• 10 months ago
Mechanical engineer and Stanford professor Chris Gerdes answers the internet's automotive questions. Why aren't solar powered cars a thing yet? How do self-driving cars see? How can Teslas crash when on auto-pilot? Is hacking a big threat to autonomous cars? Chris answers all these questions and much more!Director: Lisandro Perez-ReyDirector of Photography: Florian PilslEditor: Jordan CaligExpert: Chris GerdesLine Producer: Joseph BuscemiAssociate Producer: Paul GulyasProduction Manager: D. Eric MartinezProduction Coordinator: Fernando DavilaCasting Producer: Nicholas SawyerCamera Operator: David SmithSound Mixer: Simon GordonProduction Assistant: Nathan SandovalPost Production Supervisor: Alexa DeutschPost Production Coordinator: Ian BryantSupervising Editor: Doug LarsenAdditional Editor: Paul TaelAssistant Editor: Andy Morell
Transcript
00:00 I'm Chris Gerdes, professor emeritus of mechanical engineering at Stanford University.
00:03 Let's answer some questions from the Internet. This is Automotive Support.
00:07 @DonkeyDelph asks, "Should I buy an EV now or wait a couple of years while they're still
00:16 improving technology and prices come down?" One thing you may want to think about is Ford
00:20 has recently announced that they're going to use Tesla's charging system on their future
00:25 electric vehicles. Tesla built the supercharger network specifically for their own cars,
00:30 but other manufacturers have decided that that's actually a really good engineering option. And so
00:35 they are engineering in the Tesla charging system into their cars. So if what's out there today
00:40 meets your needs, go ahead and buy it. But if not, you probably don't have long to wait until
00:44 something comes out that does. @Punicist wonders, "I wonder why solar-powered cars aren't a thing
00:51 yet?" So if I take the Mercedes behind me, that's a really large car. I could fit maybe
00:56 four solar panels. Each of those under peak conditions would produce about 400 watts.
01:01 Put it together and I have 1600 watts, almost enough to power this hair dryer,
01:08 somewhere on the order of about two horsepower. All right, but maybe I could use this as an
01:12 alternative to charging. That same amount of power would actually be less effective than
01:18 plugging your electric car into a standard 120 volt wall outlet. It would really take forever
01:24 to charge. So if you want a solar-powered car, simply put solar panels on your roof,
01:29 generate electricity and use it to charge your electric car. @AllenBuxley asks, "Why isn't there
01:35 a huge market to retrofit most popular petrol cars with battery packs and electric motors?"
01:40 So you can take a petrol car and turn it into an electric vehicle. My team and I did that with this
01:46 DeLorean. In order to fit everything in, we actually had to design and fabricate our own
01:51 subframe for the car to hold these new components. But this is not the easiest way to go about
01:56 getting an electric car. You're going to spend a lot more time and money than you would simply
02:00 buying an electric vehicle. The challenge is that your car was designed to have an engine
02:05 and a gas tank. If you start to pull in electric vehicle components, there's a lot of things you
02:11 have to ask, like how safe is that going to be in a crash? What have you done to the weight
02:16 distribution of the car? So this is why you don't really see many electric vehicle retrofits out
02:21 there. @TheAlteriorKid writes, "How do self-driving cars see?" Self-driving cars can use a range of
02:27 different sensors. Radars can be very helpful sensing metallic objects that are near the car,
02:32 but sometimes it's hard to tell exactly what that radar is bouncing off of. Cameras give a picture
02:36 of the world that's not unlike what we see with our eyes, but it doesn't really give you distance
02:41 information. Finally, LiDAR or laser scanners. This is a series of beams of light that tells
02:47 very precisely how far away an object is from the LiDAR. You'll see that the LiDAR is spinning on
02:53 top of our X1 vehicle behind me right now, and it's painting a picture of me sitting here at
02:59 this table. So if I move my arms, you can see the image move. Each one of these lines represents one
03:05 of the spinning lasers. The color tells how intense of a reflection the laser is getting.
