to do with his over-praising of the tech?
He's impressed because he's utterly clueless.
I think you dont get it , can you choose a location on a tesla map and let it take you there. ? Thats what he did , autonomous driving, you get into the car , open map , pick your location and the car makes it happen. The others you drive to a location while sonetimes letting the car do some stuff like keeping lanes , keeping distace , mantaining ceuize speed etc . This is complete " car take me to work"
That was not a real test. How about driving on roads that don't have great road markings. Also have some tight areas with cars parked on either side of the road.
Also tesla don't use radar anymore.
So much FUD from Fully Charged, did you have Chinese minders from Xpeng telling you what to say?
What's the price difference between that car and a Tesla Model Y?
Of COURSE it was a real test, Fanboi..!
It was a test of the car's ability to self drive in a city - which appears to work .....
Stop being a pathetic, cowering slave of St. Elon - you poor, poor fool ....
There are a lot of questions I have about this: is this a swarm system? Does it only work on roadways that are already mapped and in the database? Is the system analyzing *all* the moving objects?
Ofcourse it is. You can map the road but not the traffic
It is not clear from this video why it would be better than Tesla solution. The (current) title “This car is winning the race to Full Self-Driving” smells like click bait.
... and!?
I've been in an accident when I fell asleep at the wheel. its a wake up experience in life.
This is an extremely poor episode by Fully Charged. The headline is highly misleading....no, this car is not "winning" full self driving. It is using different technology compared to Tesla and if the presenter had researched properly, he would have been able to highlight the pros and cons of each. A simple search of FSD on UA-cam will highlight how Tesla's FSD has reached impressive levels compared to this very limited Xpeng "demo". Both will still have a long way to go to really become fully autonomous but to label this as "winning" is really disappointing.
I own a Tesla and although build quality is improving, the zealots in the Tesla Vision department are making the cars worse. There’s no sensor redundancy, no radar, they’re even removing ultrasonic parking sensors. Everything will rely on the cameras despite their blind spots and susceptibility to dirt and dazzle from lights and the sun
It's very good to see competition in this sector, because that will drive up the function and drive down the cost. Better than FSD? I'm saying no after watching this. You only need to see how the car reacted to the pedestrian after the U-turn to say it's not at the same level. In time, this and others will be acceptable for 99% of driving scenarios and we'll all be safer.
This is china so even if everything he said is what the company told him, it is highly probable that this is not the ole story. This is the country that paints grass green and staples leafs in trees. He said it drove 99% without interventions but the video shows 100% without interventions so something happened that didn't make to the final video.
🤣 this thing did far better than every fsd video I have seen. But keep fan boying. Tesla can't even park itself or navigate a traffic cone in the middle of a lane. I have seen the latest vids.
It handled roundabout better than Tesla ever has in my experience. So depends on the measure.
@@jeffsteyn7174 you ok Jeff?
@9.55 in china the standard satnav has info of the traffic lights as well as the GREEN/AMBER/RED state including a countdown timer .. the satnav informs you of the GREEN light and alerts you to go .
i think that info may also be in use here ..im planning to get the G9 next month 😊
Impressive. Now they just need to advance in another area: stop the genocide, and stop making threats to other countries. Then it could be a great car.
“They”.. sigh.. the minute you class everyone in a country as “them”, then you are as bad, or worse, than “them”.
It's all about the visual systems and algorithms and processors.
How well does LIDAR read road lines and road signs?
1 it doesn't, but that info is in the maps
2 the car ALSO has cameras that can read lines, signs, etc
how well does a camera read depth?
@@frangalarza not great if you have 1, but ok if you have 2. Questionable if there is not reference, which usually dont. That is why they use laser in the industry when precision measurement is required.
Okay, this is interesting but I feel there are a couple of points to be made, firstly Tesla FSD no longer uses radar and is phasing out ultrasound so I lost some confidence in this review within the first few minutes. Secondly from what I could see 99% of the driving was done on duel carriageways, which isn't 'city' driving. I would also mention that the car was full of engineers so I'm not sure you could call it an independent test, we're talking about China here obviously.
I'd be interested in seeing another video with this car when you're allowed to drive it for a period of time without the minders then we'd see how it really performed.
TBH all this video showed was how far Tesla is ahead of the field.
I don't think there is another company within 5 years of Tesla on the self driving front. They're only using vision and are trying to basically solve General AI, whereas the others are using high res maps, radar, lidar etc.
After watching this video I feel even more impressed by Tesla's FSD in comparison. Tesla uses only cameras while the car in this video have to use cameras, radar *and* lidar to be able to drive in a limited area.
It is the amount of risk the company willing to take to release it to public.
@@林振华-t4v they're going to need to release it to the public if they want real testing done. The data they'll gain from regular users will be overwhelmingly more useful than these controlled/limited tests
@@realshnx most companies wont release alpha sofware to public and release beta sofware to Limited or selected public, some orgnization will add and Release Candidate phase to general public to sort out what you metioned, have as many people as possible to break the sofware . Tesla going out her own way asking 1.3 K for a beta (no, not me, I did not opt in the beta auto drive on my model 3), and basically having mutiple major outstanding engineering issues on pretty much every (except the Y) model at its first generation release. They pretty much try to patch issue over the whole life cycle of their first generation of the product. That means every unit sold is at a lost, because the design did not get frozen long after the mass production started. That is why they did not have a profit year till 2019 or so. Think about it, burning hundred million if not billion year by year since 2012( just for the sake of focus on the serveity of the probelm) till 2019, 6 years consecutively. That is not a sustainable model even as tesla themselves. And now look back, how far has tesla get ahead with the game? Not so much Id say, Other solutions rely on infastructure and mutiple sensors hace achived same or better result. Ditching other sensors and focus solely on visual spectrum camera will somehow reduce the difficult on system integration. Yet, the lag of data aquisition method will limit the system capability. Even us human wont solely rely on our visual for driving. We also hear, feel and even smell as well. So, what will the outcome of tesla visual only solution moving forward. There is a good chance it can only achived what other systems can achived at degrade mode. e. g. other sensors all failed while visual system still operational.
Another brilliant video from Elliot and I love these little snippets of life in China, but I am not a big fan of these systems, problem because I love driving. As for those Tesla fan boys below banging on about this video not acknowledging "Tesla doesn't use Radar any more" wrong they do use Radar, because Radar exists in their proceeding models. So it would be technically incorrect to say they don't use it. Yes they may have decided not to continue with it! But then this video isn't about Tesla is it!!
They don't use it in preceding models either. It's been disabled via software.
Sure, you can say it's used for the oldest cars that don't even support FSD - but that's irrelevant.
My Favourite presenter 😎
As far as I understand 10/12/2022 Tesla uses a camera based system. No Lidar or radar. Is that correct?
Hi Frank, absolutely correct, Elliot got that wrong. Tesla just git rid of the ultrasonic sensors too. And of any maping a few weeks back. Vision only. And of what I see Tesla works pretty much as good as this X-Peng, which would deny service once it gets to the limit of its determined area. Then the Tesla would continue to explore, analyse, decide and learn and especially: continue to drive you.
Tesla has never had LiDAR on production cars. They removed radar over a year ago (3&Y, S&X early this year) and disabled it in software.
Recently they removed ultrasonic sensors on the 3&Y being produced. They let those with existing sensors have it for now (and I wouldn't be pleased when they disable it)
I'd advise you to look up the video where XP crashes at full speed into the guy and his car (killing him) that they were travelling to help after he broke down/crashed.
Edit: as a side note, if you're around 6ft4 or upwards don't expect to be able to drive any Chinese electric car apart from the Nio ET7. They all have at least one of the following issues; low roof, too high seating, top of window is too low, seat doesn't go back far enough, headrest is fixed and pushes into your neck and not your head.
The last one used to not be too big of an issue, but considering how some electric car companies are with modifying their cars, you might just void warranty etc by changing it..
Figures..
Tesla doesn't use RADAR
"Safety is at the core of our design and engineering decisions. In 2021, we began our transition to Tesla Vision by removing radar from Model 3 and Model Y, followed by Model S and Model X in 2022. Today, in most regions around the globe, these vehicles now rely on Tesla Vision, our camera-based Autopilot system."
Tesla removed radar a long time ago, I thought?
The radar they drop is likely the parking sensor radar. Not the lidar seen here. Which for tesla, not a huge lost consider parking radar low resolution and refresh rate
@@林振华-t4v Tesla has announced that they are going full "Vision," relying only on cameras, even for parking. They even just announced they're removing the ultrasonic sensors, disabling self parking and summoning (for now).
@@QMaverick1 the parking sensor is till working on my model 3 as of firmware 2022.28.2. I will see if it get disable on next revision.
Can Lidar interfere with the Lidar systems of oncoming vehicles?
Not likely, recieving signal from other lidar will be consider as noise and throw away by the system.
Very nicely put together vehicle, in terms of road experience and appearance. It's been really neat getting a chance to see vehicles that, sadly, will probably never make it to the US.
