So we allow assumptions in news then too? Again, there are no facts, and nowhere did it say AEB or FCW (which are safety systems, not partially automated driving systems, nor advanced driver assistance systems). The title of the article is "US to investigate Texas fatal crash that may have involved Ford partially automated driving system".
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autonomous Car Progress
- Thread starter stopcrazypp
- Start date
-
- Tags
- Autonomous Vehicles
Those are my thoughts and NTSBs quote...
Given the Mach E hit the stationary vehicle, it seems like AEB/FCW did not achieve their purpose however you choose to categorize them.
NHTSA categorizes them as driver assistance technologies:
Driver Assistance Technologies | NHTSA
If that's the case, I welcome it. Since the majority of late model cars from most mfgs have AEB as a standard safety feature, and we still have hundreds of thousands of crashes every year, we need to get to the bottom of why AEB and FCW are still not helping prevent these accidents.
Digging in to the crash.
It was a Blue 2022 Mustang Mach-E (VIN 3FMTK3R77NMA43887).
The crash report lists the following (and a lot more, I just trimmed it down).
"Autonomous Level Engaged: 2 - PARTIAL AUTOMATION."
"Contributing Factor 1: 20 - DRIVER INATTENTION"
"Vehicle Damage Rating 1 - Severity: 5 - DAMAGED 5"
"Vehicle Damage Rating 1 - Area: FC - FRONT END DAMAGE CONCENTRATED IMPACT"
"Vehicle Damage Rating 1 - Direction of Force: 12 - 12 O'CLOCK"
The driver wasn't from the area (according to the crash report he's from a suburb of Houston) and has to wear glasses.
The other vehicle was a Black 1999 Honda CR-V (VIN JHLRD1849XC008389).
"Contributing Factor 1: 55 - PARKED IN TRAFFIC LANE"
"Contributing Factor 2: 56 - PARKED WITHOUT LIGHTS"
"Vehicle Damage Rating 1 - Area: BC - BACK END DAMAGE CONCENTRATED IMPACT"
"Vehicle Damage Rating 1 - Direction of Force: 6 - 6 O'CLOCK"
"Vehicle Damage Rating 1 - Severity: 7 - DAMAGED 7 HIGHEST"
The driver wasn't from the area (according to the crash report he's from a suburb of Austin).
Then the accident details...
It happened at 29.44598919, -98.35794899.
"Contributing Factors: DRIVER INATTENTION; PARKED IN TRAFFIC LANE; PARKED WITHOUT LIGHTS"
"Light Condition: 2 - DARK, NOT LIGHTED"
"Roadway Alignment: 2 - STRAIGHT, GRADE"
"Manner of Collision: SAME DIRECTION - BOTH GOING STRAIGHT-REAR END"
"Surface Condition: 1 - DRY"
"Weather Condition: 1 - CLEAR"
"Object Struck: OVERTURNED"
So basically BlueCruise drove (at night on an unlighted freeway section) straight into a black vehicle that was parked in the middle of the road with no lights on hard enough to cause it to overturn and kill the driver.
Isn't this what radar was supposed to solve, unseen objects in the dark?
Something else to note, that area of freeway is labeled as BlueCruise hands free capable.
Thanks!
Re radar: Depends on the signal quality. I don't know what type of radar (or the specs of) the radar that Ford uses. Simple cruise control radars have plenty of noise and at the end of the day there is software that need to make a decision and the OEM often doesn't want false positives...
Radar usually does not detect stationary objects.
Radar can 100% detect stationary objects. It's used as a tool for mapping underground pipes for example.
Green even said Tesla's radars can detect stationary objects, they just filter them out (or have bugs attributing them) and gave an example video.
Yeah, I just assumed others would be using better radar systems after Tesla's bouquet of oopsie daisies and the outrage associated with them removing radar. I just can't find much info on their system as every ford dealer pops up with their marketing page about blue cruise rather than a lot of facts.
diplomat33
Average guy who loves autonomous vehicles
Radar does detect stationary objects. The issue is that radar will bounce off of everything so you will get a lot a extraneous noise from stationary objects that don't matter. This is especially true for older, low resolution radar. So ADAS software will deliberately eliminate all the signals from stationary objects so that the ADAS can focus on the moving objects that really matter. And for the purpose of cruise control where you need to keep a safe distance from a moving car in front of you, that makes sense. But the side effect is that the car will not brake for stationary objects that really do matter, like a stopped car in the middle of the road. If the human driver is paying attention like they are supposed to, then they can intervene to prevent a crash. But if they are not paying attention, then you will get a crash, like we've seen many times.
To address this problem, Tesla opted to get rid of their radar since it was poor resolution anyway and just rely on the camera vision to detect objects. That works but it relies on the camera vision being good enough. And if the camera vision fails to detect the object, there is no back-up sensor. So, the human driver still needs to pay attention. Other car companies are installing imaging radar which is much higher resolution and therefore will be a lot better at separating the noise from the real objects. So imaging radar is a lot better at these scenarios of detecting stationary objects. And they do sensor fusion of both the cameras and the imaging radar to get more reliable perception of moving and stationary objects.
diplomat33
Average guy who loves autonomous vehicles
Considering that the Ford that crashed was a 2022 model, it would almost certainly not be an imaging radar. AFAIK, automative imaging radar is still pretty new and is planned for some 2025+ models.
Thanks, wasn't sure if you brought it up in a "Ford is using this" scenario (as I can't find ANY tech details about bluecruise reliably) or a "this could be a future improvement" scenario. I'll withhold judgement for the tech until I see it.
Also I can't 100% confirm this, but it seems like this may be the front end of the Mach-E in question after the crash.
This info is coming from another forum (including the picture above) but someone mentioned the stretch of road where it happened is "super dark" and has no shoulder.
Apparently (though I can't confirm this from the limited accident report I had access to) the CRV lost it's alternator, as in it fell off the vehicle, and that was the reason for it being stopped in the road. Which would also explain a lack of lights (if the 12V died).
Adding some additional detail as well from the OP of that car photo (that is speculated to be the car in question), that has since been deleted.
The filtering is exactly why most automotive radar is completely blind to stationary objects. The ones with zero phantom braking is basically doing that.
diplomat33
Average guy who loves autonomous vehicles
FYI, I believe that is lidar, not radar. Luminar is a lidar company.
You are correct, I was just adding that in as context for things coming that will apply to situations like this.
diplomat33
Average guy who loves autonomous vehicles
A streetview image at the GPS coordinates show plenty of overhead lights. Those coordinates are near an interchange. Unless the Honda was sitting under the interchange overpass, it should have been lit from above and visible to the driver.
Did the data have the speed of the crash listed? I.e., did the Mach E appear to attempt to slow down?
Similar threads
- Article
