Diesel cars banned in Frankfurt

826 posts in this topic

48 minutes ago, MikeMelga said:

Planes use a mix of GPS an inertial navigation. Spoofing does not work on planes because:

  • they are above us, so they can direct the antennas up to the sky and ignore signals from the grounds
  • most of the time they are quite far from land
  • a 500 meter error while cruising is not a big problem, except in congested areas, but usually separation is done vertically
  • On landing approach, they don't rely on GPS

 

ah sorry, I was not talking about onboard navigation, I thought brad was talking about air traffic control systems.   Still, are they really using "normal" (as in, what we here on the ground get) gps or dedicated systems?  

 

anyway yeah, on board systems are just too far away to hack from the ground...unless you can hack into the satellite itself, but that's a whole other issue with far more frightening potential if hacked.  Suffice it to say I'm not worried about planes for that reason.  it's more the mcas type stuff that freaks me out.

0

Share this post


Link to post
Share on other sites
10 minutes ago, MikeMelga said:

Planes use a mix of GPS an inertial navigation. Spoofing does not work on planes because:

They also use TRN, Terrain based Navigation

  • they are above us, so they can direct the antennas up to the sky and ignore signals from the grounds

GPS spoofing is very effective

  • most of the time they are quite far from land

Only because 75 % of the earth is water

  • a 500 meter error while cruising is not a big problem, except in congested areas, but usually separation is done vertically

Vertical, measurements is GPS worsed feature

 

10 minutes ago, MikeMelga said:
  • On landing approach, they don't rely on GPS

They use it, but its just another sensor

 

 

10 minutes ago, MikeMelga said:

PS: for the Apollo program, inertial navigation systems were developed and demonstrated. 

 

We should not forget, the the V1/v2 rocket program used inertial navigation systems to get to the target.

 

 

10 minutes ago, MikeMelga said:

BTW, it is possible to detect a brute spoofing, where the car's position changes dramatically, by combining information from the inertial navigation. But slower/small changes on spoofing become very hard to detect.

 

Correct

10 minutes ago, MikeMelga said:

 

 

0

Share this post


Link to post
Share on other sites

yeah but it's far more interesting conversation, so?  shall we go back to the poor people vs not poor people and why a ban is fair, not fair schtick?

 

eta:  and I'm pretty sure planes use altimeters, not gps, to determine their vertical position.

1

Share this post


Link to post
Share on other sites
43 minutes ago, lisa13 said:

 

ah sorry, I was not talking about onboard navigation, I thought brad was talking about air traffic control systems.   Still, are they really using "normal" (as in, what we here on the ground get) gps or dedicated systems?  

 

anyway yeah, they are just too far away to hack from the ground...unless you can hack into the satellite itself, but that's a whole other issue with far more frightening potential if hacked.  Suffice it to say I'm not worried about planes for that reason.  it's more the mcas type stuff that freaks me out.

OK, you got me.  Spoofing the GPS wouldn't work.  Shows how dumb I am.  There is no need to worry about planes being hacked.  Only autonomous cars can be hacked.  Nothing else.

0

Share this post


Link to post
Share on other sites

oh come on.  don't go nuclear.  It's just far less likely you can actually, logistically, pull off a plane hack than hacking a car driving past you on the street.

 

anything is possible, but many things are not at all probable.  And I never accused of being, nor do I think you are, dumb.  Relax.

1

Share this post


Link to post
Share on other sites
1 hour ago, MikeMelga said:

Planes use a mix of GPS an inertial navigation. Spoofing does not work on planes because:

  • they are above us, so they can direct the antennas up to the sky and ignore signals from the grounds
  • most of the time they are quite far from land
  • a 500 meter error while cruising is not a big problem, except in congested areas, but usually separation is done vertically
  • On landing approach, they don't rely on GPS

 

PS: for the Apollo program, inertial navigation systems were developed and demonstrated. One air force plane crossed USA and the pilot could not see outside, except for landing. After crossing the country he was only 3km off, relying only on instrument data. No GPS, just altimeter, air speed sensor and inertial navigation.

 

BTW, it is possible to detect a brute spoofing, where the car's position changes dramatically, by combining information from the inertial navigation. But slower/small changes on spoofing become very hard to detect.

