Autonomous Vehicle ...
 
Notifications
Clear all

Autonomous Vehicle L5 crash

22 Posts
5 Users
0 Likes
1,944 Views
RolfGutmann
(@rolfgutmann)
Posts: 1185
Noble Member
Topic starter
 

Imagine the confrontation between an Autonomous Vehicle AV-L5 and a Human Vehicle HV. An accident occurs. Both parties normally get interviewed. The HV by the officer and the AV by OBDII downloading the report.

The AV proofs, speed, angles, video in short all sensors data logged. But does not reveal how the decision making (driving policy) was made - Algo knows.

In the focus is the problem of interpretation of the traffic reality. Tech beats human in almost all aspects only by same forward looking view angle and visual information they are equal. More eyes see more than two human eyes - but there is greater brain for interpretation.

Actually law does not adapt to AVs. And also no prison cells for AVs. The car manufacturers now start a dialogue of mutual agreed law improvement. Humans should say the 'new law' is common sense and socially unquestioned accepted.

The today's traffic laws around the globe are not harmonized but defined nationally as they reflect invisible factors based on history and experience. Lets assume an AV crosses a national boarder. The AV has to switch the legal layer. But who than judges of a certain country can provide the interpretation of law? A very actual layer of national law has to be fed into the car. If you count all possible accident situations for 360 degree mapped to different street situations you may get a sense of how a 'dangerous situation' is possible. If the 'dangerous situation' causes an accident then guilty. See here a nice try to mathematical crack the nut

https://arxiv.org/pdf/1708.06374.pdf

There is one crucial question now and in 2021 full AVs arrive (Level 5)

Who caused the accident (and is therefore guilty)?

The tools to find Car Forensics - and interviewing. Hang-in -)

 
Posted : 10/01/2018 8:00 pm
hcso1510
(@hcso1510)
Posts: 303
Reputable Member
 

Legally I would suggest this is uncharted territory. Lets say you have two diverless cars. I have no idea how these things are programmed, but lets say the speed limit is 55MPH. Does the car adjust to "driving conditions" seeing how most folks drive 70MPH in a 55? Or does the driver have any input in how fast the vehicle can travel above 55MPH?

In a wreck who is at fault? Who can get sued? And what data is available? Everyone points fingers and the maker of the driverless car either says.

1. That information is not retained
2. That information can't be released because it is proprietary
3. Oops. Jokes on you! We have that data, but since it is retained overseas your domestic legal process can't legally compel us to provide it.

 
Posted : 10/01/2018 9:54 pm
RolfGutmann
(@rolfgutmann)
Posts: 1185
Noble Member
Topic starter
 

Very good aspects to take into consideration. Shall we focus on our job to proof forensic evidence of the TWO cars involved in an AV-H (AV-Human) crash? What situations LEO gets called to proof the 'technical cause of failure'? In public opinion killed little children are the worst in perception. For further speculation lets assume a little child jumps on the street and gets killed by an AV. What questions occur (as the expectation against an AV, as more technical equipped is higher)?

#1 - Why was the AV not able to brake to avoid?
%1 - "AVs are not better than humans, I never believed that!"

#2 - Was the AV better than H to avoid?
§2 - "They just say that to look positive, I do not believe it."

#3 - Is the car manufacturer guilty?
§3 - "Yes, the AV should have to brake" (not possible but over-expected by Hs)

#=probable questions to Hs
§=probable answers by Hs

Its clear that the public opinion influences the sale of AVs and the Fortune 500s' stock extremely.

If I get called to investigate the AV in the mentioned case, what comes first? Faraday the car, as the eSIM is probably on the safest location (No, just disassemble the shark-fin antenna, No, the wire to the roof is antenna enough)?

Black-Thinking AVs should highly avoid areas with little kids and convince the passenger that it was faster this way. Baby-David will kill Goliath.

Whats next?

 
Posted : 11/01/2018 8:01 am
(@trewmte)
Posts: 1877
Noble Member
 

L5 has no driver involved, it is purely an autonomous vehicle (AV).

L2+ requires 'shadow' human driver.

 
Posted : 11/01/2018 9:34 am
RolfGutmann
(@rolfgutmann)
Posts: 1185
Noble Member
Topic starter
 

L5 no driver, correct. Therefore passenger named and not driver.

 
Posted : 11/01/2018 9:57 am
(@trewmte)
Posts: 1877
Noble Member
 

My two year research into 'autonomous' deployed at e.g. NASA, aerospace, F1, UAV, missiles, etc. is that simulations are run millions of times before effective deployment (even then things can still go wrong). At this stage AV at consumer level is attempting to establish the principles, technicality, technology-context and presence, and safety is being developed alone side it.

