Autonomous Vehicles

Nasher said:
Running lights seems to be a thing in the US. I saw the aftermath of least 3 crashes because of it when I was in Vagas. Over here it's quite rare to see it (well, except for cyclists lol).
Click to expand...

Cars running red lights is not rare thing in the UK at all, I see it regularly. If it wasn’t a regular thing then you wouldnt have seen any red light cameras put up.

Only last Thursday a driver ran through a solid red (obviously wasn’t paying any attention) and very nearly had a head on with a flat bed truck
 
AI will continue to be a growing feature of self driving vehicles. AI is also being used in many other areas too.

There is a link to TODAYonline, based in Singapore, to "The Rise of AI and how lives will be changed forever".

If you have a chance, check out the humanoid robot in this link and how life-like it really is. The humanoid is called Nadine and she has life like skin and hair. First introduced in Singapore in 2015 it has made strides in the past three years.

https://www.todayonline.com/big-read/big-read-the-rise-of-AI
 
sesevans said:
AI will continue to be a growing feature of self driving vehicles. AI is also being used in many other areas too.
Click to expand...

A growing feature? AI has been involved in the development of self driving cars from the 70s. It is pretty fundamental these days, that isn't something that is growing, it has been fundamental to all models for a while AFAIK. I dunno maybe you can find some example purely relying on control theory (though I'd wager even then there is a bit of an overlap, some learning step that could equally fall within the realms of machine learning).

There is a link to TODAYonline, based in Singapore, to "The Rise of AI and how lives will be changed forever".

If you have a chance, check out the humanoid robot in this link and how life-like it really is. The humanoid is called Nadine and she has life like skin and hair. First introduced in Singapore in 2015 it has made strides in the past three years.

https://www.todayonline.com/big-read/big-read-the-rise-of-AI
Click to expand...

These things are basically publicity stunts, I wouldn't pay too much attention to them, most of the time they generate headlines and mislead the public re: what is currently possible with AI
 
Waymo must be getting very close to launching its commercial ride hailing AVs as it released this video yesterday highlighting safety I believe in calling it their "360 degrees experience: a fully self driving experience."

As AVs are still a concept for most if not all of us (having never been a rider in one), this video is the closest thing to reality:

https://www.youtube.com/watch?v=B8R148hFxPw&t=1s
 
Ars is reporting that in Uber's recent tragic self driving vehicle crash in Arizona, sensors detected the victim but the software decided to ignore the victim.

According to Ars, "
Distinguishing between real objects and illusory ones is one of the most basic challenges of developing self-driving car software. Software needs to detect objects like cars, pedestrians, and large rocks in its path and stop or swerve to avoid them. However, there may be other objects—like a plastic bag in the road or a trash can on the sidewalk—that a car can safely ignore. Sensor anomalies may also cause software to detect apparent objects where no objects actually exist.

Software designers face a basic trade off here. If the software is programmed to be too cautious, the ride will be slow and jerky, as the car constantly slows down for objects that pose no threat to the car or aren't there at all. Tuning the software in the opposite direction will produce a smooth ride most of the time—but at the risk that the software will occasionally ignore a real object."

https://arstechnica.com/tech-policy...ubers-self-driving-crash/?comments=1&start=40
 
I have been reading lately that the infrastructure spending involved in converting a significant percentage of cars today to electric vehicles by the year 2040 will be a massive undertaking. How big a spend?

Estimates are that to move one-third of the world's cars to EVs by 2040, it will require a cumulative infrastructure spend of approximately $ 3 trillion. Where will all the EVs get batteries and where will these batteries charge?

It seems projections are for not one Gigafactory (Tesla's battery factory in Nevada) but 101 Gigabyte factories that will cost $ 5 billion each to build and supply.

Currently there are 13 major Lithium mines globally. We will need 160 major Lithium mines globally that cost $ 1 to $1.5 billion each to exploit.

As for electricity generation for this global fleet of EVs, we will need to add the equivalent of the entire US electrical generating capacity spread around the world, mostly in emerging markets. That would be equivalent to 4500 Terawatt hours of capacity.

It is worth considering how massive an undertaking the move towards EVs, let alone autonomous EVs will involve.

In sum, upstream it will involve electric generation and distribution and development of raw materials (lithium, cobalt, copper, etc) and refining.

Midstream it will require EV manufacturing and battery manufacturing.

Downstream it will require maintenance and service and charging.

Significant cost and changes ahead for the global economy. We will need a continuation of the strong global economy to pay for these massive infrastructure programmes.
 
At Google I/O Waymo's CEO revealed that Waymo self driving vehicles have now surpassed 6 million miles "driven" autonomously on public roads.

He spoke at length about how AI has greatly assisted Waymo's ability to both master "perception" (ability to classify objects) and understand and make behaviour predictions. While self driving 6 million miles, the real crown jewel is their simulation training which thanks to the latest (3rd generation) Tensor Processing Unit they are now 15 times more efficient at training their vehicles than in recent past.
 
I am wondering if data from AP/AV cars can be released under legal request to serve in investigation of the driver or other vehicle drivers , accidents per se ?
the car may/will have captured behaviour of other cars, their reg plates, even a quick snap of the driver, so this could revolutionanize policing.
[ will be like the FBI trying to obtain backdoor to iphones ]
 
jpaul said:
I am wondering if data from AP/AV cars can be released under legal request to serve in investigation of the driver or other vehicle drivers , accidents per se ?
the car may/will have captured behaviour of other cars, their reg plates, even a quick snap of the driver, so this could revolutionanize policing.
[ will be like the FBI trying to obtain backdoor to iphones ]
Click to expand...

