Will we see a real self driving car in our life time?

Alec_t

Joined Sep 17, 2013
14,280
The google car relies on very accurate mapping
Hmmm. Post (zip)-coded areas in suburban areas of the UK are generally accurate to within tens of metres, but I've found a couple of errors of at least 100 metres on Google's maps of the Cardiff region. Shakes my faith in the mapping. Are our lives going to depend on it ? :(
 

Thread Starter

spinnaker

Joined Oct 29, 2009
7,830
Hmmm. Post (zip)-coded areas in suburban areas of the UK are generally accurate to within tens of metres, but I've found a couple of errors of at least 100 metres on Google's maps of the Cardiff region. Shakes my faith in the mapping. Are our lives going to depend on it ? :(

Yet another problem, accuracy of data. Combine GPS error and you have a real problem. In a recent bicycle tour, I was to stay at a fellow cyclist in Oswego, NY. I was following the GPS to his home, I unknowingly passed his street as the GPS said the turn was further down the road. I did not notice the street sign as it was only on the opposite side of the road. Not on my side. When I got to where the GPS daid the turn was I noticed the sign being wrong. Confused I got on the phone with my host and told me I had passed the road. I would like to see a computer do that.

Sometimes a route changes in seconds and humans can adjust. Say a road is flooded or there is an accident and the road is blocked. Some drivers might be out of luck as there is no where to go forward, left or right or back due to traffic. While others might recognize a detour say through a parking lot (car park for Alex) . The trapped drivers then might get a break as traffic behind them eases because the drivers behind them recognized the impromptu detour and took it. No computer today is going to be able to figure that out, mainly do to the fact that the parking lot is not mapped. For any of this to work we are going to need an extremely accurate and finely detailed digital maps. I just don't see that happening any time soon as we don't have our current maps accurate with their relative limited detail.
 

LDC3

Joined Apr 27, 2013
924
How reliant are the autonomous systems on markers such as roadside signs/transponders? I can envisage practical systems working on well-defined highways, but have my doubts as to their effectiveness on wild country tracks and mud roads with no street furniture. Driving on snowed-over roads would be a challenge, when visual cues may well be absent.
Did you see the DARPA challenge from a few years ago? All that was given was about a dozen way-points that the vehicles needed to get to and complete the circuit.
 

Alec_t

Joined Sep 17, 2013
14,280
Thanks for the link. Confirms data accuracy is life-critical. I'm not ready to trust it yet.
As for autonomy happening soon, I think not. All the 'stick shift' cars (currently ~7% in the US but ~75% in the UK and heaven knows what percentage in other countries) will have to be off the road first. 10 years at least?
 

#12

Joined Nov 30, 2010
18,224
Self driving cars will make it much easier for our government to track all civilian movement.
Self-driving trucks would cause a lot of unemployment and make much larger crashes when something goes wrong.
And, self-driving cars will make delivering car bombs much more convenient.
I consider all of these to be reasons to NOT implement self-driving cars.
Of course, all of these reasons fail if you believe that computerized systems can be designed for every possible contingency and they never have a garbled input or run slower than planned.
 

wayneh

Joined Sep 9, 2010
17,496
Those are good reasons, but I still contend that the tipping point will be determined by liability issues. Once the robot is technically superior, I can just hear the trial attorney, "You did WHAT sir? You dared to drive the car yourself, thinking you knew better than billions of dollars worth of research and technology? How DARE you! Your honor, my client deserves a jillion dollars from this arrogant bastard that thinks he can drive better than the computer."
 

Glenn Holland

Joined Dec 26, 2014
703
I used to work on elevators which transport the next highest number of persons second only to automobiles.

Elevators are regulated by local governments and based on the ASME/A17.1 Code. Despite the huge amount of persons transported by elevators, the fatality rate in the U.S. and Canada is only about 20 per year. Because of the huge risk to the public, the code requires extensive testing before a given product is allowed to go into public service.

From the standpoint of public safety, autonomous cars are in the same risk category as elevators and they will also need to undergo an enormous amount of testing to verify that the technology is safe.
 

Thread Starter

spinnaker

Joined Oct 29, 2009
7,830
From the standpoint of public safety, autonomous cars are in the same risk category as elevators and they will also need to undergo an enormous amount of testing to verify that the technology is safe.
And elevator travel is child's play when it comes to all of the various things that can happen on the route. I just don't see how adequate testing would be possible.
 

Glenn Holland

Joined Dec 26, 2014
703
As a stepping stone, the proponents of autonomous cars should first try to develop "Positive Train Control" or PTC first before trying to venture into the auto industry.

