Levi

Expert
Messages
2,715
Reactions
3,135
Well, the video is shocking, and blaming anyone (system or victim) is not the right thing. If anyone is to blame, it is not the system, not necessarily the engineers, but the 'blind believers'.

The person was crossing a straight street, no curve, nothing was unpredictable (from what we see in the video). Radar/Lidar are supposed to see better than human and night in darkness, so in this case the system should have avoided the accident that a human would not, yet it failed.

Some might blame the victim, and say it would happen with human driver, but isn't the point of the system to save the from bad human decision (of the victim)? Should the pedestrian be also be replaced by autonomous humanoid? If she is to be blamed, then she is whether there is a human or system driving, in which case what is wrong with accidents when then victim is at fault?

Of course that changes when the crash of equivalent value, not car vs person but car vs car. There are many victims, those responsible (if you want to blame) and those not (of you want to take away blame, i.e the system in this case).

What if the autonomous car worshipers, are the 'bad drivers', thus want such cars?
 

spwolf

Expert
Messages
3,511
Reactions
3,439
Seeing is better than hearing. Now I'm convinced. It was an unacceptable system failure.

Uber is known to have "bad" leadership, and I mean that - you can read up on their fight vs lyft, corporate espionage and why was their past CEO ousted... it is comically do-evil corporation. (as side note I do use uber all the time, i love the service).

So it is not hard to tie the dots here and see how these rumors of how pedestrian jumped in front of the vehicle, from the bushes, etc, etc, were paid by Uber. Anyone on the scene of crime, be it reported or cop, would see it is a wide street, no distraction, person was crossing from opposite side so no bushes, even the street lights were few meters away.
 

spwolf

Expert
Messages
3,511
Reactions
3,439
Well, the video is shocking, and blaming anyone (system or victim) is not the right thing. If anyone is to blame, it is not the system, not necessarily the engineers, but the 'blind believers'.

The person was crossing a straight street, no curve, nothing was unpredictable (from what we see in the video). Radar/Lidar are supposed to see better than human and night in darkness, so in this case the system should have avoided the accident that a human would not, yet it failed.

I think system failure will clearly be blamed as well as test driver.. here you can check how these affordable systems work:

So the Volvo was going at 63kmh, and test above is 55kmh. I am sure factory Volvo system would do at least as good as Lexus one above, since it is already older version of Lexus/Toyota system and it is base/standard system too.

So at very least, autonomous system has a lot more sensors and can react a lot faster. Both lidar and radar do the same or better in dark than light, so it is actually perfect situation for system to react, just like test above.

System and test driver did not react even after a hit.
 
Messages
49
Reactions
95
okay ill add my $0.02 on autonomous driving bc as a LTS my job is always to stay on top of the latest and greatest.

obviously, the end game here is to have self driving cars, and not just lexus, but the entire automotive industry for mass market consumers. i saw something recently i think, about a chevy spark w/o a steering wheel. the real question here is when will this actually be a real thing that cars have in well, anywhere usa. i think there are many different factors and topics on the subject, and personally i think we are maybe, maybe 10 years from roads full of cars that drive themselves. we have the technology to do this and we have for years. i think at this point, its more of an emotional/mental barrier that we as a species need to break down. radar/laser guided cruise control systems have been around 20 years, and truth be told i dont think it has come nearly as far as it could have by now, simply because of the mental barrier we hold deep in our minds subconsciously.

i will say this. as an enthusiast, although i love the idea of being on a long highway cruise, pushing a button and letting the car do everything for me (lexus LSS+ comes damn close), i dread the day the steering wheel vanishes from our dashboards. many people these days consider driving a chore rather than recreation, regardless of what they are driving. i hear something to that tone atleast once a week from one of my guests, and honestly i dont blame them from thier point of view. if you have ever experienced Orlando rush hour traffic, you know what im talking about..
 
Messages
49
Reactions
95
btw, i put the PCS / ICS system to the test a few times in the real world. -im one of those people crazy enough to walk in front of an RX 350 going 10 mph thats convinced it will stop, bc i believe in these products, im not afraid to skin my knees, and just being one of those people that want to see how things work out of pure curiosity. i also tried higher speed testing of it in one of my videos, using a cardboard box with a persons face on it. i will say this, its pretty reliable!! with radar cruise control engaged, the car came to a complete stop at 30 mph with the cardboard cutout!!! i made 5 passes at different speeds. but no PCS system is fail proof. i did hit the box, once..
 

krew

Site Founder
Administrator
Messages
3,686
Reactions
5,670
krew

The Future of Lexus Self-Driving Cars
18-09-20-lexus-ls-autonomous-driving.jpg


An exhaustive overview by Businessweek.
View the original article post
 

ssun30

Expert
Messages
3,338
Reactions
7,438
Many people don't understand the so-called Hype Cycle: https://en.wikipedia.org/wiki/Hype_cycle

While not a law, the hype cycle curve proved to be correct in predicting the adoption rate of many new technologies in the past two decades. EVs, for example, are in the stage of 'Slope of Enlightenment', after the mass extinction of EV start-ups in early 2010s. Once a new technology passes this phase, wide scale adoption is just a matter of time. That is why EV is not a scam. They will replace ICEVs and HVs, eventually. On the other hand, FCVs are in the stage of trough of disillusionment, which is why the press is overwhelmingly negative about them.