That nervousness is the result of programming. It will take programmers to dial that out. Not the car just randomly roaming the streets.
Sorry, the only model we have available that comes close your requirements is "Chinese Fishwife". 🤣
Interesting, impressive and strangely compelling/suspenseful (helped by the music I think). As others in the comments have said, I think lidar and radar are the norm in these systems and Tesla are deliberately going camera only to bring a more affordable system to market. I'm not sure how cameras deal with fog and rain but as they can have access to a greater range of wavelengths than us humans, perhaps these things aren't a problem. I don't know how Teslas system would compare to this either, but I understand that they have an advantage in that they have many more miles of data to work with than others by virtue of their massive fleet of test cars - aka customers cars on the road, and it is data that makes these AI systems more competent. I can see how a company like X-Peng could catch up on this front though as they have a huge market to sell into (but so do Tesla). The winner might be the company with more data, let's see.
It wasnt really about affordability so much as the data bottleneck. By going only with vision, Teslas system doesnt have to get the data coming in from several totally different streams and converge them together. So they can focus on getting a higher bandwidth of data on just the visual channels. So they were able to upgrade to higher resolution cameras for instance because they can process that data so efficiently now. And the data they were getting from radar was essentially redundant to or of generally lower quality than what they are now able to achieve with just vision. Even in adverse weather conditions.
Which when we say "vision" we have to understand that digital cameras are not the same as human vision and digital cameras are all capable of at least short infrared... in fact digital photography cameras have physical infrared filters so the images will look closer to human vision. So these cameras may be capable of seeing better than a human in certain adverse westher conditions even without fancy AI... which they are also working with their AI to improve object detection in fog. And remember that in really heavy rain both lidar and radar systems become just as useless as vision. Although vision may actually come out better in really heavy rain than lidar or radar if the AI's object recognition and persistence of memory is really really good.
@@patreekotime4578 the cameras on current tesla models are not even as good as the one you get from blink. Especially in low light condition. infared spectrum is simply not the spectrum the CCD or COMoS onboard of a Tesla can pick up. And the refresh rate on the current model is simply not there either. That is why for so many years, the sofware is still at "beta"
@@patreekotime4578 Thanks, that explains it well
Def more data = better resolved and safer system. DOJO FTW!!
@@patreekotime4578 the reality is, their tests with vision only FSD has been the most success they've ever had with it. That's why they are confident they don't need anything else
Tesla is not using radar anymore... it's all camera vision and Tesla just announced it is removing the sensors on the bumpers as well.
So what happens when a car with Lidar meets another car with Lidar? Do the Lidar signals interfere with each other? Same with radar.
There's going to be interference from other cars. I don't see how it can be avoided entirely.
They are aimed at a lower angle which will minimize interference. And as lasers they aren't broadcasting over a wide area, but rather as discrete rays sweeping quickly around the area which will help.
But with traffic all around you every day there will be multiple opportunities to encounter cars that can disrupt each other's lidar system. Hopefully they are designed to simply see the other source as a bright point that it can ignore instead of overwhelming the sensor.
We’ve all seen company-produced self-driving demos for years. We’re past that now. Any FSD review has to include full clips of longer drives (possibly sped up) with no cuts. Otherwise it’s a parlor trick.
What would happen if u just put the Car on fsd in a random place with dirtroads etc, not next the hq of xpeng… 🤔
do it with your tesla and report back
@@kumbackquatsta check Dirty Tesla. Dirty road in the dark. Works.
Looks like hardware assisted vision based autonomous driving is the way forward, at least it's more reliable. The hardware should be able to intelligently switch on and off to adjust itself to highway drive, city drive and rural driving etc.
Did you see that our comments are constantly deleted by the Fully Charged Show? Comments like "Reaction of Tesla Fan boys is GOLDEN" are kept but if one is discussing the "facts" presented in this video it gets deleted by the channel owner. Interesting way to handle your subscribers, guys!
@@AndreasWest I haven't had any of my critical comments removed. Yours are prpbably being removed by UA-cam itself automatically, which is far more common.
Don't include links, and try not to have a very hostile tone in your comments.
@@logitech4873 I had included links to other YT videos. Didn’t know that this gets your comment remove, silly
So that's basically FSD beta, but only on mapped roads, perfect weather, and with a lot more sensors. Well....
If you believe reports it's *Literally* FSD (early) Beta....
@@rogerstarkey5390 Nah. It's impossible to just clone Teslas system. It relies on custom hardware and unfathomable amounts of data.