Full AV (e.g. L5) large industrial vehicles on mining sites can use L5 (in tandem with a control tower) and they do a pretty good job under very limited and restricted circumstances.

From all the presentations and research to date (L5 doesn't exist at the moment in consumer terms that is). L5 could possibly, when available, be a local limited-distribution bus or delivery service etc. Passengers on a bus, for instance are most unlikely to have any liability at all.

Liability may either rest with the vehicle operator, the AV manufacturer or both. Insurance will be the main factor and regulators are making it clear the pay outs will be sky-high.

 
Posted : 11/01/2018 11:16 am
RolfGutmann
(@rolfgutmann)
Posts: 1185
Noble Member
Topic starter
 

Great input. On the scale of Level 0 - 5 only two are safe 0 or 5 as mixed-responsibilities ever fail somehow.

But the market needs scalability to cash-in after investments. They actually fear an 'AV winter' (see Amnon Shashua paper) and dont like to end in a science project dead-end. They highly strive to bring L2+ (crowdsourced RoadBook) to the masses.

But back to reality Car Forensics on a L2 car to bring-out the separate responsibilities is like smog and the disclaimers at-buy-or-lease-or-rent a car clearly towards no-guilty for the manufacturer ever.

On thing remains. What can Car Forensics really proof and what not? We search and test for what we can proof and is court-hard.

What can Car Forensics never proof (then we can stop research)?

 
Posted : 11/01/2018 11:30 am
(@trewmte)
Posts: 1877
Noble Member
 

Implicit within AV will be to define the AV system installed and operated. The vehicle operator can still be the driver between L2+ onward. The key is to determine AV-integrated and AV-add-on. If AV is used on a consumer vehicle as an add-on (L2-L5) then the vehicle operator becomes responsible under "selection and choice" - foreseeability (the man on the Clapham. Omnibus). It is not clear at this stage whether an add-on manufacturer could produce an L4/L5 device.

Car forensics may not be relevant where the data is stored in the AV add-on device, but relevant to integrated systems. Car forensics would still be fully relevant for the Electronic Control Modules (ECM).

For law enforcement, like Ed Merritt, ascertain the knowledge about an integrated system would be priority in any accident. For instance, if LiDAR was integrated into vehicle wing mirrors and one or both mirror/s got damaged or knocked off the vehicle; what happens thereafter with respect to control of vehicle? What other technologies are deployed within the AV integrated system? How would that compare to a deployed add-on AV system/technologies?

In short, it is far too soon to be writing anything off….

 
Posted : 11/01/2018 12:06 pm
RolfGutmann
(@rolfgutmann)
Posts: 1185
Noble Member
Topic starter
 

Actually the 10 seconds between car-lead and human-lead handover is legally unsecure. The car-lead until handover (variable action time of humans on same handling) responsability is clear. But the 'action-time if human not takesover' is problematic.

Look at the legal separation between 'control' and 'assist'. AEB e.g. clearly is an autonomous function if enabled. But look sharply that the moment of enabling (system time to enable unknown) brings confusion related to traffic environment.

As long as law is clear the driver is guilty. But whould you buy a new ADAS car after found guilty in a situation you trusted your oADAS-car?

The level on smog is bigger as visible. It really is like smog.

Remarks In Switzerland a Tesla driver was declared guilty on a autonomous driving crash. His lack was that he could not proof to had both hands on the steering wheel. What if he could proof, he did?

 
Posted : 11/01/2018 12:52 pm
(@trewmte)
Posts: 1877
Noble Member
 

Actually the 10 seconds between car-lead and human-lead handover is legally unsecure. The car-lead until handover (variable action time of humans on same handling) responsability is clear. But the 'action-time if human not takesover' is problematic.

Look at the legal separation between 'control' and 'assist'. AEB e.g. clearly is an autonomous function if enabled. But look sharply that the moment of enabling (system time to enable unknown) brings confusion related to traffic environment.

As long as law is clear the driver is guilty. But whould you buy a new ADAS car after found guilty in a situation you trusted your oADAS-car?

The level on smog is bigger as visible. It really is like smog.

Remarks In Switzerland a Tesla driver was declared guilty on a autonomous driving crash. His lack was that he could not proof to had both hands on the steering wheel. What if he could proof, he did?

Which Level (L1-L5) are you referring to you for the above?

 
Posted : 11/01/2018 1:32 pm
Page 1 / 3
Share: