Curious Conversations, a Research Podcast

(music)

Travis

When I was a kid, only place you would see self-driving cars or digital signs that would change as you drove was in movies. But today, you literally can see them on the streets, and they're becoming more and more common.

So I'm curious what these emerging technologies mean for drivers. What do they mean for passengers? What about pedestrians? Are they going to be impacted by some of these shifts? Well, thankfully Virginia Tech's Mike Mollenhauer is an expert in this very subject and was kind enough to join the podcast. Mike is the director of the Center for Technology Implementation at the Virginia Tech Transportation Institute. There he leads business and product roadmap development activities, including project management for commercial and government contracts, as well as evaluating in-vehicle technologies and quantifying driver safety behaviors. So Mike and I talked a little about future transportation and what this term smart mobility means, what smart infrastructure means, and what does that look like practically to me and you. We talked a little about autonomous vehicles and how they can be folded into a larger system, what the infrastructure needs would be for something like that, and some adaptive signage that they're currently working on in So if you are a driver or you're thinking about picking it up in the near future, I think this conversation will have a lot to offer you. I'm Travis Williams and this is Virginia Tech's Curious Conversation.

(music)

Travis

I want to talk to you about smart mobility, smart infrastructure, how these things get together. And as I was thinking about that, I thought maybe a good place to start would simply be when we call something smart right now, specifically in the transportation world. What does that mean?

Mike

Yeah, that's a really good question. And I think, I think it was a bit of a marketing term at the beginning, right? To get people excited about things. But I think it's more evolved into solutions that apply techniques in both machine learning and machine vision to derive, I guess, insight from sensors that are deployed, you know, in remote places in a lot of cases. But what, in my opinion, what really makes it smart is when you can combine the results that are being derived from these systems together to provide actual insight. So I can identify where I have, let's say like a problem with pedestrian crossings at an intersection because I have accumulated enough data from these systems to show me where that's at and exactly what kinds of conflicts those are. And some of that, you know, involves video recording. Some of that, ⁓ you know, involves use of AI to drive that insight. But I think having, having that kind of information, both looking at near misses and actual conflicts and crashes, makes us a little more smart in that we could be proactive about the types of countermeasures that we would develop and deploy to support solving that problem.

Travis

Yeah, that makes a lot of sense to I've heard the term machine learning before, but I don't know that I've ever heard the term machine vision. What does that mean?

Mike

Yeah. So machine vision, it's a technique that's been around for awhile that you look at success. You can look at images or successive frames of images, like you would get from a video. And through a training process, you can train a machine learning model effectively to identify objects and classify them and even track them through a scene. And so if you know the intrinsics of your camera, like the field of view and the distance and where it's located. As let's say a person walks through an intersection, you can continue to effectively draw a bounding box around that person. And then you track that bounding box across an intersection so that you understand where they are, how they're moving both themselves, but relative to the other actors that are at the intersection.

Travis

Well, it sounds like that you all are getting a lot of information, or at least that's a piece of any type of smart infrastructure. it mostly data from previous events? Are you also getting real time data?

Mike

We are squarely into the real time space now where we are running these systems 24 seven live. The data from them comes into what we call a data exchange. And that's at very low latency. Oftentimes from the time the image is captured at an intersection to the time that we get a report of that object or thing and where it's at in the intersection is less than a hundred milliseconds. So very, very quick exchange of data. And the whole idea here is to be able to perform real time manipulation of things like traffic signal control plans so that if you know somebody's present in an intersection, you might change how you change the lights. Right? So right now, when you come to a, let's say like a stoplight, you would press the cross, you know, the signal, sorry, the pedestrian crossing button in order to get the lights to give you a walk signal across the street. could do that now with just basic video analytics, right? To look and see where you are, make some assumptions about where you're headed and kind of press that pedestrian button for you. Another example would be if our cameras would detect a vehicle that's approaching a traffic signal, it's about to turn red and we're not seeing any evidence of deceleration, like they're slowing down and going to stop. You could potentially hold the crossing green phase, right, from going green so that people don't enter the intersection while that vehicle is effectively violating that red light.

