Episode Preview with TTI's Bob Brydia (audio, 24s):
Full Episode (audio):
March 2, 2021Episode 3: Promises, Promises. Turns out, building a self-driving car is a lot harder than we thought.
FEATURING: Bob Brydia
Autonomous vehicles were supposed to be more widely available by now, but they’re not. So, what happened? Those cars have to learn how to do more than just drive, TTI Senior Research Scientist Bob Brydia tells us. They have to learn how to think, too.
About Our Guest
Senior Research Scientist
Bob Brydia is a transportation technologist with more than 30 years of diverse experience with national, state and local transportation research sponsors. Bob has led several large-scale deployment projects, including testing automated vehicles in real-world environments, and has developed multiple innovations to advance transportation.
Bernie Fette (host) (00:14):
Hello. Thanks for joining us for another episode of Thinking Transportation, a podcast with experts talking about how we get ourselves and our stuff from point A to point B. I’m your host, Bernie Fette, editor-at-large at the Texas A&M Transportation Institute.
Bernie Fette (00:33):
Have you ever seen that sign on the side of the seafood restaurant that promises free shrimp? The sign reads: free shrimp tomorrow. That sign — and its promise — never change. Come back tomorrow, and you’ll get the same guarantee — for the next day. Sounds a bit like what we’ve been hearing from the companies who are trying to build self-driving cars, or what researchers call autonomous vehicles. Bear with us. We’re almost there. That was 2019, or was it 2018? So, what happened? Where are all those self-driving cars that we were promised? Our guest has a few thoughts on those questions. Bob Brydia is a senior research scientist with TTI. He’s spent much of the past few years studying what it’ll take to make those cars work. Thanks for joining us, Bob.
Bob Brydia (guest) (01:31):
Thanks, Bernie. Great to be here.
Bernie Fette (01:33):
Could we start with some basics? Can you give us a quick lesson on autonomous travel and include in that, what exactly do we mean when we say “autonomous”?
Bob Brydia (01:44):
So, probably the first thing that we need to do is to just make sure that we’re talking the right terminology. You know, in this type of environment, words matter, and people throw around “automated” and “autonomous” as if they are the same thing, and they’re really not. So “autonomy” really means the full science-fiction type movie descriptions of self-driving vehicles. “Automated” is basically highly automated systems that can assist the driver in driving down the road. But the driver needs to be aware at all times. The examples of what you might think of as automated are the typical driver assistance technologies that we see today: forward collision warning, cooperative adaptive cruise control, lane departure, keeping in your lane, steering, those types of things. Those are all automated technologies. “Autonomous” would be, I’m sitting in the seat and I’m watching a Netflix movie while my car drives me to work.
Bernie Fette (02:46):
Okay. So automated is just a stop on the way to autonomous?
Bob Brydia (02:51):
That’s correct. One of the ways that you can think about things is progressive levels of disengagement of the driver. So, first you have feet-off, then you have hands-off, then you have eyes-off. Then you have brain-off. Autonomous is all the way to brain-off.
Bernie Fette (03:09):
So, I mean, we’ve been hearing for quite some time now for the last few years, those self-driving cars are going to be on the roads very soon. Why aren’t they there now?
Bob Brydia (03:20):
The short answer is, it’s much harder than everybody thought it was going to be. You know, one reference that I like to bring up is the fact that airplanes, you know, the big jets that we aren’t currently flying in, have been able to take off and land by themselves completely autonomously, you know, without any pilot assistance for literally decades. But that’s because airspace is really a very simple environment. There’s not a lot of interference up there. There’s very set rules as the following distances and those types of things.
Bernie Fette (03:54):
Yeah. You don’t have a whole lot of competition.
Bob Brydia (03:55):
Exactly. So down here on the ground where you’ve got, instead of another airliner being 2 miles away from you, you have another car, 20 feet away from you. So the reaction times have to be much quicker. The inference, the pickup, the understanding the environment — it’s all much more difficult. And one of the biggest things is that drivers actually drive not only with the basis of the rules of the road, but they also drive by intuition and what they assume and think is going to happen. You know, you’re driving down the neighborhood street and, you know, the kids are playing a kickball game in the front yard and the ball comes out into the road. You know, a typical driver is going to assume that this child is going to run out in the middle of the road, not look, and pick that ball up. And they’re going to proactively brake the vehicle to make sure that they don’t hit the child. You need to teach a computer to do that. That’s hard. The computer has to think.