I see these demos, and it isn't starting to ice up, it isn't pouring, it isn't starting to be a snow storm, it isn't wanting to drive down a gravel / dirt country road / cow path. But they want to remove the steering wheel? Whenever they say they want to remove your ability to act, be suspicious, very much so. I assume there are conditions where these systems would just stop working, but no one is addressing these. What does happen in these conditions? How does one drive into a garage without a steering wheel or out of a business as in the beginning of this video? Lots of questions, no answers.
Yep - it's all a huge conspiracy by somebody or other to stop you doing stuff, or make you do stuff...or whatever...!
Zzzzzz.....
The car with no steering wheel for public roads is very very far away. And if it ever comes out somebody will have already figured that out so I wouldn't worry.
Wow this thing beats "fsd" by a mile. Let's see tesla do a u-turn or navigate a multi lane traffic circle.
I have seen Tesla doing some amazing things. their new FSD software is handling some really complex scenarios. From my estimation, Tesla are leading the way by some distance.
Let's see this thing drive anywhere other than it's pre-programmed geofenced area at night when it's raining, and a poorly marked road as well.
@@logitech4873 Tesla fanboys obsession with HD maps is funny. What exactly do you think HD maps do? HD maps doesn't help you avoid driving into other moving vehicles. It doesnt replace cameras or sensors. It's there for planning. Also we saw what happens when a lane disappeared because of temporary construction. Which would not be in a map. Step out of your tesla bubble. If you actually interested in learning how others do this. Go watch waymos vids about how their self driving works. HD maps is only a part of puzzle.
@@daydreamer8373 No you haven't. Any complex situations get the tesla in a tizzy. The last video I saw was last week, when a tesla couldn't even navigate a traffic cone in the middle of its lane. And that's after 8 years and billions of miles. 🤣
And let's not forget the simplest of situations... i.e. parking. Tesla even get that right.
Your estimates are based on fanboy bias.
@@jeffsteyn7174 I have been following self driving for years, and have watched many videos including Waymo. and I have seen them all mess up. Tesla's were particularly bad, and would struggle with the most basic scenarios. But I have seen how the system has progressed, and it is by far the most rounded of all the systems out there IMO.
It makes no difference to me who wins the race, I just find it interesting. From what I am seeing Tesla have some huge advantages over the competition.
When "full self driving" can cope with a *Mumbai* rush-hour, or the "Arc De Triomphe" roundabout in Paris, then and only then will I call the technology ready for general use....
10:52 there are tons of roundabouts in the US...I have some in my small town...and FSD handles them WAY better than this appears to.
And even more commo in Europe.
The more assistance a car gives the less attention the driver pays. I am not sure excessive self driving is required, just proper auto braking collision avoidance.
Interesting. Can you explain what advantages Xpeng has compared with Tesla FSD. You would have to say, none. Yes I agree, any tech that makes roads safer is great. Well done to the Xpeng Team.
For people who follow the world of self-driving tech, I don't know anyone who would dispute that Tesla is 5 years ahead of the competition.
I have followed The Fully Charged Channel from the start and I was even a Patreon supporter for 2 years. I have seen the channel go from big supporters of Tesla to the channel's current stance that other EVs are just the same.
I realise that the Fully Charged Channel needs to show and talk about all EVs available, they should give credit where credit is due.
If I only got my information from The Fully Charged Channel, I would be driving a VW iD4 Instead I have owned a Model S, a Model 3 and a Model Y.
Software amazing
Charging infrastructure amazing
Car amazing
Looking forward to FSD in the UK
I would not get any of this from another brand of car at the moment.
Colin
Tesla pursues a much harder, universally working way. No, xpeng is not solving fsd by any means. That thing has to get to a mapped main road to work somehow. The author has no clue whatsoever.
@@harry-eto perhaps they could get the maps from other people who have driven that road before ? .,.........bing!
Not the most intense test, really. Hard to say it's winning the FSD race at this point. But, I am curious to see more from XPENG
at this point no one is winning,
Completely agree..we've seen many perfect and many imperfect runs with Tesla FSD. We would need to see a lot more, diversified tests from this car before we can determine how good it actually is
Dear gods that’s impressive tech and the implementation of such.
Very impressive indeed.
Based on tea tech.
Limited to city use?
But only on pre mapped roads.
@@eclecticcyclist good thing we have Google Maps since 2005. Maps are not the hard part.
@@frangalarza Maps may not be hard, but keeping them up to date is.
@@eclecticcyclist I'd say updating a map once a day is comparatively easier than having a car predict a map with 100% accuracy in real time. But I may be wrong 🤷🏻♂
I don't really care what tech is used as long as the car can drive itself :D
I’m wondering if the roundabout in the video is actually a traffic circle. They’re different things and the traffic circle is quite horrible compared to a roundabout.