I have to believe that such visuals/video would be subject to a judge to decide in the process of discovery.

This recent article about the Uber crash in touches upon some of the issues:

https://www.reuters.com/article/us-...nt-raises-novel-legal-questions-idUSKBN1GW2SP
 
ok seems like a discussion that was already had

https://arstechnica.com/tech-policy...nvestigating-crimes-and-accidents/?comments=1

"Well sir the data shows that you are in the clear for this accident, however, here are 8 speeding tickets, and a stunting charge for all this other stuff I see...."

...Why can't we just have nice, big opt out button? If there's no reason to store the data, and for a whole lot of uses there isn't, shouldn't the driver be able to tell the car to flush the data as soon as it's no longer needed?

Imagine an AV is capturing an image of pedestrians walking on a city street as an unrelated robbery takes place. Citizen X is carrying a concealed weapon and shoots, misses the perp and hits a bystander. All this is captured and the infrared sensors can ID the weapons and who shot exactly when. You, the owner of the car, are now part of the case when the police seize all your “NFL replay” data from the AV car
Click to expand...
 
@jpaul

I believe you asked recently about the inner workings of Waymo's self driving fleet--more how they accomplish their Level 4 performance?

This article in Verge should be an interesting read for you. It really gets into the weeds on machine learning, AI, deep neural nets and Google's current advantage in these fields. Will Waymo always have such an advantage? The article discusses that too.

https://www.theverge.com/2018/5/9/1...rless-cars-deep-learning-neural-net-interview

Their VP Engineering sheds further light:

https://medium.com/waymo/google-i-o...-into-reality-with-the-help-of-ai-89dded40c63

Seems to me that Google is beginning to "open the kimono" a bit more these days in a bid to give the public a sense of security about how safe their "driver" really is.

One of my favourite excerpts from the article:

"Anyone can buy a bunch of cameras and LIDAR sensors, slap them on a car, and call it autonomous. But training a self-driving car to behave like a human driver, or, more importantly, to drive better than a human, is on the bleeding edge of artificial intelligence research. Waymo’s engineers are modeling not only how cars recognize objects in the road, for example, but how human behavior affects how cars should behave. And they’re using deep learning to interpret, predict, and respond to data accrued from its 6 million miles driven on public roads and 5 billion driven in simulation."
 
Last edited:
sesevans said:
I believe you asked recently about the inner workings of Waymo's self driving fleet--more how they accomplish their Level 4 performance?
.....
One of my favourite excerpts from the article:

"Anyone can buy a bunch of cameras and LIDAR sensors, slap them on a car, and call it autonomous. But training a self-driving car to behave like a human driver, or, more importantly, to drive better than a human, is on the bleeding edge of artificial intelligence research. Waymo’s engineers are modeling not only how cars recognize objects in the road, for example, but how human behavior affects how cars should behave. And they’re using deep learning to interpret, predict, and respond to data accrued from its 6 million miles driven on public roads and 5 billion driven in simulation."
Click to expand...

Thanks, beyond the use of neural nets for image recognition which is also discussed in the mobileeye/intel utube video,
the details of the control/stratgey seem sparse
.. I should probably first understand how deep blue manages to play chess, since that strategy is probably close to what the car has to decide ?

did see a research paper too
https://www.researchgate.net/public...control_systems_-_A_review_of_decision_making


edit - starting to read up on deep blue https://www.wired.com/2017/05/what-deep-blue-tells-us-about-ai-in-2017/ salient remark ?
.....When I covered Kasparov-Deep Blue match, I thought the drama came from a battle between computer and human. But it was really a story of people, with brutal capitalist impulse, teaming up with AI to destroy the confidence and dignity of the greatest champion the world had seen. That leads me to believe it’s not Skynet that should worry us about AI, but rather the homo sapiens who build, implement, and employ those systems.
Click to expand...
 
Last edited:
jpaul said:
Thanks, beyond the use of neural nets for image recognition which is also discussed in the mobileeye/intel utube video,
the details of the control/stratgey seem sparse
.. I should probably first understand how deep blue manages to play chess, since that strategy is probably close to what the car has to decide ?

did see a research paper too
https://www.researchgate.net/public...control_systems_-_A_review_of_decision_making


edit - starting to read up on deep blue https://www.wired.com/2017/05/what-deep-blue-tells-us-about-ai-in-2017/ salient remark ?
Click to expand...

Many thanks for the researchgate research paper, a really great in depth look at AV control and the mathematical models underlying the implementation.

I had not considered Big Blue and the chess match (or for that matter, Google's Alpha Go match vs the Go world champ more recently) but see its relevance (deep learning/neural net approach) to aiding our understanding of the strategy.

Will keep an eye out for more.
 
Last edited:
Ohio has become the fourth US state to permit testing of autonomous vehicles. They do however require a safety driver behind the wheel at all times, which is understandable in the current political climate following the fatality in the Uber crash and with the Tesla recent crash, but is limiting. An advantage of testing in Ohio is the winter weather but AVs already have that with Michigan currently.

https://www.cnet.com/roadshow/news/ohio-green-lights-testing-of-self-driving-cars-on-state-roads/
 
You must log in or register to reply here.
Back
Top Bottom