字幕列表 影片播放 列印英文字幕 tesla is very adamant on their statement of not using lidar for their vehicles to accomplish full self driving why is everyone else using lidar is lidar good is it bad let's discuss that in this video different companies are taking different approaches to full self driving no one has actually reached level 5 full self driving yet so really it's just anyone's game at this point but right now as it stands tesla seems to be the one in the lead if you haven't seen any of the videos that owners are actually posting on youtube twitter or other social media platforms about the fsd beta the rewrite that elon said would be a quantum leap then you are greatly missing out those videos that owners are posting show how well the car actually handles most of the situations now you can argue and say well waymo already has their full self-driving system active in a city in arizona and to that i would say yes it's working but there are big issues with lidar now what is lighter in the first place lidar stands for light detection and ranging and what does that basically mean well there's a spinny thingy i'm just gonna call it a spinny thingy because that's that's literally what it is and in that spinny thingy it shoots out lasers or light beams and what it does is it waits for the light beam to return back and calculate how far that item is or object is from its location and this can be very useful this is basically what radar does except using light waves visible light spectrum and this is not necessarily a bad thing the iphone 12 hasn't they're using it for a completely different purpose rather than full self driving and obviously they're using something that is much less precise whereas the lidar used for cars is millimeter precision it can measure something up to a certain distance with millimeter accuracy the iphone does not do that i don't even know if it does centimeter precision but that is highly possible and i think reports have also came showing that the iphone only does it up to about 16 feet so not that far that'd be like a car driving about three to four miles an hour and i've seen this argument made well why can't cars use lidar if the iphone has it because of this reason i'm pretty sure you don't want to be in a car that only goes three miles an hour because it can only see 16 feet in front of it so why is tesla not using lidar so tesla is very engineering based they want to simplify things as much as possible and everyone working there has a sort of engineering mindset thinking from first principles and there are some issues with lidar that present themselves before you even get started number one is cost and here's what i think about cost cost can drop dramatically over time before it costed about seventy eighty thousand dollars to just buy the equipment for lidar let alone actually installing it on a vehicle and that is very expensive that's just the entire cost of the car itself but now cost of lighter is dropping i think it's about 25 000 and so cost is something we can get over if it proves to be a valuable asset now the second reason why tesla doesn't use lighter is because it's quite unreliable i mean think about this you have physical cameras that don't move don't do anything like the ones you have in your phone pretty much and what they're doing is they're just there like they just work you could take a 10 year old phone and open up the camera and obviously maybe the image quality is not that good it's all for processing but as long as the phone works the camera should also work as long as everything works properly and that's exactly the benefit using cameras you have cameras that can detect everything and obviously lidar isn't a replacement for cameras it's just an add-on but cameras can do the same thing that lidar can and lidar is a spinny thing it's literally something that is spinning outside of the vehicle which means it's more prone to damage so because of that reason lidar is not really practical because first of all if it breaks it's quite expensive to replace secondly even if it does not break it's just something that is waiting to be broken and that's quite useless and also ruins the design of the car i mean you look at any tesla out today especially any model 3 or model y which all have the sensor suite to make full self driving viable on those vehicles you look at those cars and they're pretty much like every other car on the road well maybe they look nicer than every other car on the road that's subjective but they don't have some next external factor that distinguishes them from other vehicles on the road necessarily for example if you look at waymo or google self-driving cars first of all google self-driving car it looks like an orange it looks like a ball with something on top of it like it's weird what then you take waymo for example waymo uses a chrysler pacifica which is a pretty nice minivan if i have to say myself but you look at the sensor suite and you can tell that car is special it has some self-driving equipment and whatnot whereas tesla does not do any of that it's all just built into the vehicle and with lidar you cannot get that so it's just something to think about now the third reason why tesla doesn't use lidar is because it's redundant there's no need to use lidar because lidar does not work in most of the situations where you want something to actually work and help the cameras lidar does not work in fog it does not work in snow it does not work in rain does not work inhale it basically doesn't work in most of the natural climates we have here on earth and because of that limitation for lidar what you need is a beast of a vision system and that's what elon has been referring to lidar is useless in that sense because during those times lidar won't be used and if you try to use it it can't see anything it's like driving blind essentially even in a light rainstorm and so what you need with lidar is a really good vision system which you need even though you're not using lidar so really lidar just becomes a secondary