PTC is the futuristic technology for the railroad industry and it will supposedly replace the outdated block signal system and allow trains to run with much closer headways.

The most successful version of PTC (trademarked SELTRAC) has been developed by Thales Aerospace and used on over 100 rail transit systems throughout the world. However, the reliability of PTC is considered only "Fair" and the technology is still considered as "Not ready for prime time."

So if a railroad signal system (a "single axis" vehicular control system) is still in the proving stage, it's quite premature to make the leap to autonomous cars. The old saying applies: "You need to learn to walk before going out for the Olympic running events".
 

wayneh

Joined Sep 9, 2010
17,496
And elevator travel is child's play when it comes to all of the various things that can happen on the route. I just don't see how adequate testing would be possible.
So true. I always give the example of deer crossing. Where I live we see a lot of deer. Over time you learn that the one you see is unlikely to be a problem. It's the 2 or 3 following behind that dash across at the last second. For this reason I always slow down when I spot a deer. I can't tell you how many times I've had drivers roar past me only to hit the deer that no one saw.

Humans are (or I should say can be) extremely good at pattern recognition, risk assessment and threat avoidance. Those are all still pretty tough for computers.

On the other hand.
 
Last edited:

sirch2

Joined Jan 21, 2013
1,037
yes but computers have a broader spectrum of tools, doppler radar, laser distance, ultrasonic, infrared. I accept that it is a big step to track moving animals on the side of the road but anything in front of the current track of the car is just a distance measurement problem.
 

Thread Starter

spinnaker

Joined Oct 29, 2009
7,830
yes but computers have a broader spectrum of tools, doppler radar, laser distance, ultrasonic, infrared. I accept that it is a big step to track moving animals on the side of the road but anything in front of the current track of the car is just a distance measurement problem.
Doubtful if Doppler radar is going to penetrate thick brush. At least not very effectively. And even it can detect additional deer, how would the computer know they present a threat?

On a similar note, lets take kids playing on a sidewalk. They appear to not present a threat but most humans know kids are kids and they can run into the road at any time. A human will slow down and be extra observant when seeing the kids. Not sure a computer would be able to recognize that. The computer has the advantage of being quicker to react than the human but is not going to have the human intuition.

Nope I am not seeing it. Way too many variables in the real world that most times a human has no problem at all dealing, to a computer it is an enormous task,
 

Alec_t

Joined Sep 17, 2013
14,280
Since accurate high resolution map data is key in this autonomy scenario, whoever controls access to the data is omnipotent. Do we entrust it to a commercial corporation, or to a Government, or .....? What if access control gets into the hands of criminals or enemies? There is scope for exploitation, exorbitant charging, corruption, denial of access, subversion, economic warfare etc. A nation could be brought to its knees in an instant.
 

nsaspook

Joined Aug 27, 2009
13,086
http://www.nsf.gov/discoveries/disc_summ.jsp?cntn_id=134033&org=NSF
There are a lot of subtle cues that go into safe driving. Take for example a four-way stop. Officially, the first car to the crosswalk goes first, but in actuality, people watch each other to see if and when to make their move.

"There is a slight negotiation going on without talking," Zilberstein explained. "It's communicating by your action such as eye contact, the wave of a hand, or the slight revving of an engine."

In trials, autonomous vehicles often sit paralyzed at such stops, unable to safely read the cues of the other drivers on the road. This "undecidedness" is a big problem for robots. A recent paper by Alan Winfield of Bristol Robotics Laboratory in the UK showed how robots, when faced with a difficult decision, will often process for such a long period of time as to miss the opportunity to act. Zilberstein's systems are designed to remedy this problem.
 

tjohnson

Joined Dec 23, 2014
611
A four-way stop would be difficult for a self-driving car to handle, but I don't think it would be impossible. I remember reading somewhere about an idea for self-driving cars near each other to be able to detect if they are traveling on the same route for a while, and if so, to hitch together to be more aerodynamic and save fuel. If they had those sort of capabilities to detect other self-driving cars nearby, I think it would actually be fairly easy to handle four-way stops.

Basically, when a self-driving car gets to a four-way stop, it would check if there are cars coming in any of the other directions, and the cars would all be synced with each other to go in a certain order. I don't know how easily a four-way stop could be detected, so Google might need to make a database of them for this idea to work.

Another problem is how a mixture of self-driving and human-driven cars on the road would work. If self-driving cars improve enough to be used by the general public, the government would either have to pass a law requiring everyone to switch to a self-driving car, or else require human-driven cars to have devices installed in them that could communicate with the self-driving cars.
 
Top