Robot, You Can Drive My Car

episodes about them. We recommend listening to them there in our audio archives. But but there are a lot of other things that we still wanted to talk about, and also some interesting new information that has arisen. Yeah, And part of that was just kind of talking about the history of self driving cars and developing them, because we

didn't really talk about in those previous episodes. And you know, there was some work that was done as early as the nineteen seventies in trying to create self driving or autonomy as cars, and some of that technology has made its way into vehicles. Things like cruise control is a very simple manifestation, but we also have things like collision detection, parking assistance, all this other kind of technology that has been incorporated into various models of cars over the last

few years. And we're seeing, you know, year over year, cars are getting more and more of these autonomous features built into them. We haven't quite reached the point where you can go and buy a truly autonomous car off the lot and have it drive you home. But we're getting there. Uh, the question is how fast are we going to get there? But we'll kind of address that

a little bit later. To look at the true kind of history of the modern self driving car, you gotta look at UM a mysterious organization and part of the Department of Defense. Yeah, so, uh, Department of Defense. Yeah, United States Department of Defense has the Defense Advanced Research Projects Agency or DARPA. DARPA is actually responsible for a lot of technology that we use today. So they're building transformers. Is that it? Actually here's the funny thing. Here's the

funny thing. They are working on humanoid robots right now. That's the latest challenge. But the the challenges we're talking about our autonomous cars that were called robots, but don't transform into humanoid robots. They are not in disguise. They're just robots in car form. So so they're like half of a transformer, Yes, the transformer, the boring half, and and a lot of this is hypothetically for say, disaster

relief work, stuff like that, um military purposes. Specifically, it says they like they like talking about the disaster relief for ye. Well that's the that's the cheery way of

putting it. Yeah, because they The specific mission of DARPA, according to the website is darpast mission is to maintain the technological superiority of the U. S Military and prevent technological surprise from harming our national security by sponsoring revolutionary, high payoff research, bridging the gap between fundamental discoveries and their military use. Not quite as fuzzy and warm as they We want to to improve disaster relief response. But

I mean it's honest. So, but darp has been been responsible for technology that has given us the opportunity to have a job because it was DARPA that funded the ARPA net, which is the predecessor to the Internet. So do you guys know when, uh, when DARPA got started,

Like what what event precipitated the founding of DARPA. Can you can you take a wild guess of what event in history would have inspired the United States to launch a research and development organization to try and maintain technological superiority. I'm gonna say either, um, either sput Nick or when they canceled Happy Days. You were right the first times Sputnik. Uh. Sput Nick was in fact the event that inspired the

United States to form DARPA. So, of course, Sputnik was when the Soviet Union launched a saddle that orbited the Earth and went beep a few times. Terrifying. Yeah, No, Sputnik two is what was terrifying, because that had the dog in it, like poor La who suffocated pretty quickly

in that because it failures in the Sputnik two. So darnia. Yeah, getting back to the cheerfulness, DARPA ended up funding lots of different projects, including what would become the Internet, or at least the predecessor to the Internet, I should be clear. And then also in two thousand four decided to get into the automated car business. In a way, they actually submitted a challenge to the world, or really I should

say to the United States. Um, And they said that the purpose of the DARPA Grand Challenge two thousand four, which was the first year they held. It is to leverage American ingenuity to accelerate the development of autonomous vehicle technologies that can be applied to military requirements. UH, So we always get a military bit at the end. So UH. In that challenge, they had I think twenty five groups. Uh. And the challenge specifically involved a route from Los Angeles

to Las Vegas, which is three hundred miles. Is a three hundred mile course that these cars were expected to UH to go on, and it included both on road and off road sections. They had cleared the on road sections of traffic so these robotic cars wouldn't be careening into pedestrians and and and regular traffic. It wasn't onlike Hollywood Boulevard or anything like that. The best probably yeah, no, I've been on Hollywood Boulevard. That is for the best.