 

Exactly! Speaking again about airplanes, it is incredible the amount of situation where people try to blind pilots with illegal lasers! Making it illegal solved nothing!

Can the mods please close this Thread... It is clear that MM knows everything and everything is going to be ok...

 

 

 

ETA... Can you also kind shoot J² too... 

 

 

0

Share this post


Link to post
Share on other sites
10 minutes ago, SpiderPig said:

Can you also kind shoot J² too

 

oh no, no - I think you are looking for the hunting discussion.

 

0

Share this post


Link to post
Share on other sites

The results of the investigation of the Uber accident (the video posted here some days ago) were released yesterday:

 

https://www.ntsb.gov/news/events/Documents/2019-HWY18MH010-BMG-abstract.pdf

 

Additional bits that I didn't see there:

 

- The car operator was watching an episode of The Voice on her smartphone at the time of the accident.

- Plenty of corners were cut because Uber was already thinking in closing the program due to excessive R&D costs.

- Obama relaxed the rules for testing self-driving cars and then Trump remove plenty of rules and even closed the Self-Driving Car Committee.  The submission of reports is now voluntarely and of course the company cars now only send useless information that is basically sales brochures.  Right now you can test self-driving cars in the US without having to prove their safety before.

 

Really interesting thing from the report: The system actually detected the pedestrian long enough (5.6 seconds before impact), but it failed to determine it was a danger and it failed to predict its path.   Maybe it got confused because it was a pedestrian pushing a bike?

 

3

Share this post


Link to post
Share on other sites

That's my point

 

Unless the computer can correctly, identify the situation from a picture and various other sensors, then it will never work.

 

Maybe NN will work in the end, but I would like to show it a series of photos, and ask it to identify the bridges, other cars, people, dogs, potential threats, line of the road, like humans  can/are expected to do - before the system even gets close to the road.

 

The Uber, system, classified all things it could not identity as non-threats and carried on.

 

It is completely un acceptable, to test this system, in the really world - and watch people die - without a massive amount of testing being done beforehand.

0

Share this post


Link to post
Share on other sites
2 minutes ago, yesterday said:

That's my point

 

Unless the computer can correctly, identify the situation from a picture and various other sensors, then it will never work.

 

Maybe NN will work in the end, but I would like to show it a series of photos, and ask it to identify the bridges, other cars, people, dogs, potential threats, line of the road, like humans  can/are expected to do - before the system even gets close to the road.

 

It is work in progress, you can't expect things to work like that if they are not yet invented.  That's why autonomous driving was divided in several levels, L1, L2, L3, etc.  And we are still at the bottom of the ladder.

 

2 minutes ago, yesterday said:

The Uber, system, classified all things it could not identity as non-threats and carried on.

 

It is completely un acceptable, to test this system, in the really world - and watch people die - without a massive amount of testing being done beforehand.

 

Yes, we agree.  But Trump happened.

 

 

P.S., Another ironic thing, the cars were Volvos which include some Driving Assist system that can brake by itself in emergency situations.   In order to test their systems Uber deactivated the stock safety system and it is thought that Volvo's system would have brake.

0

Share this post


Link to post
Share on other sites
12 hours ago, lisa13 said:

 

eta:  and I'm pretty sure planes use altimeters, not gps, to determine their vertical position.

 

Of course, they use both and different types of altimeters

 

Radio Altimeters are used when close to the ground, as they do not work above certain altitudes, only really usefully if the terrain height around the airport is the same height as the airport - ie  no -mountains/cliffs around the airport.

Baro Altimeters work at higher altitudes, but are less accurate than Radio Altimeters, but not so important at higher altitudes

GPS works at all altitudes, but the lat/long is more accurate than the vertical component.

IN's were very popular up-until GPS came along, INS's lose accuracy over time, hence GPS used in combination to produce a more accurate position. There are several ways to improve accuracy

 

Its the combination of the sensors, that allows planes to fly closer to each other, eg over the Atlantic, as the position is better known - having the planes flown closer to each other, means you can fly more planes in the box of space, and hence carry more people.

