The Future of Intelligent Vehicle Interiors: Building Trust With HMI & AI - Ep. 194

0:10Welcome back to the AI Podcast. I'm your host, Katie Burke Washabaugh, and today we're taking an inside look at the future of automotive technology. On average, Americans spend more than 10 hours a week in the car. As autonomous driving continues to advance and attention is freed up, we'll be looking to intelligent vehicle interiors for ways to spend that time and communicate with the car. The industry has some ideas to help you bide your time on the road, including movie theater-like features and video game streaming. Just as important, however, is how these interiors build trust with their passengers.

0:48Joining me today is Andreas Binner, Chief Technology Officer of Rightware, a company developing the future of in-vehicle human-machine interfaces, or HMI. Their platform, Konzi, works with NVIDIA Drive IX to provide automakers a complete tool chain for building customized vehicle interiors for the next generation of transportation, including detailed visualizations of the car's AI. Hello, Andreas. Welcome to the podcast. How are you today? Thank you, Katie. I'm fine. Thank you for having me on the show. Yeah, well, we're excited to have you. This topic of trust when it comes to autonomous vehicles is so important, and there are so many different facets to it, which is code for me saying I'm going to be asking you a lot of questions.

1:35But to start off easy, you have a long resume in both the automotive and tech industries. What drew you to automotive technology in the first place? And how did you end up specifically in HMI? Yeah, actually, I ended up in automotive during the financial crisis of 2009, more or less by accident. I was working more in the chip industry before, and then I wound up at BMW as a software architect in the infotainment area. And funny enough, it was a Linux-based head unit based on an NVIDIA Tegra at the time. It was the first one, I believe. So, and from my background, infotainment seemed to be a pretty good fit for me, having a background in consumer electronic software.

2:25and to be honest, I always had a little bit of a crush of good user interfaces in general. So that's how I ended up in automotive with my first shop with BMW. So the first car that I ever drove was a 1999 Toyota Avalon and thinking about what infotainment looked like back then versus what we have today seems light years apart. In your experience, how has infotainment and the way the industry approaches infotainment changed over the past decade? Well, I think, yeah, 1999 and also after that, I mean, things were a lot centered around radio and music from physical storage devices. That was a thing.

3:13And maybe the most complex application that showed up was then the GPS navigation. And I think that's also what HMI took the first baby steps in map rendering and maybe towards 3D. But it also required more powerful SLCs like the ones from NVIDIA at the time with powerful GPUs. I think that's where it was an enabler. and also I think with the smartphone in general, people got more and more in interacting with devices much more than just turning on the radio. And looking today where we ended up in this immersive multi-display environment that includes the rear seat and the customer, so only the driver.

4:04And one other important thing that came in was the internet connectivity at one point that opened up a whole other world of infotainment and also HMI needed. And then probably then the final step was ecosystem integration, talking like CarPlay and Android Auto, which OEMs didn't like at the beginning letting them in. But I think it is basically requested by customers. I think a lot is driven by what the customers want in their car. And also a lot came from the complexity of the car functions that HMI suddenly became an essential part. Rear-view cameras, 360 surround views and things like that just required much more than just have an indicator light for Blinker to show the driver what's going on.

5:05And I think that is roughly a quick light of what happened here. It's technology and it's also what customers want to do in their car. Yeah, I use CarPlay every single day. I absolutely love it. I would be very sad to see it go. Are there any trends in HMI that have taken off that have surprised you or vice versa trends you expected to see but never ended up actually happening? I'm surprised. I put it, I never thought maybe 15 years ago that we have that many displays in a car, that it really took over. And also that it has more and more replacing physical controls, which not everybody likes. And I have to think maybe there's a little bit overdone in some places.

5:55So, I mean, some go back to physical. So that was a surprise to me a little bit. What never happened, that maybe I'm still waiting for the full windshield AR display. I mean, head-up display still is a small window. You look through. At CES this year, you've seen prototypes that go a little farther. That would be my dream, that the windshield is a full AR display. Yeah, I feel like I definitely associate that with CES, those cockpit setups with the full AR. That would be very cool to see on the road. How has AI influenced HMI design and what are some novel features that have been enabled by AI?