A series of way points to find the routes, and the vehicles were to follow those way points, and there were also some checkpoints where the teams were allowed to stand by in order to do quick repairs or refuel their vehicles before the vehicle would continue on. However, every single vehicle had to be unmanned and autonomous, which meant

you could not control it by remote control. It had to be moving on its own and the winning team would be the first one or the one to navigate that route in the shortest amount of time and they would win one million dollars. So among those twenty five teams were some racing teams UH. There was also cal Tech and Virginia Tech were in there. There was even a high school team, the Palos Verdes High School Racing Team, and other teams involved robot enthusiasts and other companies and

research facilities. So who won nobody none. None of the vehicles were able to to go across that three hundred mile routes successfully within the timeframe. So then they held another Grand Challenge in two thousand five. It was a little more um conservative, I would say. Instead of three hundred miles, it was a hundred thirty one point two miles long and UH it was so that made it

less than half as long as the two thousand four challenge. UM. They had five teams completed UH and Stanford went away with the first prize, winning two million dollars. Some of the other teams included a couple of teams from Carnegie Melon and a team from Princeton UH, but the Stanford

team completed the challenge in six hours fifty three minutes. UH. Their car relied on lasers, optical cameras, and radar for environmental perception, and the data gathered by sensors was mapped against a drive ability map that was used to set

the direction and speed of the vehicle. And all the technical papers for all the teams, all of that is available on an archive site at DARPA will try and link that in social so that you guys can see Because you can actually go and read their technical papers and see their approach to building an autonomous vehicle. It's really fascinating stuff, um, particularly if you're an engineer and you can understand all the terms. Even for me, where I could understand maybe twenty of them, I was really

I was really excited to read them. And then you finally had the Urban Challenge, which was in two thousand seven. So the Urban challenge was stepping up the difficulty from the two thousand five challenge. It wasn't as long, but it was in an urban environment, meaning they were going to create traffic patterns. They had people driving actual cars driving around to to make that more like an actual town setting. Uh. They used an old uh no longer in use Air Force base to create their little fake

town and and have the three various courses. They had a national qualifying event. Uh. There were I think eighty nine teams that had applied to be part of this. Thirty five were selected to participate in the National Qualifying Event, which took place over eight days, had three test areas. One course was one where vehicles had to merge into and out of two way traffic in a circulating course. And out of all the test days, only one vehicle

was ever struck by a robotic car. And in my notes I said for getting fresh, So apparently a robot car just Actually, there's some funny stories from this, I'll tell you in a second. The second course tested cars ability to stay within a lane on a meandering road. So if the road has lots of turns and curves in it, can the car actually maintain the right lane? Uh, in order for it to not very into oncoming traffic? Well ideally yes. Yeah, we'll have a lot more to

say about that in a minute too. They also had to go through a a narrow street where they had parallel parked cars on either side of the street, and that section was called the Gauntlet, And the last part of that test was the robot had to locate and park in an assigned parking space and then pull back out and move on to the third course, which was a series of four way intersections that the robot had to negotiate, so the rolways had to detect traffic and

obey the rules of the road. So you know, when you come up to a four way stop, who gets to go through first? It all depends upon when you arrived at the four way stop and your position in relation to the other drivers. And also they put in roadblocks on the course and the robots had to identify the roadblock, make a U turn, and then plan some different Yeah. Yeah, it had to still get to its to its end location. Um. And what was interesting was

that eleven teams went on to the final challenge. That challenge required the robots to visit specific checkpoints and the teams were not notified of the checkpoint locations until five minutes before the launch, so they had a map of the entire space, like they knew what the digital file that had all the map of the fake town, but they didn't know where the checkpoints we're going to be on that map until five minutes before they were pulling out the star the route and exactly they had to

do it right there in real time. The course also had thirty vehicles with drivers moving through the course during the test, so you had real live drivers on the course at the same time, and about half the vehicles were removed from the course for various errors. And here's the fun part. One of the errors was the Terra Max vehicle that nearly drove into the old Commissary building on the Air Force space, so they were able to

stop it just before it drove into a building. Um. And also the team you you see f car spontaneously decided to park at a car port and just take a break, just pulled into a car poard and stopped, just like y'all, I give up, I am done with this test. They eventually disqualified it. Cornell and uh Cornell and M I. T S vehicles gently bumped into each other.

I like to say they were trading paint. They were trying to share a lane together, but they were allowed to continue the course once they were separated and totally given a little time out. UH. And then first place on that went to a team called Tartan Racing out of Pittsburgh, Pennsylvania, and the second place went to Stanford again. UH.

First prize was two million dollars again. So ultimately six teams finished the challenge that year, which showed that even though they wrapped up the difficulty, they still had six teams from across the United States complete that course, which really showed the development of technology in the span of two years, because we remember in two thousand four no one finished, well three years really, because it was two thousand seven when they did Urban Challenge, right, And of

course at this point in two thousand seven, many cars, if not most cars, had on board computers. We're controlling a lot of different systems. It was mostly monitoring, but yeah, there was some control as well. But of course after these challenges, some tech companies a little curious. Are you talking about a company that rhymes with schmoogle? Um? I might be Okay, I remember when this news broke because I have a friend who works at Google who told me that he didn't tell me about the car. He

just told me. It's like there's something that's so cool going on at Google, and I can't wait for it to become public because you're gonna flip out. And it took about half a year before the news became public from when he told me, and sure enough I flipped out. They actually recruited from the challenges they did. In fact, the guy who is the project leader UM is in fact a one of the members of the Stanford team

that won Sebastian. He won that two thousand and five challenge, so he was part of the the winning team back then. And uh, and there are a few others who also are working with Google that participated in the Grand Challenge, and other people who participated in the Grand Challenge are working for specific car companies, so not just Google. Google was not the only one to pull talent from this group.

There are other car companies that that did as well, right, But so Google got to work um decking out some other proprietary vehicles. They weren't creating their own car from scratch. That's one thing to make clear. Yeah, the first one that I remember being released to the public, or the information that's released to the public, was a Toyota Prius actually, but they also they've used Lexus, Yes they have, which

is also a Toyota company. Yeah, but so they in two thousand and ten, I believe it was, they went public with r Yeah, they have been working on it for like a year, but that was when we found out about it. Yeah, there was a big Google blog post where they announced that, look, you know where there are all these traffic accidents, um cars, they're not driving efficiently, and people are spending way too much time behind the wheel.

We can solve all three of these problems at once if we create self driving cars, right, just take people out of the equation and everything goes better because removing quote the loose nut between the gas pedal and the steering wheel. That's my favorite. That's my favorite joke. Mechanics. Yeah, as as we talked about in the previous episodes, the National Highway Traffic and Safety Administration estimates that more than road crashes are due to human error. Yeah, I've seen

I've seen figures everywhere between that that's huge. Like if you could remove that then you would significantly decrease the number of injuries and deaths, perhaps terribly upset the car insurance industry, but which you know, that's okay. Uh. You

also have efficiency to consider. Machine drivers are going to be way more efficient than human drivers because they have like established rules that govern when they gas and break, especially once you set up a network of these things and these at times cars start working together to clear

out traffic exactly. Yeah. In two thousand twelve UM and I Trip Police study, they they estimated that they think a highway filled entirely with autonomous cars instead of with human driven cars, can increase the capacity of that highway five times. Um. Yeah. Now, now we just have to get to the point where we don't need jobs anymore, and then we'll just stay at home in our cars will drive everywhere for us for no reason. Perfect. Uh. Yeah,

you know what's interesting to me? Do you know what they You know what the Google car uses as a primary sensor, right? Is it light ar? Yeah? Have you heard about this? Okay, So imagine the lasers l sixty four lasers. That's not radar for detecting lies, Um, L I D A R. Yes, it's it's sixty four lasers all right. So imagine that you've got a pack of sixty four lasers and a turret top of a car, and that tour. It rotates at six centered revolutions per minute.