Bernie Fette (04:59):
What you were mentioning earlier about the autonomous nature of a big commercial jet landing by itself — I think that that’s something people just don’t consider, a car to be as complex a machine as a commercial airliner. So part of what we’re talking about here is not just the vehicle itself, whether it’s a plane or a car; it’s the environment that vehicle is operating in. This is about more than just autonomous vehicles. Isn’t this at least partly about the infrastructure that those vehicles need to operate on?
Bob Brydia (05:32):
Absolutely. So you look at a couple of things. You mentioned about the vehicles themselves, and I’m not going to get the exact numbers here correct, but basically a vehicle on the road today — it’s actually fairly surprising. They have about three times as much software code in them as an airliner does. So, we have a much smaller vehicle operating in much lower speeds, and yet it has, you know, two to three times as much software code as a commercial airliner flying through the air.
Bernie Fette (06:04):
Yeah. And operating in much closer proximity to other vehicles in that environment.
Bob Brydia (06:09):
Correct. And so then you, you add in the fact that there’s all those complications in the environment. There’s all these rules. There’s signs. There’s pavement markings. There’s work zones. There’s weather conditions. There’s people doing weird things. There’s squirrels that run out. All of these things are inputs that, that the computer system that’s in the vehicle has to deal with that other environments don’t have to deal with.
Bernie Fette (06:33):
And in addition to it being a technology challenge and an infrastructure challenge, we’ve also got the challenge of policy, right? I mean, what about the issues related to traffic safety and law enforcement?
Bob Brydia (06:46):
Those are really good questions. You know, the industry has not yet determined what the answer is to all of those questions are. For example, who’s at fault when an autonomous vehicle crashes? I mean, it will happen. It has happened. But, you know, the consequences of who’s at fault, whose insurance company pays, you know, what, what type of data can you get out to support from it? All of those questions are still out there. It’s one of the reasons why most companies out there that are working on autonomous vehicles are taking it essentially very slowly. They’re certainly pushing fast, but, but they’re doing it very methodically. They’re doing it with very careful assurance to a rigorous test protocols and all of those types of things because, you know, nobody wants to read about an autonomous vehicle killing people.
Bernie Fette (07:41):
Right. The margin of error here is really pretty slim.
Bob Brydia (07:45):
Yeah. And it’s really kind of strange because, you know, as some of our listeners might know, typically about 30 to 35,000 people die per year on the roads in America from just regular accidents, you know, so … but yet if we had 35,000 people being killed every year in vehicles that were being driven by computers and robots, there would certainly be an outcry about that. So the standards are very different.
Bernie Fette (08:11):
You mentioned earlier, Bob, that teaching a machine to think, teaching a machine how to drive has been harder than we expected it to be. Can you give us a few examples of how that has become apparent?
Bob Brydia (08:25):
So, I want to caveat my answer, Bernie, by saying that I’m not deep into the actual design of, you know, those computer systems and algorithms that folks are using.
Bernie Fette (08:35):
Bob Brydia (08:35):
So, this is, this is from observation only. But you know, what it really comes down to is it’s pretty easy to program a computer, to follow a series of steps — if A, then B. Or if I have a choice and this one has a better outcome, I would pick that choice — you know, those types of things. We do those things all the time, all day long, and our brain does that as well, but it really comes into the judgment call type of situations. What do you do when you’re driving down the road, the light in front of you turns from green to amber and you’re 20 feet from the intersection? Lots of people punch the accelerator, try to go through. Other people brake hard and try to stop. You know, those are individual choices based on their thought process of their risk, what they are willing to accept, how people are moving through the system, what the environment is around them.
Bernie Fette (09:38):
Yeah. And how much experience they have with driving.
Bob Brydia (09:41):
Exactly. And, and the weather conditions. You would make a different decision, hopefully, if it was a rainy road or a snowy road. And so, you know, an autonomous vehicle’s brain, the computer, has to take all of those inputs and assimilate that and make a choice in a nanosecond.
Bernie Fette (10:01):
And it might be a decision that has to be made with a set of conditions or circumstances that are just very rarely seen.
Bob Brydia (10:10):
That’s correct. One of the problems is, you know, there’s kinda two different paths that manufacturers are following to try to create autonomous vehicles. There’s a mindset, you know, in one camp that is saying, okay, I’m going to put sensors on the car. And those sensors are going to be able to detect the environment all the way around me. And I’m going to be able to interpret that environment and make any decisions I need on the fly, so to speak, without having to follow a path or, or anything else on the road, essentially with no prior knowledge. Like when we’re driving down a road, you know, we’re going to a new address and it’s a hilly country road. Uh, and you know, you kind of adjust your driving based on the fact of, you know, what you’re seeing and all, because you’ve never been on the road before.