Autonomous driving is now at the peak of inflated expectations, with 'tech companies' setting an unsustainabe fast pace in their race to L4. The truth is, the trough of disillusionment will come sooner or later, as tech companies ignore a lot of basic engineering principles to accelerate progress.

The sad thing is that world economy is becoming too hot especially in the tech industry. This results in a 'hype-driven' investment culture, where investors rely too much on (often false) promises instead of actual progress and financial results when making decisions. That's why Tesla is still afloat and why Alphabet is able to pour $10b into autonomous driving (and yes, that's more money than Musk spends on his Mars rocket).

Under such culture, if a company says anything negative about autonomous driving, or is not spending multi billion dollars working with or buying out a tech company, or is just a year or two late to the game, it will be labeled by the media as 'uninnovative' or 'resting on its laurels'.
 
Last edited:

RAL

Moderator
Messages
1,217
Reactions
1,756
hmmm ... from the Bloomberg Businessweek article: "... 55% of consumers surveyed wouldn't ride in a fully autonomous car, but more than 70% would ride in one that was partially autonomous - some of the very customers Toyota is targeting."

Change is coming; it is just a question of how fast consumers are willing to adapt to and trust new tech.

The parallel of innovation and development in transportation between the early 20th century and the early 21st century is interesting.
 

ssun30

Expert
Messages
3,338
Reactions
7,438
I have always failed to understand why they would use their most expensive car as the platform. They could save so much money by buying back used Prii. In the end all they need is a sensor carrier with enough power generation. Even if they need the largest vehicle to represent a worst case scenario they could just modify used RX hybrids.
In fact, it bugs me why the autonomous vehicle industry as a whole use expensive platforms while there are plenty of used Toyota hybrids lying around. They need very large fleets (thousands or even tens of thousands) to gather the data to train the AI, and $100k cars aren't the most economical.
 

Joaquin Ruhi

Moderator
Messages
1,529
Reactions
2,434
I have always failed to understand why they would use their most expensive car as the platform. They could save so much money by buying back used Prii. In the end all they need is a sensor carrier with enough power generation. Even if they need the largest vehicle to represent a worst case scenario they could just modify used RX hybrids.
In fact, it bugs me why the autonomous vehicle industry as a whole use expensive platforms while there are plenty of used Toyota hybrids lying around. They need very large fleets (thousands or even tens of thousands) to gather the data to train the AI, and $100k cars aren't the most economical.
I totally agree with your comments, but would point out (as you noted) that "the autonomous vehicle industry as a whole use expensive platforms". The carmakers currently selling semi-autonomous vehicles (GM and VAG) are offering the option on their most expensive models (Cadillac CT6 and Audi A8), so it may be a simple matter of Toyota expecting to launch consumer versions of their Guardian and Chauffeur suite of semi-autonomous features on the Lexus LS. Also, the Toyota Global Newsroom release announcing the TRI-P4 automated driving test vehicle doesn't say how many copies will be made for testing.

Perhaps Ford is being smarter on this, since I often see their semi-automated Ford Fusions (as opposed to Lincoln Continentals) driving around my Miami hometown.
 

Will1991

Moderator
Messages
1,573
Reactions
3,205
There's something that always bugged me on autonomous vehicles.... Wille Tesla believes it can do the job without expensive sensors, all the rest of the industry is using some lidar sensors and a lot more cameras... Toyota is even using thermal image cameras...
 

ssun30

Expert
Messages
3,338
Reactions
7,438
There's something that always bugged me on autonomous vehicles.... Wille Tesla believes it can do the job without expensive sensors, all the rest of the industry is using some lidar sensors and a lot more cameras... Toyota is even using thermal image cameras...

Tesla is wasting their data-gathering advantage by not integrating LIDAR and Musk WILL pay for his stubbornness or eat his own words. Data gathering is one thing, but it is really how effectively you use these data to train the AI that matters.

I really like how TRI is incorporating more sensors to the package to further boost information variety. Sensor fusion is a powerful performance multiplier and the more variety you gather the better the quality of the data (best example: imagine how differently you perceive the world with just your eyes). A well-trained machine vision AI can still fail to tell a pedestrian from a person on a billboard, with a thermal camera it's not a problem.
 
