Last year was a buzzworthy year for autonomous vehicles. The pedestrian death in Arizona raised concerns over safety while many AV OEMs slowed down production of advanced levels of automation due to regulatory issues. Meanwhile, China continues to push forward, faster than ever, in an AV race for commercial deployment.
In addition, press headlines are not in agreement with analyst projections. Promises that driverless vehicles will be on the road this year are in sharp contrast with analysts who say AVs have entered the “trough of disillusionment” with anywhere from five to fifteen years before they will be commercially available.
In episode 1 of Tech Lightning Rounds, Beth Kindig of Intertrust Technologies goes directly to the source of AV expertise and hosts discussions with technologists who specialize in the field. Beth holds discussions in “lightning round” format, which are rapid interviews for immediate depth on the topic.
Jill Sciarappo, Director of Autonomous Driving at Intel, is candid as she discusses areas where AV companies can improve on safety and how a new coalition called PAVE can help. PAVE stands for the Partners of Automated Vehicle Education initiative, which was launched at CES 2019 in Las Vegas, to help bring “realistic, factual information to policymakers and the public.”
Jill and Beth also chat about the confusion consumers have around AV and why press headlines need to be a little more accountable. Jill is from Arizona, where a lot of AV testing occurs, and she = points out that the vehicles will have to undergo best practices and validation before they’re allowed on streets, drawing parallels to seatbelts and drivers licenses. She also discusses why going open source with mapping infrastructure is the future for autonomous vehicles as the data will be critical for progress in safety.
The second guest, Michael Fleming CEO of Torc.AI, has been developing autonomous vehicles for twelve years, which is a very long time for this particular technology. Torc placed third in the 2007 DARPA Robotics Challenge and has since developed AV technologies for the military and mining industries. Michael discusses some of the history around AV and how it was initially developed to avoid military fatalities from roadside bombs.
Because Michael has been developing the technology for quite some time, Beth also asks him how AVs can improve on safety from a technical standpoint. Michael discusses how classifying objects will improve AV response and how to account for pedestrians doing things that they shouldn’t, such as stepping in front of a car at the last minute.
Anuja Sonalker, CEO of STEER, is the third guest on the episode. Anuja has a PhD in network security and held executive-level security positions before founding the self-parking technology STEER, which is designed to let your car park itself similar to a valet. She discusses the importance of security-led design in order to have “intrusion and fault tolerance distributed in the architecture” in order to avoid falsified data attacks.
Anuja and Beth also discuss the complete security life cycle from conception to implementation. Anuja describes cybersecurity as an artform, encouraging security professionals to think both like an attacker and like a defender.
Please Subscribe and Leave a Review for the Podcast Here
The podcast is now available on the following podcast directories: iTunes, Google Podcasts, Google Play, Spotify, Stitcher, TuneIn, iHeart Radio, YouTube, Soundcloud, Digital Podcast, iPodder, Podcastpedia, Soundcloud, Acast, Amazon Web Services, Cast Box, Bluebrry, Spreaker, Double Twist, Downcast, Libsyn, Listen Notes, Otto Radio, Overcast, Player, Pocket Casts, Pod Cruncher, Podbean, Podcast Addict, Podcast Blaster, Podcast Republic, Podcasts in Color, Podiant, Podkicker, Podomatic, and Radio Public,
TECH LIGHTNING ROUNDS EPISODE 1: TECH EXPERTS DISCUSS THE FUTURE OF AVs
00:01 Intro: This episode is brought to you by Intertrust Secure Systems. Today’s hackers have changed the security threat paradigm and forced companies to protect both enterprise infrastructure and customer data. Intertrust Secure Systems provides application security to protect software applications, mobile apps and IoT devices. Our solutions shield applications against reverse engineering, tampering and unwanted modifications. Go to intertrust.com for a trial.
00:34 Beth Kindig: Welcome to Tech Lightning Rounds, I’m your host, Beth Kindig. This podcast interviews key people with deep expertise on one topic for a 360-degree view. One difference between this podcast and the other podcasts you listen to is that I hold short interviews called lightning rounds, with the goal of giving you a lot of compelling information very quickly so you can get on with your day. In this episode, I went to CES and spoke with three experts on the topic of autonomous vehicles. The people I interview come from Intel, Torc and a company named STEER. I asked probing questions about how to improve autonomous vehicle safety after the pedestrian death in Arizona, whether China will be the first to commercially deploy driverless vehicles, whether driverless is really just a bunch of hype. Or will they become in reality in the near future?