03:11 So autonomous cars see by using a variety of sensors and AI to turn all of these sensor signals
03:18 into a picture of the environment. @Kindman72672657 asks, "Why do autonomous vehicles have
03:26 steering wheels? Our federal motor vehicle safety standards are generally interpreted to mean that
03:31 all cars have to have driver controls. It's easier for manufacturers to simply leave the
03:36 steering wheels and driver controls in the cars. Engineers like to talk about levels of automation.
03:42 A level one vehicle is a system with an adaptive cruise control, something that uses a radar to
03:48 follow another vehicle safely, or a lane keeping system. A level two combines these two systems,
03:54 but the driver needs to remain with their hands on the wheel. A level three autonomous vehicle,
04:00 that's this Mercedes EQS here, that car will actually drive itself in limited conditions
04:05 in traffic jams. The driver can take their hands off the wheel and their eyes off the road. A level
04:11 four automated vehicle is a vehicle that can drive itself within a specified area. These driverless
04:17 robo taxis that you see companies like Waymo or Cruise. Level five autonomous vehicle is really
04:22 the highest level. That's a car that can drive from one point to any other point that you want.
04:29 That car obviously won't need a steering wheel, but that's a long ways off in the future.
04:32 @Coochariat wonders, how will flying cars operate? There are several manufacturers out there
04:39 looking at very light aircraft that could take off and land vertically and might travel
04:44 electrically over 50 to 60 miles, looking a lot like bigger versions of the drones you may be
04:51 flying. These aircraft could in fact operate autonomously. In some ways, that's an easier
04:56 problem than trying to make a robo taxi. You don't have to worry about pedestrians or many
05:01 other vehicles if you're flying fairly low over the ground. @Heyzel asks, can anyone explain to
05:08 me why Miatas are cool now? It's a f***ing Miata. They have a 50/50 weight distribution. When you
05:14 turn your steering wheel, what you're trying to control is the yaw rate of the car or how fast
05:19 it's turning. In many cars, what you'll see is a yaw rate response that tends to oscillate. The
05:24 car kind of wiggles as it's getting into its turning rate. But with a 50/50 weight balance,
05:30 like the Miata has, the car has a neutral steering characteristic. When you turn your
05:34 steering wheel, you get a nice smooth response. So if you want to learn to race, you're going to
05:39 find that this neutral steering characteristic in the Miata is the perfect way to learn racing.
05:45 @Ywen8698 writes, drove down I-40 earlier and hit a long and indistinguishable object
05:52 falling from the car in front of me. My car drifted at high speed for about five seconds.
05:56 Wonder how autonomous vehicles will react to this? Autonomous vehicles need to react to
06:02 unexpected situations on the road with skills that are as good as the very best human drivers.
06:08 That was the reasoning behind building our autonomous drifting DeLorean, Marty.
06:12 Marty is able to drift intentionally, putting the car sideways so that it's very,
06:18 very controllable and it can track exactly the path that it wants, even in emergency situations.
06:24 @Therizofan420 writes, you built a drift machine out of a DeLorean? Yes. So in the hands of an
06:30 inexperienced driver, you're likely to over-rotate and spin out, but a skilled drifter can keep the
06:36 car sideways and maneuver the car with pinpoint accuracy. This is what we were trying to do
06:41 with our software. We've now largely achieved the ability to control the vehicle's path within
06:47 a few centimeters while sideways. @Ween0 asks, why were cars back then more boxy and more curvy now?
06:56 Older cars were certainly more boxy. They came from carriages. In the early days,
07:00 would often be built upon a frame and somebody else would build a body that went on top of it.