And what's doing is it's just constantly scanning its environment so that it can detect minute changes as as it's traveling through a space, so it knows not only what stationary objects are nearby, it can track when things are moving and then be able to plot against that. UM. Now, granted, I think I think in every case that I've read, Google only allows people who are using these cars to operate them on highways. Once they pull off onto surface streets,

it has to go under manual control by Google's policy. UM. And of course these are not cars that are allowed outside of the realms of Google. But that is changing. Yeah, I mean, what's what's interesting is that, you know, it was two thousand ten when Google unveiled this, and there weren't really any laws on the books that would allow for autonomous cars. In fact, most of the laws in the world specify that you have to have a driver behind the wheel in control of the vehicle at all times.

But actually that is now changing, and in a few places in the United States it's already changed. Right in California, Nevada, and Florida, there have been laws passed that specifically allow for the testing of autonomous cars. Nevada was the first in in two thousand eleven, they passed Assembly Buill number five eleven, And what that basically did was it allowed the state Department of Transportation to revise their laws to

accommodate for autonomous cars. Now, in this case, there's still has to be a driver behind the wheel, but the driver does not have to be in control of the vehicle at all times. Yeah. Um. And so Nevada actually followed through with that and in two thousand twelve they officially approved a license for autonomous cars. Yes, I saw that May of two thousand twelve. I know that Florida, Florida being one of the other states that allowed for autonomous car testing. This is I honestly don't know the

reasoning behind this. I'm just gonna throw it out there and see if you guys think it's realistic. Do you think that it's possibly because of the aged population in Florida that this was something that Florida was interested in. Well, I don't know, but I mean I've read that a lot of the reasoning behind autonomous cars is giving more mobility to people, say who like don't see as well as they used to, um, but would still like to be able to have some independent movement around the city.

I would guess that that Disney might have something to say about that as well. Automos cars are legal in California and Florida. Those are the two places where Disney parks are in the United States. So um, Google is also in California. So collusion is a little is a little, a little on the flimsy side. Uh. So that's where it is so far. I believe d C also, I think has been working on this, but now um, there

are a bunch of states that are starting to considering it. Yeah, and the federal government is kind of saying you know, of course, in the United States, for those who are not in the United States, we have state governments and

we have the federal government. The federal government has been on record as saying, guys, you might want to slow down a little bit before you start passing all this legislation on the state by state basis, simply because we don't have enough data here to be able to legislate responsibly. Whereas the technology companies are saying, look, all the data that's coming back is showing that this is way safer

than the alternative. So right, a report published by the Insurance Institute for Highway Safety, which which is a nonprofit although it is funded by the American auto industry um and said that collision warning and breaking systems are already helping drivers avoid accidents. Um that that statistically, if you were drive, if you are in a car that has these systems, you are less likely to be in an accident. Yeah.

And also it's good to keep in mind that typically these laws, uh, they sort of unveil these new motor rights in different stages, so they're opened on private roads before they are on public roads, right, Um. And I guess that's to allow for testing. And and it's it's not just Google Old that's getting into this game. Actually, there are quite a few companies that are all investing in their own automated car technology. Toyota is another one,

BMW four GM, Volvo, Audi, Mercedes, Benz, Nissan, Right, right, Toyota. Yeah, all of these have at least in the testing stages various forms of these things, and a lot of them are saying that within the next ten years they're going to be rolling out, if not fully automated cars at least um auto assist cars with more, because again, this may be one of those things where because of the legislative barriers, there may be some time that we'll have

to wait to see automated cars being available for the average consumer. I'm sure they'll be prohibitively expensive for most people when they first when they first hit the roads as well, But I think it's gonna be. I think it's gonna be a few years just because of the legislative side. But we will still benefit from that technology in other ways, with more more advanced cry prevention systems,

parking assistance, that kind of stuff. Yeah, speaking to the price, I believe right now an autonomous car is running about a hundred and fifty thou dollars. Yea. I saw one guy say that that there's always the possibility that before we see fully autonomous cars rolling off the manufacturing plants out of all these these companies, that we may see aftermarket kits where you can convert a car over to an autonomous vehicle. But even that would be in the

hundred dozen dollar range, so still not cheap. Sure, Yeah, And you know the kinds of companies that are putting out things that you can buy right now that are partially autonomous are like Mercedes Benz, which has a radar and stereoscopic cameras that will help you determine, you know, help your car determine the speed of traffic around you and also your position within a lane and keep you

there up to speeds of about eighteen miles per hour. Yeah, and I think, uh GM, they've said that they're they're going to have partially autonomous cadillacs by well, we'll see. That's really close. Yeah, yeah, yeah, and partially autonomous. I mean that's vague enough where you could argue that there are cars on the market right now that have that one of the upcoming systems and that I'm excited about. And and another another sponsor plug, This one is from Toyota.