Bob Brydia (10:59):
Then there’s the companies that are working on things that are more in the camp of rigorously mapping, all of the available roadways. So that there’s a baseline of information, which would be equivalent to us, you know, knowing the road that we drive to work every day, because we’ve driven it 10,000 times, those are two very different driving environments. And those are really kind of some of the different models that companies are using to develop fully autonomous vehicles.
Bernie Fette (11:28):
And all those have to be taken into consideration in teaching that car what it should do.
Bob Brydia (11:33):
Right. And then you have to throw in your unexpected conditions. You’ve driven your work-to route 10,000 times, and this morning, and you drove it and there was an accident, or there was a work zone that wasn’t there yesterday. Your brain knows what to do with that.
Bernie Fette (11:48):
Bob Brydia (11:48):
But if you’re an autonomous vehicle and you are looking to follow rules that you already know, and that information isn’t in your rule set, that could be challenging.
Bernie Fette (11:59):
Bob Brydia (11:59):
But it could also be challenging for the vehicle that is interpreting the environment as you go to make accurate assessments, because it could be changing very rapidly.
Bernie Fette (12:10):
Yeah, speaking of those unexpected conditions that you just mentioned, we’re a little bit over a year into a global pandemic. Has that set of circumstances changed the research path or slowed it down in any way or sped it up in any way?
Bob Brydia (12:28):
So again, I don’t work for any of the companies that are actually physically doing this, but from what we hear, it’s both helped and hindered in some ways. The testing and the focus that they’ve been able to maintain doing some of those trials on-roadway in areas because of reduced traffic and everything have probably actually been helped a little bit. But at the same time, you know, the whole concept of physical separation, social distancing, and those types of things and adhering to COVID protocols, when you need more than one person, you know, working in a closed space, that’s made it a little harder. But I will say that the industry that will be rolling out autonomous vehicles faster than anybody else is the freight industry. And they’ve been making really good progress over the past couple of years, including during the time of the pandemic, to continue to accelerate their research and innovation and on-road testing.
Bernie Fette (13:31):
Okay. Bob, can you talk a little about how the research that you and your team are doing? What are you working on and how might the work that you’re doing help to move things along a little more quickly in terms of getting those self-driving cars out on the road?
Bob Brydia (13:47):
So there’s research going on in all kinds of areas related to autonomous vehicles. One of the really interesting aspects of some things that we’re working on with a team of researchers across the country, working for the United States Department of Transportation, is really looking at the future of autonomous vehicles and how they’re present in all forms of transport that we see. So, the freight industry, you know, transit buses, picking people up at the grocery store and delivering them to home, automated deliveries from — did your order from Amazon come into your door… How does all of that fit into the system and the infrastructure and the road types and everything that we have in place today? Uh, and what changes — would we need to consider to make these vehicles operate much better than they might in the environment we have today? We’ve talked previously about the environment, the physical infrastructure, the roadways, the conditions, the guard rails, the driveways — all of those types of things. The environment might have to change. We might have to make changes to it in order to help these vehicles. We’re also looking at specific examples of problem situations, such as work zones, and could autonomous vehicles that not only sense the environment around them and understand what they’re seeing, can they also communicate that to vehicles that are behind them and give those vehicles an alert so that there’s additional warning time? And then not every vehicle that comes onto the situation has to understand it and interpret it for itself. So that brings in both connected and autonomous vehicles, where the lead vehicle understands the situation, and then communicates it back down the roadway.
Bernie Fette (15:38):
Considering everything we’ve talked about from your vantage point as a researcher, what do you think people can expect in the next 10 years, maybe?
Bob Brydia (15:47):
So as I said earlier, the industry or the portion of the industry that’s going to experience autonomous vehicles first is freight. So those 18-wheelers that you see driving across the interstate, across the nation, and, you know, on any freeway and roadway, those are going to progress faster towards autonomous vehicles than any other vehicle out there. It’s a simpler environment for them. You know, they’re doing long-haul trips across interstates that are largely well-maintained and have good infrastructure. We also get asked, when am I going to be able to have a vehicle like this in my driveway? When am I going to be able to get up in the morning, go out to my car, tell it to take me to work and catch up on last night’s TV show? That, my opinion, it’s a few decades off, at least, just to be able to make sure that these things are affordable; that they’re safe.