These are all concepts that have been tested and we think we can show benefits with those and we're looking for opportunities to deploy and test them in the field.

Travis

Are there any areas right now when I'm driving around the real time data and smart infrastructure? Like are there things happening when I'm driving that I just don't really even know about?

Mike

Yeah, right here in Blacksburg at South Gate and Beamer, there are sensors set up to track traffic going through that intersection. We are gathering data at this phase and understanding how we would manipulate and or change the signal, the traffic signal timings at that intersection that will eventually be hooked up to go live here in the not too distant future. So hopefully we see an improvement in the reduction, right? In the amount of waiting time that you spend waiting at that intersection when there's no cross traffic, cross traffic, sorry. I know we've had complaints about that intersection in the past, so this is an opportunity to see if we can improve it. We did select that intersection because it is right down there near the stadium. And one of the aspects that we evaluate on the technology is how well does it do when you have game day traffic or a Metallica level traffic, right? So, you know, we're trying to figure out what the limits of the capabilities are. We have one other intersection at Industrial Park in South Maine as well where we're testing a similar type of a concept, but it's a clearly a different type of intersection with a lot less pedestrian traffic.

Travis

So it sounds like that the idea would be that I could possibly wait, wait less, which sounds awesome to me, especially at a clear intersection.

Mike

Yeah, that's the whole idea. Waiting less, burning less fuel, spending less time. In fact, improving safety. We know that when vehicles are moving smoothly through an environment, they're much less likely to be in conflict with one another.

Travis

Well, I've heard also, I guess I've heard that UOs are also working on maybe some adaptive speed in certain areas. Is that a thing that you're able to use real time data to change?

Mike

It is. So we are actually working ⁓ in partnership with the Virginia Department of Transportation on what's called a variable speed limit corridor. So that is on a 15 mile stretch of road from Fredericksburg South on Ive. And the northbound lanes have been outfitted with a special radar devices that are about every third of a mile. And so they're evaluating the flow of traffic in that, in that, and when a slowdown occurs, let's say like at one end of that region, we can change the speed limit signs proactively to slow the flow of traffic coming into that area. And the idea is that you don't want to get into this sort of shock wave of start and stop driving that we see, which will ultimately result in more rear end collisions. So that has been tested and that is showing a significant reduction around 20 to 25 % reduction in crashes along that, that stretch of roadway. So we know that that They get those kinds of benefits, even though compliance is still relatively low from the human drivers, right? They see the new speed limit signs up there and people being people, they tend to drive, you know, seven to 10 miles per hour over that, that limit anyway. The new concept that we are developing is called automated speed control or intelligent speed assist. And this is a function that would run on your vehicle that would take information coming from that same traffic operation center that's computing these prescribed speeds. It'll go right into your car. And if you have your cruise control set, it will do all those speed changes for you. So you don't have to worry about reading the signs and you don't have to worry about complying or not. If you just hit the button and say, follow those speeds, it will do it. And so our first study is a pilot study where we're going to have 125 drivers drive with this solution over the course of six months. So we'll be giving, giving vehicles to people to drive and we'll be studying what their, uh, utilization of the system is. Do they want to use it? We'll get their subjective feedback about how they felt with it, you know, turned on. Was it, you know, working properly? Did they feel like they were going too slow? There are, there is an aspect of the study where we'll allow them to add a buffer to that prescribed speed. So just like on your regular speed, your regular cruise control, you can bump up the speed a couple of miles per hour if you want to, to go a little higher than the compliance speed. The whole idea here is to study how a human would use it sort of in a naturalistic setting. And so with that, we'll also look at safety related events. So if we have, you know, a third of the vehicles out there that are complying with the speed limit and everybody else wants to go a bit faster, does that cause any road rage or safety type events that we need to be ⁓ cognizant of? So this is really interesting study, one of the first of its kind here in the United States. So we hope to learn a lot so that we can inform the Virginia Department of Transportation about whether investment in these types of systems makes sense moving forward. And then can we also use the data that's being transferred between the vehicles themselves and that traffic operations center to eliminate the need for having these variable speed limit signs and having the radar units out there because it's a little bit expensive to deploy all that technology. So the more we can streamline that process and eliminate the need to deploy hard infrastructure, the more likely it is we'll see the benefits from that system.