UM is a cooperative adaptive cruise control, part of which uses transceivers to broadcast speed, acceleration, and breaking in between different cars on the road, so that way you can maintain the proper vehicle distance between you and the car in front of you. That kind of thing, right with something a little bit more interactive than just the the lasers. The receptors that are that are figuring out what's going on more than just an alarm or something on those lines. Yeah.

In fact, um M I T had a great little post about some recent research that came out of the Google experiments. A. Google has been running these these cars for something like over three hundred thousand miles accumulatively. If you had them all up together, it's like three miles of travel time. They've had one accident, and according to according to Google, that was done while the car was under manual control, not under computer control. So I think

they also got rear ended ones. Yeah, but the when they took the black box readings, because a lot of what is going into these autonomous cars is also a very intricate um recording system for everything that's going on in the car. I mean, you know, they are in the testing stages, so of course that data is very valuable. Um. But it proved that the autonomous car breaked appropriately and that the guy behind him that rear ended him was not paying attention and just just just hit him because

because the car did exactly what was supposed to do. Yeah, a lot of this research that's going on right now is about how to manage the interface between humans and computers, and you know where a human should take over, where a computer should take over, what specialties both of these operators can have. And some of the questions being posed are about, you know, whether or not reckless driving behavior

is actually really important in some situations. You know, whether or not a computer could handle a situation wherein you kind of need to go over a line, or you need to speed up or slow down at a at a rate of speed that isn't otherwise advisable, or break a traffic law. Now, it does seem that for the most part, the uh, the autonomous cars are able to respond much much faster than any human could, in a more smooth transition than humans. However, that's that's based upon

a certain set of rules. Well, yeah, and and assuming that everyone is in an autonomous car, that's fine. But when you are the one person in an autonomous car and you're driving in a town full of crazy people like go on no Atlanta, then maybe that your car is doing everything that's right, but you know, it may still not save you from a crash. Of course, we can't plan for everybody's behavior, obviously, you can only do

the best you can. One of the other interesting things from that m I T post was that the data shows that automated car spin time and what they called near collision states, meaning that that that's how you usually drive. I don't drive at all, so so that makes it easy. Uh yeah, but ner if I were to drive, I'd

be in a constant near collision state. And some might argue that Georgia is a near collision state, but the you know, the the actual definition there is when you are engaging in behaviors that are more likely to cause the collision, things like accelerating too quickly, breaking too quickly,

following too close, following much too closely. In the automated cars, because you set up these very specific rules for them to follow, they have to follow those rules, so they tend to operate within much safer parameters than your typical

driver would. Though also I've read if if you were to imagine entirely automated highway, those cars can follow much more closely, much more safely, if you were able if everyone were automated, Like if if either of these we have that infrastructure and all the cars are talking to one another, then the theoretically you could do that and also ours behind you know, larger vehicles would even benefit from drafting at the point you're talking about saving a fuel,

But you know that's building out an infrastructure is is really really hard, and uh, I think I think we'll see automated cars before we see automated infrastructures on a wide scale basis. Though, an interesting thing about the most recent developments and automated cars is they might not depend

on that much infrastructure like we used to imagine that. Well, in order to really have automated highways, you need like you know, computers in the roads that control everything, right, But we're creating fleets of automated cars today that do you just find simply by reacting to each other, right, and or through differential GPS systems so that they're paying attention to to not only your position via satellite, but

via markers in the road. Right. Like I said, I think, I think there are a couple of different barriers that we have to get over for automated cars to become a reality for the average consumer. The big one being legislative, because I think it is going to take a lot of work to get to a point where automated cars have a legal designation that is recognized, not just within a state, but between states and within between countries even for them to be a viable form of transportation, I think.

