Tesla VS Waymo - Who Will Win the Race to Full Self Driving? + LiDAR VS Computer Vision

In the current

race

towards our autonomous

driving

future, who

will

be the first company to offer a

full

autonomous

driving

function. I'm Johnathan Stewart and welcome to a cleaner. I recently read an article on autonomous driving and this article had a graphic from Navigant's research. his leaderboard and showed which companies he thought were leading in autonomous driving and which were lagging behind, interestingly enough they had Tesla behind, they had a lot more GM and Ford in the lead, so I took a few minutes and dove in in the data. for each of these companies and I found out very quickly that really only way mo had some serious data some serious autonomous miles and really promised autonomous driving of course knowing what I know about Tesla and their autopilot system and how it all works that.

I know they are leaders too, so in this video I want to take a moment and talk about the top two leaders in the

race

to autonomous driving. and mo, Tesla and I want to focus on a few key metrics technology differences and determine based on this data who is likely to be the first to

full

y drive themselves, so the first category I think is important to talk about is the set of sensors. Do you use each of these companies to be able to get to full autonomous driving so that the car can function on its own in its environment?

One of the major sensor differences in the system that way you use Mo versus Tesla is Tesla's

computer

vision

versus a Relient

lidar

system like grind only if you don't know it's simply a light and range sensing system that does bouncing lasers off objects at millions of pulses per second and then the car measures changes in distance and the laser pulses bounce back to give these readings instead a camera-based system uses

computer

learning to train the neural network, which which allows the car to recognize elements in the environment and react accordingly. Here's a clip of Elon Musk talking about

lidar

vs. cameras on autonomy day earlier this year everybody's going to throw away lidar was my prediction walk my words i should point out i don't actually hate light very much or as much as It may seem like it but in SpaceX their basic dragon uses lidar to navigate to the space station or dock not only that SpaceX developed their own lidar from scratch to do that and I personally spearheaded that effort because in that scenario lidar makes sense and size is fucking stupid is expensive and unnecessary and as laurie said once you wear the

vision

it's kinda useless so you have expensive hardware worthless on the car we have front radar which is inexpensive and useful especially for occlusion situations , so if it's foggy or dusty or if you know snow the radar can see through that if you're going to use active photon generation don't use the visible wavelength because once you have passive optics you go you've dealt with all the visible wavelength and stuff you're wondering if you want to use a wavelength which is the occlusion penetrating like radar so the tread lining is just active generation of photons in the visual spectrum if you're going to do active photon generation do it outside of the visual spectrum on radars in the radar spectrum like twenty point eight millimeters versus 400 to 700 nanometers that's going to be much better occlusion penetration and that's why we have radar Ford and then we also have alpha 12 ultrasound for near field information in addition to the eight cameras and the Ford radar the youngsters need the radar in all four directions because that's the only direction you're going really fast so I mean we've weathered these mobile times as usual we are sure we have the right sensor suite, if we add anything else know that the advantages of a camera based system are of course that is very cheap to implement cameras in a car it can be trained through computer learning to recognize objects and so you could train this system to recognize objects drop the car somewhere or never been before and you could still orient your

self

a system based on a camera allows the car to read the signs our current road system is set up for human vision of course and therefore a system that is based on human vision of course a camera which is computer vision allows you to read signals and react accordingly just like a human would and of course the cameras allow for a very sleek built in design which doesn't take away from the aesthetics of the car the downsides of the camera based system are just that takes a lot more computer learning and data entry work, and this time and computing power is very intensive, so now that we know the difference between lidar based system and camera based system, i want to go over the sensors in that way, no. they currently use in their cars to navigate their environments of course as we mentioned before they use lidar sensors on the top front and rear of the car they also have a vision system so you have a camera that helps in th the surroundings and the driving the car they have a radar system in both the front and rear and then they have add on sensors which they describe as their audio detection system that can listen to police sirens and emergency vehicles and they also add a GPS for Know where you are based on a map Tesla's Autopilot sensor suite includes eight cameras, twelve ultrasonic sensors and a forward-facing radar system.

Ultrasonic sensors are simply used to measure distance by using ultrasonic waves. It lets you know how close you are to an object. Of course, they also use cameras and radar to get a glimpse of the environment around you. The second category I want to talk about is focus. How does each of these companies approach the problem of solving fully autonomous driving? We currently know. wham-o is using a lidar based system and currently they have to map the whole section that way that Moe can drive, they have to traverse and do 3D mapping for an area and then they have what they call their geofence which means you can only operate wham-o in full autonomous driving features on this geo-fenced map and here's a picture from way Mo's website and it shows what their current

self

-driving turf is in the greater Phoenix area and to what they hope to expand in the future and so Mo's focus is to use lidar and geofencing and map specific areas in the city and I

will

allow their cars to be used in those areas Tesla on the other hand is using computer learning and vision by computer, which means that once its role is complete with fully autonomous Tesla vehicles, they will be able to drive autonomously almost anywhere in the world, geofencing will not be required.