Travis

Yeah, that sounds like a fascinating project. know anytime on the interstate that, you know, that kind of, Hey, we all saw a deer on the side of the road and everyone suddenly shifts and slowly, that can be very stressful.

Mike

Absolutely. And especially when it gets into that start and stop driving, that's when you really start to see an elevation of rear end collisions. And then from there you get secondary collisions because everybody else has stopped and waiting for that traffic to clear. So yeah, the more we can do to avoid significant differentials in speed in the traffic, the better off we'll be on the safety side. ⁓

Travis

Well, how does automated driving systems fit into this equation?

Mike

So to date, most of the automated vehicle developers have been very, well, I'll call it proactive. They want to kind of control their own future. So they're not waiting for infrastructure-based solutions to move ahead with their products. And so they pretty much rely on onboard sensors, meaning their own cameras, their own radars, their own LIDAR to understand, you know, where they are in the environment and what's happening around them. We believe that if we had vehicles that were connected, that have a special type of radio that allow the vehicles to talk to each other, but also allow the vehicles to talk to infrastructure, that we could deploy a safer automated vehicle solution. So we can tell that vehicle about what is around it that it can't see because of normal blockage. Let's say like if it's following an 18-wheeler right on the freeway, we know we have a hard time seeing around these vehicles to see what's ahead. If you could communicate, if there's a hazard ahead of some kind, like maybe there is a traffic queue, then that automated vehicle could prepare for that scenario and be ready to respond when the time comes. And, you know, this would really help eliminate what we call, you know, the blind spots around that vehicle.

Travis

Yeah, it sounds like you would just be able to give those automated vehicles more information to make decisions with.

Mike

Correct. And sometimes that means, everything is safe. This is a good place to drive. Or it might mean, hey, we've had a traffic crash and there's going to be a lot of backed up traffic. There's going to be an officer there who is using hand signals to divert traffic. Since your system doesn't really understand that all that well, just avoid it altogether, right? Take a detour, get off that road for a segment and just avoid those scenarios because sometimes it's really hard to teach an automated vehicle to do something like understand hand signals from an officer.

Travis

Yeah. Well, what are some challenges to implementing smart infrastructure, smart mobility, this technology?

Mike

The technology is pretty young, which means that it's changing rapidly. It's got a cost to it, like any other infrastructure based solution that we would put out. But they, I think we're having to do a lot of baseline evaluation of the technology to make sure that it lives up to its marketing, I guess, promises. And so that's one of the activities that we do regularly for VDOT is to look at these different products and then say, Hey, that works really, really well most of the time, but when it rains, it doesn't work at all. Right. Or, you know, it works pretty well, but it over counts by a certain amount when it's counting traffic and pedestrians. So we like to work with VDOT to give them that information, but we also like to inform the vendors. We're not here just to tell them what doesn't work about their product. We're trying to inform them about what we see and try to help them come up with solutions for how to make their product better. And so right now, because that is in such a rapidly advancing space, it's difficult to stay ahead of where they're at and keep up with all the developments that they've had in the product lines and then make sure that we're choosing always the product that is correct for how we intend to use it. So it's really matching up the technology to the use case, right? To make sure we're getting what we think.

Travis

Yeah, you mentioned some of that living up to the, to the promises. It makes me wonder, is there a human component, like a perception or an education curve that it makes this also challenging because like I didn't know anything about what you just told me until just now. So I'm curious about that. Is that challenging?