07:05 At that time, there wasn't a lot of attention paid to aerodynamics or how lightweight a car body
07:10 could be. In fact, if you went a little bit further, there still wasn't a lot of attention
07:14 paid to aerodynamics, but more towards tail fins and what the car looked like. The size of the tail
07:19 fins became an important marketing feature with new cars. So in the 1970s, there was a severe
07:25 energy crisis and that made manufacturers want to develop cars that were lighter weight and more
07:29 fuel efficient. The way to make them lighter weight was to use unibody construction, design the car
07:34 really as one unit. The way to save fuel was to make cars more aerodynamic. And that really changed
07:42 the shape to the sleeker shapes that we associate with modern cars. In the future, one area that
07:47 manufacturers are really looking at is removing the side mirrors, which create a fair amount of
07:53 drag and replacing these with cameras. @KubekGeorge asks, why are vehicles from the 70s outlasting
08:00 vehicles from 2022? Cars in the 1970s had very few computers. With a reasonable set of tools,
08:07 you could maintain that car then and continue to maintain that car today. The price that we
08:11 pay for cars that are more fuel efficient and more environmentally friendly is that they've become
08:17 more computerized. Cars increasingly have over the air software updates. Will you be able to
08:23 continually update your software in the future? It's become impossible to do all of the work
08:28 yourself on a modern vehicle. @planetaryjim asks, why you put spoilers on race cars? To make sure
08:36 that they can grip the ground better. So if you think about an airplane, they generate so much
08:40 lift that they can overcome the force of gravity and fly through the air. We can put something like
08:45 an upside down airplane wing on the car, which is a spoiler. You can generate so much downforce
08:51 with aerodynamics that you can overcome gravity. Imagine that I create a Formula One track upside
08:56 down. If I get my car to run that way, my downforce becomes upforce. It could actually stay
09:02 stuck to this upside down track. Some people just attach something to the back of their Honda Civic
09:08 to look cool. That's not an engineering thing, more of a styling choice. @lumenrio1 asks, so
09:14 you're telling me that if someone puts up a sign with the image of a green traffic light,
09:18 it will confuse autonomous cars? Maybe. If it's using cameras to try to make sense of its current
09:24 environment, it may very well be looking for traffic signals. If it finds one on the back of
09:28 a semi and not actually signaling an intersection, it could get confused by that. One of the difficult
09:33 things about programming autonomous cars is you have to think in advance about things that might
09:38 confuse them. News reports from San Francisco show that people have been able to confuse Waymo cars
09:44 by putting a cone on the hood of the car. That's something I doubt any engineer would have thought
09:48 about when they were designing the automated vehicle. @litterpickerpro1 wonders, will cars
09:52 of the future have wooden wheels? Can't get much greener than that. Are we going that direction
09:57 in the future? No. But on the other hand, you may have tires that are coming from shrubbery.
10:03 Bridgestone has been working with Y-Ule, which is a shrub that produces hypoallergenic latex.
10:10 Tires could be biodegradable. We've paid a lot of attention to pollution from the tailpipe of cars.
10:16 As tires wear out, they produce little bits of rubber and petroleum products that can actually
10:23 cause health issues. @mrxavierm writes, get at me. Ferrari versus Porsche. Both of these marks have
10:30 tremendous racing history. On the Ferrari side, you've got that Italian styling, but I'm going to
10:36 have to actually come down on the side of the German precision. If you want handling designed
10:42 precisely to go where you're commanding with your steering and your throttle, it's hard to beat a
10:47 Porsche. @realgreenboy0 asks, why does people get in crash when they have on Tesla autopilot? So
10:54 psychologists know that if we don't have enough to do, our mind wanders, and it's hard to keep
10:59 our attention on being ready to take control if we need to. Tesla's autopilot are level two
11:06 automation systems. They require the human driver to be engaged and in control all the time. This is
11:11 a challenge that the whole industry faces, and they're working on different solutions to warn
11:17 or re-engage the driver. @emvdn, self-driving cars running with AI present a new version of
11:23 the trolley problem. How does the car make an ethical decision about who to kill by running
11:28 over? So in one classic version of the trolley car problem, you're a bystander and you see a
11:32 trolley headed towards five people on the track. You can switch to another track where only one
11:37 is killed. Do you pull that switch? Automated vehicles have to consider situations where there
11:42 may be loss of life, but they're not really using this sort of moral calculus. Part of my work with
11:47 a safety advisor for Ford, I actually helped develop a set of rules that autonomous vehicles
11:53 could use. And this just simply involves following the duty of care that the law outlines. Now, if
11:58 other people violate their duty of care to the car by jumping in front of it when it doesn't
12:03 have enough time to avoid a crash, you can program the autonomous vehicle to do the best it can, to
12:09 do everything under its power to avoid that crash. Autonomous vehicles are out there protecting people
12:15 who follow the rules and not making determinations about who's more worthy to live or die. @theaoti
12:22 asks, "If my self-driving car gets into an accident, why wouldn't whomever made the car be at fault?
