Hi I Wanted to talk about this recent incident with a fatality. the first fatality involving an autonomous car. In this case, it's an autonomous Uber car in Tempe Arizona in the US and unfortunately there was a fatality. Someone was crossing the road and they got hit by this autonomous uber car.

There was a driver behind the wheel, but they weren't in control of the car at the time. It was in fully autonomous mode apparently, and it's the first accident of its kind. It could have our wide reaching ramifications for, well, at least the near-term future of autonomous cars. So I thought it's important to have a look at it and talk about it.

and the Tempe Police have just released the footage from inside the car of the accident. They just released it like an hour ago. Which really raises an interesting point because before this it was just all speculation about well, did the person just dash out in front or whatever could the autonomous car? it should it have detected this and avoided it and I'll show the footage are shortly. But let's have a look at the original Y news report from this shall we? BC 15 Big questions about this early morning crash in Tempe A bicyclist seriously hurt in a crash with a self-driving Buber vehicle on Mill Avenue near Curry.

We don't know right now whether there was a human driver at the controls at the time of the crash or not. We also don't know the condition of the bicyclist. It appears the injuries are serious. Tempe PD Just letting us know that Mill Avenue has reopened.

This is just north of Tempe Town Lake Now unfortunately, as I said, the pedestrian did actually died in this incident and there's a huge investigation into this. so I'm sure that you know the full details will eventually come out. But it was very interesting that very soon after the incident, the Tempe police chief came out and says an early probe shows there were no. There was actually no fault by Uber.

From viewing the videos, it is very clear it would have been difficult to avoid this collision in any kind of mode. Otama autonomous or human driven based on how she came from the shadows right in to the roadway. More? Who's the police chief said they have not released the video, but now they have just an hour ago and we're able to take a look at this to see how bad it is Now of course, autonomous cars. They shouldn't just have visual based systems because as they said they're if they've come out of the shadows, you can't see them.

Oh, this is a very typical of pedestrian and other you know, animal accidents and things like that. When you're driving at night they jump out of the shadows and slam. You know you don't have time to see them come in or to stop. But of course autonomous cars are expected to be better than this are expected to have.

You know lidar based radar type bar systems that are supposed to You know, ping out in front of the car in or or in all directions essentially and know where objects and other things are. They shouldn't just rely on visual aids. And here's the actual world. Volvo We're talking about the Uber one here with their this is Ubers own driverless package.

It's not actually a Volvo thing is that's got a top-mounted lidar unit provides 360-degree three-dimensional scan of the environment. so that's actually a rotating laser that spins around and then they receive the data from that, they can build up a lidar map image of that, a forward-facing camera. a way to focus on both close and far field watching for braking, vehicles, crossing, pedestrians, traffic lights, and signage. So like the problem with just if you've just designed an autonomous car that used optical like, camera only based art systems, they're just not as good as the AI.

There's a lot of stuff that goes into the technicalities of this, but generally the dynamic range of the eye is better. It's much better up at peak in, you know, low light. Do you know little glints and movements and other stuff then? especially a video camera based system, But of course it needs. That's why this Uber car has all these other sensor packages as well.

Like this: 360-degree radar coverage so not only lighter, but radar as well. Where those sensors are I don't know, but they're kind of showing them on the front here. a roof-mounted GPS of course, side and rear faces and cameras working collaboration to construct a continuous view of the vehicles surroundings. So it's seven cameras, one laser, inertial measurement units, custom computer and data store at 360 degree radar coverage.

Something obviously Rhett went wrong there with the algorithm or whatever. This stuff has more than enough technical capability to see somebody pushing a bike in front of the car. Unfortunately, something went wrong. Now from the news footage, I've been able to determine the exact location where this happened.

It was just over the Tempe Town Lake here, just past this overhead a bridge right about here on North Mill Avenue just before East Currier the intersection with East Curry Road Here and if we actually go in and have a look at the street view down here, we can see as you'll see in the footage very shortly, there's a lamppost right here. This lamppost here provides the light coming across where the pedestrian comes out of the shadow here. So the impact happened like just right about here and the bike ended up some way up here just past this blue sign here with the palm trees as you'll be able to see in that news footage before. So the bike ended up here, the impact was down there.

and what's actually sad about this is this is actually a bike lane going along. Here there you go and further up here. the person could have actually crossed up at the lights up here. It is a long way, but hey, look, let's not put blame on anyone for cross in anywhere or doing whatever we're just talking about.

The technicalities of the should the self-driving car have actually detected the person crossing across like this in this case with their bike going across the road. But with the lasers and the lidar and the radar and everything else is along with the 20 cameras on the car, should it have been able to detect such a simple thing as something coming across in front? First let's take a look at uber self-driving car the one involved in the accident and the modern ones that they're using are a Volvo Xc90 I Believe they started out using a Ford Focus test cars like this, but have a look here. The cars themselves were packed with around 20 cameras, seven lasers, GPS radar and lidar technology that measures the distance reached by outgoing lasers so cars can see and interpret the action around them. so it's got all the bells and whistles.

It's got everything that should be required to detect somebody walking in front, even if they're coming out of the shadows. As the police chief said, and as we'll see in the video, this is certainly the case. A human probably would not have been able to avoid this accident I suspect based on the footage, which you'll see shortly, but it certainly had everything there that should have done the business. So and it was.