6:43I mean, the first thing that comes to mind, also in the car, is assistance. And unfortunately, how crappy they were at the beginning when OEMs tried to do it. There are funny videos still on YouTube, I think, that people try to get the navigation destination and it's just not working. And at the same time, fear the end, there was a system came. I think that is, for most of us, the first time we realize something towards AI and how helpful that can be. I would say the next, well, contextual aware functionality, I would call it. I mean, that is where maybe AI shows not that much in the car yet. Sometimes they predict where you typically go on Sunday morning or whatever, or on a Monday.

7:38But I think there is a lot of leeway to go for having more contextual aware. and helping the driver at the right time, the right situation, whatever is needed. But that's all I think in the early phases. So I think there's a long way to go for AI in the car, to be honest. I mean, JetGPT and all that stuff is just exploding currently. That's still outside the car, but I think that will change pretty soon. Speaking of these new technologies, we mentioned this at the top of the podcast, But I want to circle back and really kind of dive into this idea of why is trust so important as these new technologies start to roll out to consumers?

8:25I think that's a very simple answer. In a car, human life is at stake. So we have machines controlling and softwares controlling all parts in the car. And so we need to trust the technology not to fail in critical situations. And I think that's basically it. I mean, if the smartphone is crashing, well, it's maybe annoying, but in a car, it's a different story. What role does HMI play in building trust between the vehicle and its passengers? You know, we hear a lot talk about autonomous driving up to level five. And it's no secret that probably takes longer than we all thought. Also that the company thought.

9:14And as long as we haven't reached level five, that is a need for the human supervising the machine, right? I mean, so in these levels, we have two and three, maybe currently. It actually is by law that you need to look after what the car is doing. You cannot go for sleep. And for that, obviously, how do I do that? So you need some form of human machine interface to tell the driver what's going on, to give them more means than just looking on the street or suddenly the car is turning. Maybe you want to know that before what's happening. So we need some form of audio and visual human machine interface so a human can mount into the hardware and the software because he can't look into it.

10:07So is this relationship just important for the driver who kind of has these varying levels of responsibility of supervision over the car, or does it extend to all the passengers in the vehicle? I think at the beginning, it's mostly for, it's my opinion, for the driver. And I still think even if the car is level five autonomous driving, so it's perfect. Let's assume that's possible. I still think we need some form of HMI for quite some time to give driver and passenger the comfort that everything is okay. I think there's an interesting story at the early days of elevators. there was kind of the lift operator or elevator operator.

10:53It was at the beginning that it was actually needed to operate that. But then it stayed in there, even the time when it was perfectly okay, you enter the elevator and push the button to give the people the comfort that everything is okay. I think it will take quite a while that we have some HMI for driver and passenger to comfort them. even the machine is presumably perfect. So, yeah. Yeah, so this idea of communicating with the driver and passengers, could you talk about autonomous vehicle visualization and what role that plays in that communication? Yeah, I think autonomous driving cars are basing their decisions, what they're doing.

11:41Not only in complex software, but a lot of sensors around. There's video and radar. and ultrasonic and LiDAR, and all that is fused together to this environment model. This is how it typically works. And I think the most common visualization is what's called sensor or confidence view, depending on who you talk to. Most people know it's from Teslas because they had it from the start. It was an integral part that you see in a screeny way what the car is seeing. You see the other vehicles, you see it's a truck, it's a passenger car and all that stuff. It's getting better. And I think that is probably the most obvious visualization we will see in many cars because it is telling you what the car is seeing.

12:32So on the other side, maybe there's an opportunity to help and entertain the jobless driver at that point. And that's probably also opening up a new world of visualization. So I think it probably is split in something to monitor the car and also some opportunity to do something else useful in the car. Autonomous vehicles have so many different sensors and sensor types. You have LiDAR, camera, radar, taking in so much information every single second the car is driving. How do we decide what information to show to passengers to let them know what's happening, but also making it easy to understand at a glance?