Doesn’t their solution require maps? So if roads change; the car needs to be made aware of it? Tesla autopilot is made to map in real time so it can drive theoretically anywhere
Definitely a smoother system then what Tesla has! Just watch how smooth the steering wheel turns!
Just watch how little turning it does in this video in general... Its basically glorified lane keeping system.
this episode will NOT age well. mark my words. "Lidar is the fools errand" - Musk 2019
It already hasn't aged well.
Er, he's not the Oracle, you know - he's just a bloke with far too much money for his own good .....!!
I like the system voice. There is no system voice on model 3 other than navigation system. And it is in English, fench or Espanion...
This video is simply wrong, and makes me think it's deliberately miseading.. it then makes me think I wonder who is paying you to make this rubbish ??
•Tesla do not use radar as you claim, currently it's only a camera based system.
•Xpeng system is currently a geomapped system, and this is what is featured in your video, this means that it can only drive autonomously on areas of the planet where a geomap has been created for its system, in this case probably Guangzhou only. The amount of work required to create geomaps is massive, and the bigger problem as you probably know, is as soon as you publish a map it's out of date the next day.. the system needs to be able to deal with the real day to day road environment variables, to be fully autonomously it needs to be self-learning, this is Tesla's method. As an engineer myself, and my son working on a geo based navigation systems, we know that Tesla's method is only system at the moment to achieve self-learning autonomous vehicles.. although some labelling is still checked by humans but this is reducing quickly and the aim is to be zero human checked.
Even Xpeng themselves have stated that their aim is to achieve system similar to Teslas which does not use geomapping. The geo mapping method is probably being used in the car you tested for ensuring safety, & used in tandem with their non geobased system. From your video evidence it seems Xpeng I have made great progress dealing with all the object variables in the driving space, and it seems to me that maybe this is the second best system available at the moment, and it's good to see that Tesla may have some real competition.
I suggest checking Teslas recient 'AI Day' video.
You promoted that having radar, lidar and cameras is better than having only cameras, because you said "Tesla only have cameras" .. with my engineers head on again.. remember what Elon says..'the best part is no part' .. he's referring to components in car.. as this reduces costs massively and probably increases reliability.
I understand that in certain situations radar can produce better object data, for instance in poor visibility, and for this reason Tesla is reconsidering radar, however the data that Tesla is receiving in such driving conditions has proven that the cameras only are orders of magnitude better than human human beings at driving safely in these such conditions.. but hey let's make it even safer by radar if its worthwhile.
Checkout utube channel AI DRIVR if you're interested in seeing how Tesla system deals with driving across San Francisco using cameras only.
Totally agree with your excellent answer
Lol, no.
As someone who has worked with LiDAR, it doesn’t work so well when someone is wearing black either.. but much better than radar, and certainly better than image processing like Tesla does.. LiDAR is super duper accurate for depth info too.. down to mm usually
Lidar is a fools errant, just like hydrogen. Lidar does not help you one iota to drive your cars better. Its all in the Software. Lidar just measures the obstructions in a road better to mach the route with the built in HD map more exactly.
Lets drive on a unmapped road. Now lets take cars with 25 or more sensors like BYD, Nio or XPeng and then take a car with 8 sensors - a Tesla. Which one will drive the unmapped road more safe? The one with lidar or the one with the better software?
@@wolfgangpreier9160 : I'd like both personally. knowing exact depth information of pixel im looking at when evaluating image data is kinda important.
Depth info is useful, it's why we have 2 eyes.
@@billybollockhead5628 I want a working self driving vehicle. I don’t care whether they use LiDAR, vision or magic. I can not imagine that XPeng comes into our little village in the middle of Europe and makes a HD map of every road. That’s why I count on Tesla. They do not use maps. And their autopilot works today.
LIDAR is great for certain things, but Tesla has a point when they said it's a crutch. It makes things _too easy_ in certain situations, but sets you back severely once you start relying more on it - for example it can struggle to work properly in adverse weather conditions (rain/snow), and you need vision anyway for all kinds of traffic-related visual context.
@@logitech4873 : How exactly is 100% imaged based technology (like tesla is planning on) going to work in heavy snow/fog?
Xpeng have done a great job of stealing Tesla's FSD code and using it as a base. So much easier, quicker and cheaper than doing it all yourself.
Tesla doesn't use Radar anymore. They're also phasing out Ultrasonic as well. It will be Vision (Camera) only in the near future and going forward, at least for time being.
that's why they will LOSE. 🤣🤣