Mike

I mean, of course there are human concerns. call them concerns, but there's really opportunities as well. Right? So some of the concerns are that we over-trust a technology that is not a hundred percent trustworthy. We see that in automation. We've seen that on the Tesla vehicles, you know, with full self-driving, they are really, really good in terms of the technology and how it can drive. But that system is really meant to have a human driver ready to jump in and take over at a moment's notice. But we find that people over trust in those kinds of systems and then are not in a position to be able to take over or the abuse it do things like hop in the back seat and take a nap. We've all seen the videos of the abuses. So I think really kind of educating people on where the technologies are, what some of the, I guess, deficiencies could be and that they're not a hundred percent trustworthy all the time right now is, is, is pretty key. And that includes everybody from, let's say like the driving public, but also the people at the DOTs who are potentially purchasing and installing these pieces of equipment, they have to kind of look at it with a bit of a scrutiny, right? And a little bit of an eye to try to find where they are not working. And that is squarely where we can come in and help. And we have helped communities such as the Virginia DOT. We work with Alexandria up in Northern Virginia and Falls Church up in Northern Virginia. This is exactly the type of work we do for them to help educate them about what the capabilities really are.

Travis

That's awesome. Well, you mentioned a little bit about the concerns, but you also mentioned the word opportunity. So I'm curious, what is the opportunity that most excites you in this area?

Mike

Yeah, I think it has to come down to me to improving safety, right? We at Virginia Tech Transportation Institute are very focused in on safety and that's one of our number one missions. And so if we have the opportunity to save just one life, you know, that, that is a, you know, by doing something like holding a light so that a red light violator doesn't come, you know, come through and hit somebody else. I think, you know, the potential there to really work towards what has been deemed a vision zero, which is a future we envision where there are zero crashes, right? To be able to get meaningful steps in that direction, I think that's the real opportunity. And technology can play a significant role in that as long as it's applied correctly, right? We don't want to cause harm either by, as I mentioned, over-trusting or trying to use a technology for a use case that it's not appropriate for.

Travis

Yeah. Well, I think that sounds great. And I think that it would be hard to argue against that, that vision and heading towards that vision.

Mike

Yeah. Well, there's a lot of other visions out there that we hope I'm a little less excited about things like. Well, there, there has been a model proposed by some like a, prioritization scheme. So if I'm a certain type of a driver, and if I want to pay a little bit more on a license, I might get a little bit of priority at the next traffic light. I I'm hoping those things don't really come to fruition and I don't think they will necessarily, but you know, we want to make sure that we don't open up opportunity to have the haves have more in the. The less have less, right? So, so like we tried.

Travis

So like green lights, green lights for money is what that would be Something like that?

Mike

yeah. mean, effectively we're already doing that with transit. So, you know, there's a technology called Transit Signal Priority that looks, you know, it evaluates buses and if they are behind schedule, it will hold a green light for them to get through the light so that they can try to get back on schedule again. This is all about improving the reliability of that transportation mode. And I think everybody's willing to...give them that priority because it's a mass transit solution. It's something that we're trying to use to improve the efficiency of our overall system. So they're willing to provide that. You could do the same thing with freight, right? Because it costs a lot more to slow down and stop a class eight heavy truck, right? And get it going again, as opposed to maybe having a couple of cars sit a little bit longer while a truck goes through. So all of these are being evaluated and tested. I really hope we don't see it based on, you know, somebody who wants to pay a little bit more, but you know, when you have Traffic environments like up in Northern Virginia, you've already got express lanes, right? Where you can pay more and have a reasonably free flowing lane of traffic out of town, right? Rather than being sitting in bumper to bumper traffic on a ⁓ regular general purpose lane. And so, you know, there's a bit of a precedent for it, but we'll see how it works in the end, you know?

Travis

Yeah. Well that, that is fascinating and I'm excited to see how it develops and I'm glad that you all are doing what you all are doing so that we can be at the forefront as it develops.

Mike

It's exciting times and it's a really interesting time to be in transportation and there are probably more problems out there to solve than we can even count up and think through. Hopefully we'll make a, again, our mission is to make a dent in it and make sure that we're delivering something good for those who are sponsoring and funding us to do the work.

(music)

Travis

Thanks to Mike for helping us better understand smart mobility and the future of transportation. If you or someone you know would make for a great curious conversation, email me at traviskw at vt.edu. I'm Travis Williams, and this has been Virginia Tech's Curious Conversations.

(music)