13:18Well, we kind of define here term functional value, right? Where we think, so HMI and visualization isn't just, just look good. We also need to think what information has a functional value. So what do I, how do I reduce the information? So for the particular function the car is currently doing, what does the driver really need? So it's about reduction and of course great design. You need great designers for that. But if you look at the surround view, it's really much reduced. So you have a 3D car that everybody understands. You have this bird's eye view. So it is pretty much reduced with a lot of computation and stitching together that it looks like you're looking for the top of your car and everybody understands that immediately.

14:12And it has an immediate value because you see the obstacles. So I think I would say is the trick to focus what has at that situation a functional value to the driver and just abstract that away. Otherwise, it's overwhelming. Also, maybe if you have screen capabilities, you might try to overdo it here and there and have too many effects. Maybe you need to take that down a little bit in that areas. We've talked about the proliferation of screens in the car a few times. As the size and number of screens in the car continues to grow, how can automakers ensure that the driver stays focused on the road in times when they are in control of the vehicle?

15:05Again, it comes a little bit to what I said before, that at that time, And we really have to limit what you visualize to get the driver focused. But of course, I think there is also new technology that automakers could incorporate and sometimes have. You could have eye tracking technologies. You could want for distraction, but you also could potentially, if you have multiple displays, track where the driver is looking. and also kind of context aware put the important warning where he's currently looking. So there might be a couple of things. Also, AI could help here maybe to, again, to present contextual relevant information for the particular situation.

15:54So here could, AI could be very helpful. So certain information might not be critical in that situation than something else. and then also change the HMI to reflect that. So I think new technology, AI or eye tracking, other things could help to get the focus back or at least one if the focus is not there. Aside from what we've already talked about, what are or are there other features that could help build trust between the vehicle and passengers? It might sound stupid or obvious, but I think the quality of these active functions need to be flawless. If driving assistance fails often, it will reduce the trust.

16:50In the same manner, false warnings in otherwise regular situations also reduce trust. So I think it's not that much about HMI. I think almost, and everybody knows that, human gain quickly trust. At least I did when I first had an ACC. I got quickly trust that the thing is holding the distance and didn't question it. But if it's one, two, three times a drive suddenly is doing stupid things, this trust is quickly collapsing. So I think the most important thing is to have flawless working functions in the car for the trust. from an HMI point of view, I mean, it might sound even contradictory, but you may be offering the passengers some more leisure activities, like watching a movie or playing a game that could even give them the feeling that everything is okay and they can trust the machine, otherwise they might not even offer that to you.

17:47So yeah, it might sound contradictory, but I think that could also be a way that you convey trust that, you know, you are offered with special things to do. the car. Yeah, it's the age-old adage, actions speak louder than words when it comes to building trust. So final question, what are your predictions for intelligent interiors over the next decade? It's difficult because I think I was wrong there in the past, but I try. I think, as I said before, more contextual aware, the right information at the right time. Maybe we have to peak with so many displays. Maybe I also can see that it's scaling down, or at least high displays more.

18:35We have to assume that CES, but certain displays kind of fold down into the cockpit, you don't see it and fold up if you need it. Also, I think more aesthetically pleasing visualizations. I think we're much better here, but there's also a way to go. So I think maybe other means of user in and output that could be, as I said, full screen visualization. So other display technologies I could see in the next decades. User input, I'm not sure we have voice and touch and everything, but maybe there's something else also coming. Also, I would wish, at least, it's more or less a prediction that we find the right balance between physical and tactile UI and virtual controls.

19:22I think that has gone a little bit overboard the past few years. So, yeah, that's probably summing up, but maybe I'm not the most visionary guy when it comes to it. I'm too much of an engineer, but that would be my prediction and wishes. I don't know. It sounds like we'll be coming a long way from 1999. So that all sounds great to me. Well, thank you so much for joining us today, Andreas. this was a really enlightening conversation. Yeah, thank you again for having me here. And yeah, anytime again in the future. All right, until next time. Thank you.

20:46¶¶

20:51Thank you.