 

Radar is generally used by commercial Airports, to locate and coordinate landing and takes off around the airport / controlled space.

0

Share this post


Link to post
Share on other sites

 

Quote

Maybe NN will work in the end, but I would like to show it a series of photos, and ask it to identify the bridges, other cars, people, dogs, potential threats, line of the road, like humans  can/are expected to do - before the system even gets close to the road.

Google images...? This already works for a decade! Identifying objects is very easy. Identifying what they want to do is harder, but perfectly in reach. In any case, without NN there is no self driving.

 

1 hour ago, yesterday said:

The Uber, system, classified all things it could not identity as non-threats and carried on.

Uber is a shitty company that stole driving software from Google, and they are going down to bankrupcy in a few years. That is a very bad example. You must look at the current leaders, Waymo, Tesla and maybe Mobileye. All others are years away.

 

Quote

It is completely un acceptable, to test this system, in the really world - and watch people die - without a massive amount of testing being done beforehand.

Ok, first, cool down, you don´t know much about the topic and you are completely overreacting. "People die" is relative, if you see data from Tesla (although details are not public, and subject to debate), when using AP you actually have much less accidents!

Second, testing HAS to be done on the road! You don´t go far with simulated data!

Third: it only has to be slightly safer than a human to pay off. 100% safety is not the trigger for release

 

Tesla is about to release a massive update to FSD, within the next 2 months. This will use the full potential of the new HW and it should be a giant leap in this field. Then my guess is that it will take them 1-2 years to fine tune it, but anyway either this shows a massive improvement or their approach is flawed.

0

Share this post


Link to post
Share on other sites
4 minutes ago, MikeMelga said:

 

Google images...? This already works for a decade! Identifying objects is very easy. Identifying what they want to do is harder, but perfectly in reach. In any case, without NN there is no self driving.

 

Maybe it works  - only time will tell

 

4 minutes ago, MikeMelga said:

Ok, first, cool down, you don´t know much about the topic and you are completely overreacting. "People die" is relative, if you see data from Tesla (although details are not public, and subject to debate), when using AP you actually have much less accidents!

 

Interesting when you say "Tesla (although details are not public, and subject to debate), ", where have you got the data from, are you a personal friend of Musk, and he gives you the data and not to the public.

 

4 minutes ago, MikeMelga said:

Second, testing HAS to be done on the road! You don´t go far with simulated data!

Third: it only has to be slightly safer than a human to pay off. 100% safety is not the trigger for release

 

Sure testing has to be done on the road at some point, but there should be government controlled tests in the lab, to ensure the system can meet minimum  requirements before it gets to the road. What we have at the moment, is the programmer says yes its good to go and then its on the street. Sounds quite like the Boeing, saying its 737 is safe to fly and the FAA, just says ok and it flies.   

 

4 minutes ago, MikeMelga said:

 

 

0

Share this post


Link to post
Share on other sites
Just now, yesterday said:

Interesting when you say "Tesla (although details are not public, and subject to debate), ", where have you got the data from, are you a personal friend of Musk, and he gives you the data and not to the public.

The data is public, I´ve posted a link a few pages back. What is not public is Tesla´s definition of "accident".

 

Just now, yesterday said:

Sure testing has to be done on the road at some point, but there should be government controlled tests in the lab, to ensure the system can meet minimum  requirements before it gets to the road. What we have at the moment, is the programmer says yes its good to go and then its on the street. Sounds quite like the Boeing, saying its 737 is safe to fly and the FAA, just says ok and it flies.   

Lab tests are worthless. The best tests government can do are defining what constitutes an "accident" and then get statistics.

And it is not as simple as the programmer deciding it is good enough. For example Tesla operates their changes on "Shadow mode" for months, on each car, to check if it improved or regressed. Then they make a very slow release to fleet, to people who actually agreed to beta test it!

Bottom line: self driving is impossible without real world usage, and you need LOTS of it.

0

Share this post


Link to post
Share on other sites
On 11/20/2019, 11:06:24, fraufruit said:

Final question - Are diesel cars banned in Frankfurt?

 

I think Frankfurt is now moving towards diesel ban only in selected streets, just like the Berlin diesel ban that will start next week.

0

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now