I think the thing that's going to help with that is is the parallel to all of this, where in insurance companies are providing black boxes as an opt in option for opt in option, that was a really great turn of phrase for for for their drivers to potentially get a lower rate on their car insurance. And not many people, but maybe like about a third of people I supposedly are signing up for these things, and so they're starting to gather data on driving habits a lot more.

On top of that, the other barrier I would say is we've seen this happen a lot in various networks, right where you get a proprietary network where other networks can't necessarily communicate with it. So, for example, we talked about an automated house, and right now, you can do a lot of automation at home. But the easiest way of doing that is to buy all your stuff from one source, because buying it from multiple sources means you have to find some way of making them communicate with

each other. Sure, however, if you have for example, and UM Apple TV and also an iPhone and also right, if you're all within one ecosystem, you're great, but if you're not. If you're if your ecosystem is not homogeneous, then you're in trouble. Well, as we know, traffic is not a homogeneous situation. People have all sorts of different cars out there from different models, models and different makers.

So unless we were to have an agreed upon SEV standards that all vehicles were using in order for their UH information to communicate properly, then you have potentially a fleet of cars that doesn't really communicate with each other very well. That they need to all speak the same language. And um, that's something that maybe has yet to be devised. In fact, I think we talked in our previous podcast about this about maybe the need for something like a

traffic protocol. Yeah, yes, yeah, And it's one of those things where you you need either there needs to either be a leader in the space and everyone ends up following suit because they just they've defined it, or there needs to be some sort of regulation. Yeah. Like like and where did i P camp come from? Oh, well, that came out of our bonette, So that was like a public initiative that created the protocol that all of

the private organizations would end up using. Right, so you could get it some form of independent entity like maybe even the I E. Being able to come up and you know, a group of engineers that come up with a A a standard language that would be used. But I mean, all of this is jumping way ahead. We still aren't entirely sure when we'll see these cars um become something that we could actually purchase. To mention i E. Again, they predicted it also last year that in their view,

by of cars on the road will be autonomous. That'll be how that's a that's bold. I would love that future so much. Just I'm imagining how traffic would become such uh so much more smooth. Studies showed the young people like this idea. It's more popular more. I'm liking it even more now. Uh well, young young people are certainly the ones who are most likely to pick up these black boxes from their insurance companies. So I so

I guess that would that would make perfect sense. Yeah, But I mean I think I like it the most because I don't drive, So for me having something like this where maybe I do, you know, drive occasionally, but I have a vehicle that does most of my driving for me would give me uh even more freedom, or more importantly, would mean my wife is no longer my chauffeur.

Oh sure, And as we talked about in that previous episode, it would certainly open doors to people who are who are unable to drive for various physical health sure yeah or not health. But I remember I do. I did read one thing from uh M I T Professor John Leonard, who has worked on navigation systems for autonomous cars, who said he doesn't believe we'll see a totally autonomous taxi

service within his lifetime. He doesn't think that's ever going to happen while he's still uh, you know, walking treading the earth so uh, and he's he's working on the problem. And it's not again, it's not just a technological thing. It's also a social thing. And that was kind of what he was pointing out, that that taxi drivers um provide an important role in society, particularly if they are

efficient and nice. Uh. He was talking about this about Manhattan, but said, really, I mean more of the European style taxi drivers. I was about to say, alright, so I think I think we're good on self driving cars for the near future. Of course, I say that probably tomorrow there'll be some breaking news they'll necessitate us to do a whole new podcast. But that's okay because I love this topic. So guys, make sure you go to FW

thinking dot com. That is our homepage where you can find the blog posts, podcasts, videos, articles, all sorts of cool information about the future. Go check that out, and also look for us on Facebook and Twitter with the handle f W Thinking and we will talk to you again really soon. For more on this topic in the future of technology, visit forward thinking dot com brought to you by Toyota. Let's go Places,