option for something that is less relevant and actually even less useful because it doesn't even work most of the times when you need some sort of help now radar is different radar uses radio waves to determine the distance between the vehicle and front and what radio waves do is they can see through rain fog hail snow just anything pretty clearly if there's any rainfall it can see other cars pretty clearly and so that's why they use radar for adaptive cruise control to keep you a safe distance away from the car in front and radar is actually useful it's not that expensive it doesn't break easily there's no real moving part and it actually aids the cameras it can actually detect distance with pretty good accuracy and see through most of the natural climate we have here on earth so radar makes total sense lidar not so much even though it's trying to compensate for radar if that makes any sense so why is every other company using lidar well for one no one has actually reached full self driving yet so we don't really know what's the right method it could be that tesla's method is completely wrong and they just have to start from scratch and add lidar and whatever you know mission failed full self-driving capabilities not coming to any of the cars that elon promised it would tesla would probably refund their money at that point but i see that as a highly unlikely scenario i mean think about it humans drive with two cameras their eyes and this is what elon argues all the time we have two cameras on a pretty bad gimbal system our neck or head and we drive pretty badly but pretty well for how we should be driving i mean with two cameras only looking straight with a little bit of peripheral vision we should be driving much worse than we are right now so in that sense you can say that cameras do a really good job and now tesla has the 360 degree cameras looking all around the car with a computer that reacts in milliseconds not at the second time scale your brain reacts at the second time scale if something happens it takes a fraction of a second for your brain to compute and perform that task whereas for a computer it takes milliseconds it's a thousand times quicker than you are and here's going to be my final point about lidar what everyone else is doing other than tesla and comma ai is using lidar and mapping out the roads so what you do with mapping is you have a vehicle or many vehicles that actually map out the road beforehand measure exactly to millimeter accuracy where everything is and then the car would drive kind of like in a video game but real life isn't a video game random things happen all the time for example if there is construction on a mapped road the car is likely to have a much harder time to comprehend how to navigate that situation versus a car that relies on vision and that works like your brain when you arrive at a situation when there's construction or pylons or really anything you know how to navigate that situation maybe not well because the humans you know we create traffic jams but we know how to navigate those situations we're not just completely stuck we don't just turn off our cars we're like oh we're done that's it we just have to wait for this construction to go we're not like that tesla's system will be exactly the same it'll work more like a human brain than a computer and this can be seen by all those fst beta videos the owners are putting out it behaves a lot like a human it waits it navigates roads and navigates around other cars other objects around people much more like a human and it drives much more like a human rather than what you would assume a computer would drive like so when the waymo vehicle approaches a construction zone in its path it has to essentially rely on vision and just rely on everything that whoever coded the system taught it and what to do in that situation whereas tesla system it works like that it's literally intended for it doesn't even have to think about it it's almost like it doesn't even care that there is construction its main purpose is to navigate the system and the tesla vehicle is just gonna go through it like it's nothing whereas waymo might get stuck and this is exactly why waymo looks to be ahead when it actually isn't right now it's only working in a city in arizona on certain streets try taking that exact system just take it outside of the city where their roads aren't mapped and it would essentially fail not saying would necessarily fail entirely but it won't be able to do as good of a job or even close to as good of a job or remain as safe as it was in that playground as i call it what you have to do is take the car out of the playground once the car is out of the playground you can determine if the car can navigate certain situations that humans can navigate without even thinking about you have to design a full self-driving car like a human and lidar just sort of restricts you like that we don't have any sort of system that determines distance we just have two eyeballs and we just know okay this is how far we are from the other cars and so far we should remain and we do a relatively good job sure some people get killed on accidents but based on what computers have taught us there should be a lot more accidents so we're a lot more safer than we were intended to be with driving so in the end it looks like tesla is actually going to be first to achieve complete full self driving just because of what they're doing and it also seems like lidar may not be the right approach i think england was right the entire time that lidar is just not the right approach but we have yet to see real full self driving with no human supervision and whenever that happens times can change maybe waymo will decide to drop lidar entirely tell me what your thoughts on this in the comments down below and i'll talk to you there thank you for watching hope you enjoyed this one and i'll talk to you later peace you
A2 初級 美國腔 为什么TESLA不使用LiDAR?(Why TESLA Isn’t Using LiDAR?) 51 3 joey joey 發佈於 2021 年 05 月 24 日 更多分享 分享 收藏 回報 影片單字