12:27 I wasn't driving." If the car is driving itself, manufacturers will in general be liable for that.
12:35 Mercedes has already made a very strong statement with their level three drive pilot that if it has
12:40 a crash, the company is responsible and not the human driver. So we're shifting really into an
12:45 entirely different world. In fact, in San Francisco, they've already run into the problem of
12:50 how does a police officer give a ticket to an autonomous vehicle? That doesn't make sense when
12:55 there's not a driver. @mbj2042, "Why would I buy a hydrogen car when I could charge my electric
13:02 vehicle at home?" Think for a moment about over the road heavy trucks. In order to recharge them,
13:08 they would actually need vast amounts of electrical charging capability. If they wanted to do that in
13:13 a reasonable period of time, they'd also have to carry massive amounts of batteries. Hydrogen may
13:20 actually make sense for buses, over the road heavy trucks, probably not going to replace your electric
13:27 vehicle and your charger at home. @digitalbladeca wonders, "If you had to solve real world driving,
13:32 would you go down the path of programming the rules of the road and expect the car to obey,
13:37 or let a neural network learn to drive like an experienced Uber driver?" You really want a
13:42 combination of these techniques. So Tesla, they use neural networks to learn from their camera
13:49 system and what the cars are likely to do around them. In contrast, Mercedes level three system
13:55 tends to rely on rules. Engineers can program specific behaviors into the car to ensure that
14:01 it follows the traffic laws or follows a vehicle at an appropriate distance. @chumicer asks,
14:07 "I'm really concerned about security when autonomous cars take over. What if they're
14:11 hacked?" Hacking is a real concern with automated vehicles, particularly those systems that can
14:15 receive over the air software updates. There's also a threat as we move towards AI systems that
14:21 hackers could get involved with the data stream. AI in an autonomous vehicle is trained by feeding
14:28 it lots of data, which could be, for instance, a series of camera images. If you could hack that
14:33 set of images, you could make it so that the vehicle failed to recognize certain objects in
14:38 the environment, and that could create some serious safety risks. @raverbrogan asks, "Why do
14:43 cars have four wheels?" Because we found out that is a really great combination to design vehicles
14:50 that are stable, comfortable, and high performing. Motorcycles have two wheels, and it's up to the
14:56 rider to balance. With a car, the suspension handles all of the balance. As my car goes around
15:01 a corner, my tire forces increase on the outside and decrease on the inside to keep the car
15:08 upright. If you have a three-wheel car, you actually have only two of those wheels helping
15:13 to keep the car upright. And what this means is that the weight can shift very dramatically from
15:17 one wheel to the other. These cars have a much greater tendency to roll over. @mountaindewmlg
15:23 asks, "Why aren't we all driving electric cars by now?" It's not a technology that makes sense
15:28 for everybody. Range remains a limitation. If you want to take a long road trip, it is going to take
15:33 you some time to stop and recharge your car. More time than it would take you to fill up a gasoline
15:38 tank. So for many people, an electric vehicle is cheaper because they can lug it in at their home
15:43 and use their own electricity. But if you have to charge your vehicle all the time at a commercial
15:48 charging facility, you're likely not going to save that much money. @dubifuture asks, "Will
15:54 autonomous vehicles reduce the number of accidents?" So autonomous vehicles will certainly eliminate
15:59 some forms of human error. Autonomous vehicles don't text while driving. They don't drive
16:05 impaired. They're never distracted because they can see 360 degrees around the vehicle. Autonomous
16:11 vehicles won't eliminate all errors, but they will shift from the human driving errors to the
16:16 programming errors. Over time, that should be a shift towards safer cars. So those are all the
16:20 questions for today. Thanks for watching Automotive Support.

Recommended