What about the cost of these vehicles, of course the cost will be important. as they implement these features into cars whether that means a Robo taxi network or they decide to sell these vehicles to the public the cost will be great currently retail price of the MO c The hrysler pacifica minivan they use to become in his way mo taxi it has a retail price of around $40,000 it's a hybrid and if you look at that in the pacific a 2019 hybrid is around thirty nine thousand dollars then of course you realize there's a number of lidar sensors and you start to add about $7,500 for each of them plus the computer system and all the rest of the technology that they put in the car and a very conservative estimate would be that it would be about a hundred thousand dollars for this car but I suppose that in reality it is much more but a minimum of one hundred thousand dollars for each of these taxis that they put on the road Tesla on the other hand has a model the model three sr+ that has a retail price of thirty nine thousand nine hundred and ninety dollars so the same base price of the chrysler pacifica hybrid minivan that way uses mo but the

tesla

car already includes all the chips all the sensors and everything you need to a full autonomous drive, also layer in the fact that the current leases for the model 3 don't allow you to buy the car because Tesla has mentioned that they're going to turn these cars into their Robo taxi network so essentially they'll get these deeply discounted cars where the customer has already taken much of the depreciation on that vehicle, so the actual cost to Tesla will be much lower and this is not even a factor than the fact that Tesla has a margin of between 15 and 20 percent on their cars is also another important factor to compare between the two companies is the current size of the fleet how many cars each of these companies have in real world situations well based on my research wham-o currently has 600 autonomous vehicles on the road they have an order for 62,000 plus chrysler pacifica z' and they have also placed an order for 20,000 jaguar i paces.

Tesla, on the other hand, has over 600,000 cars on the road right now with hardware version 2 and higher that are capable of self-driving once. Tesla rolls out those features via a software update on Lex Friedman's website. He put up a nice graph that he created and you can see this. I'll put a link in the description, but he estimated that there are currently 625,500. and seventy autopilot cars to hardware plus and that was as of Q3 2019 so by the end of Q4 2019 of course there will be about a hundred thousand more cars and on top of that we will break the mark of 700k by the end of 2019 now the next important thing to look at is the data collected by this fleet so of course as we mentioned Wham-o has six hundred vehicles Tesla has over six hundred thousand and really close to 700k vehicles in the road by collecting data and giving them feedback, so with all the information Moe has recently posted that they have driven over ten million miles on real world roads autonomously since 2009 and while that sounds impressive, especially when you Compares to Less Freeman also estimates that the rest of the Tesla competition will hit two billion miles driven on autopilot with full-featured driving at autonomous included for January 2020, so when you look at this for a minute, 10 million sounds impressive, but then you realize that Tesla, by the end of this year, will have data on over two billion miles driven with its hardware systems.

This is an incredible amount of data and is a huge benefit to Tesla engineers who are working on computer learning and computer vision. Tesla has recently launched. features like navigate on autopilot that allows the Tesla vehicle on autopilot on the highway to change lanes and also take exits by itself they also launched features like smart summons that allows the vehicle to drive itself out of a parking lot and come to meet you at the curb or come pick you up wherever you are and it's a pretty amazing feature that is a step towards full autonomous driving so in the race towards autonomous driving ving data is king as you know the world real world is very unpredictable people humans do very irrational things and don't always follow traffic laws and laws that are around on signs so because of this most real world data you can be collected, the better, this allows you to anticipate and solve the equations that are needed for all the corner cases and so you can solve them and make sure that the auto s react accordingly as they should Tesla has all the real world data for their global fleet if you think about that for a minute they have a global fleet they have data not just for the United States, Canada, Europe, the whole world, they have this data they're currently building, billions of miles of data, up to this 10 million miles of data. wow that's a lot of data just for specific areas that have been mapped they have data for the greater Phoenix area mainly Tesla is also working on solving computer vision and has hundreds of thousands of their owners helping them with t your data wham-o have to pay your engineers to drive all these miles and map and geo-fence these areas and that's very expensive and time consuming so based on all these data points we've discussed who is going to win the race, well I think it becomes very obvious when you think about the data, when you think about the features and you think about the number of cars on the road with the necessary hardware to be fully autonomous, I think it becomes pretty obvious

tesla

is poised to get there faster than any other company now wham-o is operating at level 4 according to the sae they have a list of what they consider autonomous driving they have n Levels 0 to 5 currently say that Tesla is operating on Level 2 officially and wham-o is operating on Level 4, so at first glanceI would say ok of course Mo is in the lead they are doing level 4 Tesla is only at level 2 but this level 4 is in geo-fenced areas only they have been very intensively mapped and they are using allied expensive resensors to put in these cars to do this once this is all done the cost of a fleet and the cost of preparing the maps for these areas is very substantial Tesla on the other hand is tier 2 but once they figure out computer vision they are good to go all they have to do is just flip the switch and the whole fleet is ready to be autonomous their biggest problem right now to get to full autonomous driving is the millions of cases of corner that exist in the real world just like us.

As mentioned before, the real world is unpredictable and Tesla currently has to train its neural network to deal with all these corner cases and in my opinion Tesla is likely to figure out full autonomous driving and have a whole system of features that can not be approved for use yet but they'll have a full featured system by the end of 2020 and by that time they'll have over a million cars ready out of the box so as I mentioned I think Tesla is ready to go a long way faster than anyone. I think they will be fully featured by the end of 2020 I think regulation and stuff like that will take quite a while to officially allow it but maybe by the end of 2021 2022 we'll have the regulators approve it and it might be ready to just go to sleep and let the car take you around the city companies like way mo and others that use lidar systems i think we are still five years to a decade away from achieving this thank you very much for watching this video i hope you learned a few things and i hope you enjoyed reviewing the data comparing these two companies, if you like the video please consider subscribing so you can see future content coming out in the future, also if you liked it please go ahead and click the like button it helps get the video seen for more people, thank you very much.