Bernie Fette (16:43):
And you’ve got quite a few drivers out there who really want to hang on to their ’57 Chevys.
Bob Brydia (16:48):
Absolutely. There are lots of people that just love driving and love the feeling of it, and being in control and making the choices. It’s relaxing for them. You also have a heavy percentage of people who, you know, don’t trust this technology. They haven’t seen it enough. They haven’t experienced it enough. And so they’re like, “no, I’m not going to let a robot drive me. I’m going to make my own choices.” You know, giving up that level of control is something that the industry is also going to have to face to get this to a much more widespread rollout than just in a trucking industry.
Bernie Fette (17:23):
So, Bob, whenever you’re talking about the, I guess the public acceptance of self-driving cars, you’ve actually done a little bit of work in that area, uh, not all that long ago here on the Texas A&M University campus, right?
Bob Brydia (17:38):
Yeah, we had a really great project here where we brought in an autonomous shuttle. And we put it on the road in mixed traffic, on a route with two stops, and we ran it for three months to see how people would accept it and would they use it and what their thoughts and and feelings were on it.
Bernie Fette (17:56):
What were your takeaways from that?
Bob Brydia (17:58):
Our takeaways are that the technology is hard to achieve consistency, but that over time people did learn to trust it more. In particular, we focused on the operators; we had a safety operator or a safety driver in the vehicle at all times. And you know, so our trust factors that we looked at were largely their trust factors at the beginning of their employment versus at the end of their employment. And they definitely increased, but that’s just something that shows that it’s going to take time and experience for, for people to really meld this new type of technology into their everyday life.
Bernie Fette (18:37):
What I’m curious about, is lots of us, you know, have interests in certain things about what we want to be when we grow up. I’m just wondering whenever everybody else was out playing baseball, were you working with robots, or what inspired you to get started in this area?
Bob Brydia (18:55):
Well, I’ll make this answer really quick. Growing up, I was always more science and math minded and not nearly as athletically inclined, but what really started me in transportation was, in college, I knew I wanted to be in engineering. I took a transportation course and the professor that I had at time who was visiting my college from Germany, and who was one of the renowned experts in geometric design in the world, was lecturing. And he just brought it alive. He brought everything that he said, you could just see it in your brain. It just enthused you. And literally that said to me, this is what I want to do.
Bernie Fette (19:40):
So, this is slightly different… What is it that motivates you to do the work that you’re doing?
Bob Brydia (19:47):
What motivates me and really, what is the most fun about what I do is that for all practical purposes, I don’t ever do the same thing twice or, or even, you know, two days in a row — we always have different projects. We’re always looking at new problems. We’re always trying to figure out how to make things better, how to make things more efficient, how to make things more safe. Doing that, and the constant challenge of evolving and being innovative. It’s — it’s a privilege to be able to do that.
Bernie Fette (20:21):
Okay. Bob, given your predictions, we should check back with you in a few years, maybe to see if you might have a fortune teller career in your future.
Bob Brydia (20:28):
That sounds good. I’d be happy to talk to you about where we are with freight and see how my predictions for vehicles in our driveway are holding up.
Bernie Fette (20:37):
Thanks for talking with us today.
Bob Brydia (20:39):
Thank you. Appreciate it. Glad to be here.
Bernie Fette (20:44):
This discussion with Bob reminds me of a brief study TTI conducted just a few years ago. Researchers asked a sample of drivers in one city whether or not they saw a self-driving car in their future. The responses were split, right down the middle. Half of the drivers said yes, and half said no. AAA published their own study in 2020, showing that people are still concerned about issues like liability and safety. That suggests that a lot of Americans still aren’t entirely ready for self-driving cars. And maybe that’s just as well — because the cars themselves aren’t yet ready, either.
Bernie Fette (21:27):
Thanks for listening. In our next episode of Thinking Transportation, we’ll visit with Greg Winfree, TTI’s agency director and a former U.S. assistant secretary of transportation. Greg will be sharing his thoughts about what we might expect to see from the new leadership in Washington, D.C.
Bernie Fette (21:47):
Thinking Transportation is a production of the Texas A&M Transportation Institute, a member of the Texas A&M University System. The show is edited and produced by Chris Pourteau. I’m your host and writer, Bernie Fette. Thanks again for joining us.