The car has been determined. It was in full autonomous mode at the time, so it looks like based on this it with all that technology, something went wrong. unfortunately and it should be noted that the one in the actual actual accident look, it's got the the scanning looks like there. that's probably the scanning laser on top, plus the cameras and everything else.

So it's got that full car with sensor package actually going across there, so it looks a bit more streamlined than you know. It's just like the package on the front, so it's obviously like a newer one than the full-on one that they actually had in these early prototypes. So they seemed to have like I improve that, but it should have the full sensor suite in there now. I'm going to point out this interesting article that I found and I will link it in down below: it is a year old from March 24 2017 but and it's from recode inside Ubers self-driving car mess.

Apparently I won't go through the whole thing, but of course Ubers future pretty much depends on self-driving cars because they're not doing very well financially. They know the future self-driving cars, and if they can get the drivers out of the loop and have a fully autonomous service, it's going to probably have a much better return fall. And so they're betting big time on autonomous cars, as they probably should know. I Blame Sons for that.

But anyway. apparently. oh, like a year ago. Granted, this was a year ago had they said here that the companies are autonomous efforts are in turmoil according to exclusive interviews they had with current and former and current employees of their self-driving unit, which they got.

They bought a truck self-driving truck company called Auto and that's where a good lot of the talent came from, but turmoil, internal tension among its executive leadership. Granted, this was a year ago, issues, a wave of key talent departures, and problematic demos. At least 20 of the company's engineers have quit since November that would have been 2016 at the time anyway. I'll link it in down below and you can read it for yourself and make up your own mind in terms of that.

But yeah, it wasn't looking good back then their autonomous car division. They did have issues, so here we go. Let's take a look at the footage. the Tempe Police Department have just released this.

It is public and I have actually taken this footage. It's done as two separate footages, one from the dash cam looking presumably the dash cam looking for it or it is that the cameras one of the cameras up on the top. I'm not sure, but anyway, it's looking forward out of the car and there's the other one looking back in at the driver as well. and what I've done is actually taken this footage.

I've inlaid the driver with the front forward-looking dash cam footage I've tried to line them up as much as possible. The police chief said that actually the first time the driver knew about it was when the impact actually happened. so that's when over synced it up. So warning, some viewers may find the footage disturbing.

So as you can clearly see here, basically the pedestrian walking their bike going across the road did actually come out of the shadows there. And granted, this is a very common way that accidents happen day at night, pedestrian animal accidents and things like that. especially here in Australia no kangaroo jumps across the road, they come out of the shadows. there's almost no way to see them and respond in time.

So even if there was a driver like behind the wheel, actually, they may not have been able to prevent this accident unfortunately. but that's not where what we're discussing here. It's like sure this autonomous car have actually detected this with all its sensors and really looking at this footage. I've got to say I mean this is a classic case, wasn't just a person, it was a person with a bike So you know it was like a it's a pretty big target to miss I would say so you've got a I guess question and once again not necessarily putting blame on Uber here or there technology.

but let's look at this. A person was walking straight across autonomous cars as I said I think we expect better from these because of all the sensor fusion technology. the later, the lasers, the lidar, the radar, and the 20 cameras on this thing. Granted I don't think the cameras could have actually picked up that but there I Can't see how the lidar or other systems should not have picked this up.

So do they have an issue with their sensor fusion technology and stuff like that? Granted, these are still essentially prototypes that they got out there driving around on these roads. So I'm like I don't think the driver could have actually seen the pedestrian coming out of the shadows here. I So yeah, I I Can't see how you can necessarily blame the driver in this case, but as you can see, he was actually distracted. probably maybe looking at a mobile phone or something like that.

but in this case, the sensors really should have picked that up. I think so that's a like a classic case of something walking in front of a car and that's a reasonably big target. So anyway, there you go. That's the footage.

Let me know what you think should this uber self-driving car have picked this up? Do they do? you know about this all this sense of fusion technology? If you've got any further technical details on exactly how the Uber system working and actually fuses all these different sensors together, does it process them separately? Does it do fuse them together and then do it based on the final output of that, please leave it in the comments and as always evey blog forum down below. But if you want my opinion on these are self-driving cars and a lot of people have asked and we've discussed it on the amp hour a few times. All the talk in the last couple of years is fully autonomous cars are only a couple of years away. Like that's what everyone was saying and even before this accident I was saying no way that they aren't even close to being able to do what humans can do.

Yes, in limited circumstances geo-fenced areas with fully mapped modes. but they I can't see how the technology at this stage can detect stuff that humans can detect. just basic stuff you know, like a big city like Sydney different work zones and Contra flows and you know, lollipop people were standing there. The stop go and like all the like let alone navigating car parks and navigating.

You know things like that. Looking for parking spots and stuff like that? you know? can the cameras see other human faces over there and see them walking towards a car? They're walking towards their car. Looks like they're going to leave. You know I can wait for a spot there and things like that.

How like there's countless different scenarios I can think of ones I Can't think of that if you've got them. put them in down below where current autonomous technology, no matter how good it is, is not going to be able to replace humans any time soon. except in specific. You know you know narrow sort of circumstances, so it's a very interesting question.

this autonomous car thing. it's going to happen. But I don't think it's going to happen on as large a scale. And as soon as people think, comments down below catch you next time.