Gen 3 Insight Forum banner

1 - 5 of 5 Posts

·
Super Moderator
Joined
·
4,098 Posts
Discussion Starter #1
AAA just completed a study of pedestrian detection systems, and reports that they are mostly ineffective, perform inconsistently, and are more likely to fail at nighttime.

Tested conditions included 'blind' approaches (vehicle turning right into path of adult and child darting into traffic), two adults standing in road, day and night conditions, and speeds of 20-30 mph. The 4 cars tested were the 2019 Chevy Malibu, 2019 Honda Accord, 2019 Tesla Model 3, and 2019 Toyota Camry.

During daylight, the Advanced Driver Assistance Systems (ADAS) avoided crashes 40% of time at 20 mph but mostly failed at 30 mph. During nighttime, no system detected the simulated pedestrians at night. In the scenario with 2 adults standing in road, it seems the system gets 'confused' because one adult is to the side (ignored) and the other is in the road.

This AAA link has path to the full research report, summary fact sheet, and "b-roll" video of the tests performed:
https://newsroom.aaa.com/2019/10/aaa-warns-pedestrian-detection-systems-dont-work-when-needed-most/

And this news clip nicely recaps the learnings:
 

·
Registered
Joined
·
836 Posts
This should support Verdiers claims... at 30mph, his car rarely "detects" a stationary object. Also lends credence to my theory that the system is more active to moving objects, although they didn't test it...
 

·
Super Moderator
Joined
·
4,098 Posts
Discussion Starter #3
What I found compelling is that tests suggest these failures were true across all four systems (Chevy, Honda, Toyota, Tesla). We know that Honda outsourced the development of their ADAS system (thru Bosch), per the the WSJ article shared by andrew28. Yet the failure isn't unique to Bosch-related systems.

Too bad AAA didn't include Ford's system in testing, to allow for comparison. The article below lists key development companies and manufacturer partnerships for "future innovation" on ADAS systems. It shows Ford affiliated with 3+ companies, while other manufacturers link to 1. Ford seems very dedicated to having the best technology in this area, which supports Verdier's experience with Ford "Co-Pilot 360" being better than Honda "Sensing." Hopefully AAA will expand future testing to show this as a differentiator.

https://medium.com/@chelsie.may/top-9-adas-software-development-companies-for-automotive-9699bd772303
Electrobit (Germany) = Ford, Volkswagen, Audi
Continental Automotive = Volkswagen, Ford, Honda, GM
FAAR (France) = Nissan, Ferrari
Harman (US) = Toyota, BMW, Chrysler, Lexus, Mercedes
AISIN (Japan) = Mazda, Nissan, GM
Green Hills (US) = Mitsubishi
Wabco (Belgium) = Ford, GM, Hyundai​
 

·
Super Moderator
Joined
·
2,713 Posts



I think the cars feel the same way that we secretly do when we see people like the above crossing the street whenever and wherever they feel like it. ;)
 

·
Super Moderator
Joined
·
4,098 Posts
Discussion Starter #5
I think the cars feel the same way that we secretly do when we see people like the above crossing the street whenever and wherever they feel like it. ;)
Those crazy jaywalkers... the second picture even includes a child (wearing purple) who's on the side nearest traffic.

But just so we're all on the same page, jaywalkers = target practice, right? (j/k!!)

AAA's test results mention that SUVs passed these same tests, while sedans did not. There's something to be said for SUVs being able to mount the camera and radar system even a few inches higher, to allow for better range/view of risks than on sedans.
 
1 - 5 of 5 Posts
Top