Last edited:

Joaquin Ruhi

Moderator
Messages
1,529
Reactions
2,434
As Toyota expands its U.S.-based Toyota Research Institute to a Japan-based counterpart (Toyota Research Institute Advanced Development Inc., or TRI-AD), a report by Automotive News' Asia Editor Hans Greimel includes some Lexus-related bits. I've highlighted and bolded those in my cut-and-paste of the article below:

February 09, 2019
Toyota's moonshot: Self-driving car for sale - in a year
Goal: ‘Most powerful supercomputer on wheels'
HANS GREIMEL

TOKYO — They call it Toyota's moonshot: In just one year, Japan's biggest carmaker wants to start selling a self-driving vehicle that it says will be the "most powerful supercomputer on wheels."

But the company has a lot of work to do first. It must develop hardware and software for the vehicle, dramatically slash the cost of the technology and ramp up the hiring of the software engineers who will make it all happen.

That's the challenge facing James Kuffner, CEO of Toyota Research Institute Advanced Development Inc., or TRI-AD. His Tokyo company was set up in March to spearhead Toyota's attempt to bridge the gap between research and the showroom floor.

The first milestone comes in 2020, when Toyota plans to introduce vehicles capable of driving themselves on highways. Kuffner said they would be rolling supercomputers.

"The prototypes and the preproduction vehicles that the team is building here at TRI-AD are going to be … the most intelligent supercomputer on wheels," Kuffner said late last month at TRI-AD's temporary office near Tokyo Station. "We've called it the moonshot of my generation to build this technology and bring it to market."

Toyota teamed with Toyota Group suppliers Aisin Seiki and Denso to invest $2.8 billion in the new company to create the software that runs self-driving vehicles.

Kuffner, an American, took the helm after a stint as chief technology officer at Toyota Research Institute, its future-looking research counterpart in Silicon Valley.

TRI-AD is tasked with making the 2020 technology, called Highway Teammate, a reality.

"If you think about building a research prototype, making a demonstration is pretty easy, but making a product is really hard," Kuffner said. "Whenever we talk about our company, we often talk about being a bridge of the prototype to the product."

Doing so means building a software company from scratch inside one of the most bureaucratic and traditional automakers in Japan. And the company is only partway there.

Chasing talent
TRI-AD has about 500 employees crammed into temporary space on two floors of a Tokyo high-rise. But this summer, it will move into new digs next door with room for 1,000 engineers on five floors. It will take several years to fill those ranks, Kuffner said, noting increasingly fierce competition worldwide to hire the best software and artificial intelligence engineers.

Toyota has more than one advanced prototype of the 2020 vehicle. A modified Lexus LS sedan is the mobile laboratory. Kuffner said the 2020 vehicle will also be a Lexus.

That is partly a function of cost. The car will need hardware and software not yet in production, Kuffner said. The array of advanced cameras, lidars and radars won't be cheap.


"This is our first automated vehicle," said Ken Koibuchi, TRI-AD's chief technical officer. "It has very high-performance computing and very rich sensors. So, it means very expensive. To deploy in middle-class vehicles, we have to reduce costs dramatically."

Koibuchi declined to give a figure but said cost must drop by more than half of today's levels.


Creating a culture
Part of the challenge will be cultivating a Silicon Valley-like corporate culture in the middle of Tokyo. TRI-AD will target development of "production-quality software for automated driving," partly by harnessing big data collected from connected vehicles.

"We have to embrace software and ride the wave like a software company," said Nikos Michalakis, who worked on cloud computing at online video streaming service Netflix and is now TRI-AD's vice president in charge of the software platform. Indeed, TRI-AD will institute a so-called Software Dojo as a kind of in-house school to teach Toyota employees computer culture.

"Our goal is to teach a Silicon Valley mindset here," Michalakis said.

Toyota and its partners already team on software development, but Toyota hopes to jump-start progress by combining efforts under one roof with an expanded staff and bigger budget.

"We know we can't do it alone, so we've forged strong partnerships," Kuffner said.

Creating the self-standing autonomous driving unit may help Toyota lure top computing talent as automakers increasingly compete against high-tech companies for talent.

Although TRI-AD is in Tokyo, English will be its business language. About 80 percent of the staff is Japanese, but several of the top officers are non-Japanese.

"People actually respond well to our mission," Kuffner said.

"I can tell a top-talented software engineer, 'Would you like to write software to sell ads, or would you like to write software to save lives?' " he said. "And they'll join us."

https://www.autonews.com/mobility-r...le-year?utm_source=dlvr.it&utm_medium=twitter
 

Will1991

Moderator
Messages
1,573
Reactions
3,205
"most powerful supercomputer on wheels", that's quite a statement... Let's see what 2020 will bring.
 

mediumhot

Follower
Messages
456
Reactions
606
They signed an agreement with nVidia to supply them with the fastest GPU accelerators they could offer. These things are literally supercomputer in a box.

All of the automotive autonomous technologies are built on Tegra chips cause they are the best in business right now. Maybe some use custom AISIC but from what I know all the major manufacturers are working with nVidia