01:23 Jill Sciarappo: What level of best practices or validation do we need to have these companies prove they can accomplish before we let them on our city streets?
01:37 Michael Fleming: Some of the reasons that mining and defense were early adopters of this technology was, well, on the defense side when we were at war about a third of our war fighters were dying from IEDs or roadside bombs.
01:48: Anuja Sonalker: And cyber security is an art, also it’s a technology, it’s a capability. Cyber security is something that you have to be groomed for… It is a mindset. You have to think like an attacker and you have to think like a defender.
INTERVIEW ONE: JILL SCIARAPPO OF INTEL
02:00 BK: My first guest is Jill Sciarappo, Senior Director of Strategy and Marketing for Autonomous Driving at Intel. My interview with Jill is candid as she discussed areas where AV companies could improve on safety and how a new coalition called PAVE can help. We also chat about the confusion consumers have around AV and why press headlines need to be a little more accountable.
02:23 BK: Right now, we’re at a level two, partial automation on our roads today. What do you think is the future of autonomous vehicles?
02:36 JS: I think that companies are gonna continue to invest and evolve the technology that gets us to an autonomous vehicle space. I think there are several things that we need to take in account along the way. Number one, first and foremost, is always safety, and safety shouldn’t be a differentiating factor, right? If you get in autonomous vehicle A, it should be just as safe as autonomous vehicle B. So I think that in the next decade or so we’re gonna see some standards start to be established around what it means to be safe in an autonomous vehicle, any vehicle, as we continue to add technology, much like seat belts, anti-lock brakes and airbags have gone through. It took them decades, though, right? Those three technologies took decades from invention to standardization. We’re gonna see the same thing with autonomous vehicles. But additionally, we’re gonna see lots of companies continue to add their own thumb print on their solutions, just like OEMs do today, which is gonna be great.
03:34 JS: I also think that we’re gonna see some consolidation and standardization in mapping infrastructure. It takes a lot of driving and data collection to create accurate maps of our infrastructure, and that really shouldn’t be as proprietary as is today. We should all share maps. We should all share incidents we see on the road, so that collectively we’re making everyone safer on the roads.
04:02 BK: Do you think we’ll continue to improve the experience for human drivers for some time?
04:06 JS: Oh, 100%.
04:07 BK: Before we move into a driverless vehicle?
04:10 JS: Yes. We love that question, because one of the things we’ve really seen over the last two years is the notion of pulling research and development for AVs into the ADAS features available in cars today, and that’s good for everybody, because we’re using future research and making it pragmatically available for cars you can buy today and in the next five years. The BMW in our booth today is a very good example of that with the tri-focal camera. It’s one of the most advanced ADAS systems you can purchase, and it’s good for all of us if we continue to see those technologies come into the vehicles. For example, forward collision warning itself has been proven to reduce your probability of being in an accident by over 25%, just that forward collision warning, that beep if you’re gonna rear end somebody. That kind of technology is amazing when it comes to what we can do in vehicles we can buy today.
05:09 BK: There was a survey done, and I blogged about this, where they went out and they surveyed a bunch of consumers, and mistakenly, they believe that there are a lot of driverless vehicles on the road right now.
05:20 JS: Yes.
05:20 BK: There seems to be a lot of publicity around these Arizona testing sites, etcetera. Why is public perception not matching where vehicles are?
05:32 JS: Well, there are probably a couple of reasons for that. My mom can’t open the paper without reading about autonomous vehicles, and most of the things in the paper about autonomous vehicles are articles written by the companies doing autonomous vehicles, right?
05:46 BK: Right.
05:46 JS: Everybody wants to win. It’s become a big race to the Moon. So sometimes what’s published is not fully matched with reality. Or not everything is shared. I firmly believe making sure that public perception and understanding of this technology is imperative. We have to get people to trust, understand and respect the technology, even in our ADAS situation, right? Reports of it delivering Level 1 or Level 2, people are taking advantage of it and taking their hands off the wheel and going to sleep and not paying attention, it could severely hurt this industry. So while people believe there are cars running around without a driver in them, on the other hand, it’s our job as a marketing community and a technical community to educate the public on what’s really happening, where they should be looking for this technology when they go to purchase a car, and what they should be looking forward to. I think this is a public education issue, and in fact, a new coalition was established this week. It’s yes, called, PAVE, which is all about educating the public on the technologies that are coming out for vehicles.
07:02 JS: In fact, I’m from Arizona, so the State of Arizona is really closely looking at what level of best practices or validation do we need to have these companies prove they can accomplish before we let them on our city streets to test, because Arizona has been very good about it, but now there are a lot of cars testing in the state, and so one of the things I think the whole industry needs to look at is, what is that minimum level of validation a car needs to prove it can handle before we let it out on public roads. This is where RSS comes in. RSS can be an element to that model to say at any point, is this car always keeping itself safe?
07:49 BK: Intel has a partnership with Baidu, so I took the opportunity to ask Jill about China and how likely it is China will be the first to commercially deploy autonomous vehicles. We also discussed Mobileye, a very large acquisition Intel made in 2017 with a price tag of over $15 billion, which helped Intel become a leader in the AV space.
08:10 BK: At this event I went up and I visited the Baidu booth and I was in one of these Apollo vehicles, and they were kind of showing me around a simulator. What are you doing with Baidu and do you think China is gonna be the first state really to hit the road with more autonomous features?
08:24 JS: Yeah. And so China has always been really good about delivering new technology. They’ve always been right there at the forefront, right there with the leading technology and Baidu’s been a great partner of ours. Over the last year we’ve announced that they are going to be adopting our responsibility sense of safety model into their project Apollo platform, which is fabulous. This is a way that Mobileye has put out a mathematical formula to measure and verify that a vehicle is always keeping itself safe. So although your policy, and the intelligence in the car may be making decisions, this is a way to check to make sure, is that car making the right decision? And so, Baidu has partnered with us to put this into their project Apollo. We put this out there as an open mathematical model, an open concept, and we’ve called to the entire industry to join us in coming together to define what those safety standards should look like.
09:30 BK: A little earlier you had mentioned Mobileye. Just for the listeners who aren’t completely aware, you had a big acquisition. What is Mobileye, and what do you see Intel doing with that acquisition?
09:41 JS: So Mobileye is the world leader in vision-based technology for vehicles. They’ve been around for 20 years, and they’ve been working on technology that goes into vehicles to make them safer for 20 years, first with the cameras and the warnings, and then they actually got into cameras and actuation. Those are the baby steps, the derivatives that lead to autonomous vehicles.
INTERVIEW TWO: MICHAEL FLEMING OF TORC.AI
10:09 BK: My second guest is Michael Fleming, CEO and Co-founder of Torc, who has been working with AV software for over a decade, which is a very long time for this particular technology. In this interview Michael shares an interesting story about how autonomous vehicles got started with military and defense. We also discussed Torc’s partnership with Caterpillar, one of the only companies currently operating driverless machinery.
10:30 MF: So I’m Michael Fleming, I’m the CEO and co-founder of Torc.
10:35 BK: Can you give us a baseline? How does Torc compare to Waymo or GM Cruise?
10:39 MF: We provide just the software stack. We work with OEMs and service providers to integrate our software stack on their fleet of vehicles which are deployed in different environments. So it’s the same business plan that we’ve used in some of the early adopter markets, such as the defense space, the mining space and we’re doing the same thing in the mobility and transportation space.
11:02 BK: Regarding defense and mining, we hear a lot about consumer autonomous vehicles. Has defense and mining been an easier market entry? And why?
11:12 MF: So, Torc started about 12 years ago from the DARPA Urban Challenge. So we’ve partnered with Virginia Tech, and we were one of three teams to cross the finish line. We were a little bit ahead of our time, if you will, from a transportation side of things, but the early market adopters were mining and defense.
11:30 MF: So we gravitated towards getting our technology out sooner rather than later. And it’s kind of interesting being in the space where for over 12 years, we’ve been waiting for the transportation and mobility space to really take off. Now, some of the reasons that mining and defense were early adopters of this space, of this technology, was well, on the defense side, when we were at war, about a third of our war fighters were dying from IEDs or roadside bombs. So we integrated our technology with the services and defense OEMs to remove the war fighter from having to be inside the vehicle. So vehicles could autonomously go down a road, detect and neutralize roadside bombs. At the same time we partnered with Caterpillar, one of our longest-standing customers, to automate large surface mining trucks in Australia. And while you may walk through CES and see a lot of self-driving technology talking about what folks are going to do in the future, Caterpillar through a partnership with Torc and some other organizations, they’re running fleets of self-driving half-trucks without anyone inside the cab.
12:34 MF: Now, defense vehicles and mine vehicles are a little bit more expensive than the consumer car that you and I would buy, so they can justify a higher price point for autonomy. What we’re focused on now is driving down the cost of this self-driving technology, increasing the safety and the reliability, such that you and I and some of your audience can basically drive autonomously from home to work every day, other than getting frustrated in some of the gridlocked traffic, to free them up to maybe take a nap, talk on the phone, or catch up on some work or emails.
13:09 BK: I was going to ask you, there was a high-profile accident involving an autonomous vehicle in Arizona. How can AVs improve on safety?
13:18 MF: So as I mentioned, Torc has been in this space for about 12 years. Self-driving is a very difficult problem. It’s a very complex problem, But in reality, the software architecture is, think of it as hundreds of different software modules all being interconnected, which is pretty incredibly complex. Now, we’ve been working in this space for over 12 years, and these complex technologies do not come together in short order. And for that reason I think it’s important that the organizations take a slow and methodical approach to not only developing, but deploying self-driving vehicles. Having been part of Torc for the last 12 years, having run self-driving vehicles on public streets for over a decade, we’re accident-free. One of the reasons that we’ve done that, we’ve taken a very slow, responsible and methodical approach.
14:08 BK: Because Michael has been working with AV technology for such a long time, I thought I would ask him, how can autonomous vehicles avoid hitting pedestrians in the future from a more technical level? How does a vehicle tell the difference between inanimate objects and a pedestrian?
14:25 MF: Sure, that’s a great question. So I think earlier in the podcast I mentioned we like to bend our self-driving software stack into sense, think and act, and if we take our sense component, we can break it down into two sub-components. The first one being detecting the object and the second one being classifying the object, and we’ve been a big believer in three different modalities, since our success in the DARPA Urban Challenge: Vision, radar and also Lidar sensors. So those are pretty straightforward sensors to conduct the detection. The classification uses some advanced algorithms, deep learning, machine learning and so forth where we are able to detect a car as a car, a person as a person, or let’s say a light post as a light post. As we classify those objects, we have a little bit more understanding of how those objects move. Because you and I move, well, let’s say, two to three miles an hour and we can kinda move in any direction that we want. You don’t really see cars move laterally. You don’t really see light poles move. I guess you could probably come up with some extreme situation.
15:31 MF: So by classifying we have some context into how these objects can move. Now, we also like to overlay the map, if you will, or the environment. So if you think about it, if I’m a pedestrian and I’m walking down a sidewalk, more than likely, I’m going to continue to walk down a sidewalk. There may be a cross walk. More than likely, I will probably walk through that crosswalk. Now, some pedestrians cross that cross walk at the proper time, but we have a lot of rule breakers out there as well. And we also would have to take into account, What about that possibility of a pedestrian stepping out in front of the car at the last minute? And we’ve got some great video here at our booth at CES, where one of the edge cases that we’ve captured was we had a handful of customers outside of a restaurant in Blacksburg where a lady stepped in front of our self-driving car on the road without even looking at the oncoming traffic. So we, as self-driving technologists, have to account for you and I doing things that we shouldn’t.
16:40 BK: My third guest is Anuja Sonalker, CEO of STEER, who has created autonomous parking technology. Her company is setting out to solve one of the most frustrating parts of owning a vehicle with the concept of self-park. What is automated parking?
16:54 Anuja Sonalker: So automated parking simply means taking the human out of the equation in parking. So imagine, if you will, you drive yourself to wherever you’re going and now, instead of having to look for a spot to park your car, you literally get out at the curb and go and enjoy what experience you’ve come there to enjoy. Pull your phone out, press a button and your car knows where it needs to go to park, find an available spot, pull in and let you know, oh, I’m parked. So at all times you know where your car is, and you enjoy your experience. When you’re ready to leave, just summon your car back and your car comes to pick you up. It’s like having your own personal chauffeur, except you don’t have to pay anybody do it.
17:32 BK: What kind of time are you saving? Or what is the cost to benefit ratio there?
17:38 AS: So, there’s tremendous value in time savings and parking. We don’t normally realize this, but an average human spends about 20 minutes each time they’re trying to park their car, just trying to locate a spot, pull in, park and then walk to the mall or wherever they were trying to get to. On average, the commuter spends about 110 hours a year looking for parking. That’s 28 days of time that you didn’t give your family, you didn’t give yourself, you didn’t give your work. It’s a vacation that you could have taken that just went in parking.
18:20 BK: Oh, yeah, it’s a total waste of time. I’m from San Francisco so I’m sure it’s even more than 28 days.
18:25 AS: Absolutely.
18:25 BK: Anuja has a background in cyber security and lives in Maryland, which is the global hub for cyber security. I asked her to elaborate on autonomous vehicle security and what STEER, or any other company, can do to prevent hacks when a human driver is not present?
18:41 AS: We call it the principle of the security-led design, where you start from a minimum design with a minimum attack surface and maximum default security. What that means is, right from conception of the design, where the attack vectors are, where the threats can come from, security, safety, theft. All of these are considered. All the open connectivity elements are considered. The architecture is designed with security in mind so we have intrusion and fault tolerance distributed in the architecture from the sensor suite to the decision-making, to the actuation, everything is security-monitored and controlled. So there is cross correlation, we can avoid falsified data attacks. You may have heard media researchers have demonstrated attacks where when the incoming data is bad, so you fool the car, the car is bound to make decisions that are bad, because it’s relying on information that is bad. So even attacks like that, we’ve taken that into consideration in the life cycle. It’s a complete security life cycle from conception all the way to implementation.
20:00 BK: When you say you’re from cyber, you have a cyber security background, you’re actually from a really special region for cyber security, right, out in Maryland?
20:07 AS: Yes.
20:08 BK: Can you just, for some of our listeners, describe how much of a hub Maryland is actually for cyber security?
20:14 AS: Absolutely. So, big shout-out to Maryland. Maryland is the center for cyber security in the United States. Especially where I come from, Howard County, we are the epicenter of cyber security. Per capita we have more white hat hackers in Maryland per square foot than anywhere else in the country. And if I may say, anywhere else, even globally, such a strong workforce developed with cyber security principles. And cyber security is an art also. It’s a technology. It’s a capability. Cyber security is something that you have to be groomed… It is a mindset. You have to think like an attacker and you have to think like a defender. And an attacker’s job in some senses, is slightly easier than a defender’s. Because imagine a fish net, an attacker needs to just find one hole and get in and that’s his job is done. And hackers usually they are lazy creatures. They want to find the first easiest thing they can get in and they get in. A defender’s job, if you will, is to plug every hole on that fish net, and as you have to know and find and keep finding iteratively every possible and yet you would not know did you find every single hole in that fish net.
21:26 BK: Thanks for joining me for the autonomous vehicle episode. Be sure to check out future Tech Lightning Rounds on Robotics in 5G. Please also leave a review on iTunes to support the production of this podcast.
21:40 Outro: This episode was brought to you by Intertrust Secure Systems, a comprehensive solution to protect software applications. Visit intertrust.com for more information.
Intertrust Secure Systems blurb:
Intertrust Secure Systems is a leading provider of application shielding solutions to prevent hackers from reverse engineering and tampering with code. Our products are backed up by superlative support and professional services to help you achieve your business goals quickly and efficiently.
Our application shielding portfolio consists of two products:
- Code Protection: provides application developers with a comprehensive suite of anti-reverse engineering and runtime application security protections to help protect your applications.
- Secure Key Box: is an advanced white box cryptographic library that protects cryptographic keys for critical security functions such as device authentication, secure communications, and data encryption.
I consult for financial firms. Inquire here.
Sign Up to Receive Bi-monthly Insider Analysis:
I’m an industry insider who writes free in-depth analysis on public tech companies. This year, I predicted Facebook’s Q2 crash, Roku’s meteoric rise, Oracle’s slow decline, and more. Be industry-specific. Know more than the broader markets. Sign Up Now. I look forward to staying connected.
Originally published at beth.technology on February 15, 2019.