Issue with speed sign recognition system

[scazea]

Thanks for your suggestions. I also have issues when I come up behind a very slow driver.

Don't we all!!!

Sponsored

 

[Donj02]

It's far from perfect, my F150 ironically seems to do better most of the time. My Mustang will read a bridge weight limit sign as speed limit too. So if it says 40 tons it interprets that as 40mph. I've had it do the opposite too driving on cruise at 35 and had it ramp up to 65 after seeing a 65 ton sign.

 

[John G]

It's far from perfect, my F150 ironically seems to do better most of the time. My Mustang will read a bridge weight limit sign as speed limit too. So if it says 40 tons it interprets that as 40mph. I've had it do the opposite too driving on cruise at 35 and had it ramp up to 65 after seeing a 65 ton sign.

That is good to know, thanks for your post.

 

[roadpilot]

I've got nearly 60K miles on my 21 F150 and it's never read anything but a legit speed limit sign as a speed limit sign. I only have a few hundred miles on the DH, so I haven't even really used ACC all that much yet.

 

[John G]

I've got nearly 60K miles on my 21 F150 and it's never read anything but a legit speed limit sign as a speed limit sign. I only have a few hundred miles on the DH, so I haven't even really used ACC all that much yet.

We have a 2020 Ford Escape, my wife's vehicle, and we have not had any issues with the display regarding the speed limit sign. So, I know where you are coming from.

Congratulations on your Dark Horse! Hopefully, this will not be an issue with your new Mustang.

 

[DontHitMeSafetyYellow]

I prefer to drive my cars vs let them drive me…

yeah… as stated previously I do work in robotics research and autonomy. A few years back work sent me to an AUVSI Autonomous Vehicle Symposium. I walked out of there deathly afraid of autonomous cars. I know WAY too much about how they do and don’t work and where the pitfalls are and limitations.

I’m all for driving aids (things like lane departure warnings, emergency braking, ABS, TCS, etc. things that help a driver but requires them to still actively drive the car.

Where I get concerned is all these techs that allow the driver to not have to really pay attention to driving, autopilots, etc. particularly when the argument is given that if the autonomy fails for whatever reason, the driver can just take back manual control. Ok sure that’s somewhat reasonable IF: 1) the driver is experienced (new drivers don’t have this luxury and if they are only ever driving cars that auto drive they will never build this skill set) and 2) are paying attention and ready to step in, which over time they will trust the autonomy so much that they WONT be paying attention.

Case in point - look what happened to that Air France flight that went down over the Atlantic back in the early 2000s. Very experienced flight crew, flying on AP, were having a long conversation in the cockpit and not really paying attention to the instruments for a while. Pitot tube froze over causing the instruments to show bad data. Autopilot disengaged because it was getting bad data. Pilots were totally confused by what was happening and kept pitching the nose down trying to gain airspeed because the airspeed indicator was showing low airspeed when in fact it was frozen over and they put the plane into a nose dive all the way down from 40k ft. You can’t expect people to quickly reengage and reorient themselves in an emergency situation when they have been totally disengaged.

Dont even get me started on concerns over software updates, software security, or maintenance of all these sensors (what happens when somebody replaces a faulty sensor with a cheaper non-OEM one? Does the software know it’s potentially getting crappier data?) what happens when a sensor fails while autonomy is driving? Does it just pull over? Does the car stop suddenly in the middle Of traffic? How does it know HOW to “pull over” if it’s a sensor critical to doing that maneuver?

Yeah, no, I’ll stick to driving myself thank you. Machine learning is impressive but it is NOT foolproof. You will NEVER get me in a Tesla driving under Autopilot…. More like auto coffin.

 

[Upacurb]

yeah… as stated previously I do work in robotics research and autonomy. A few years back work sent me to an AUVSI Autonomous Vehicle Symposium. I walked out of there deathly afraid of autonomous cars. I know WAY too much about how they do and don’t work and where the pitfalls are and limitations.

I’m all for driving aids (things like lane departure warnings, emergency braking, ABS, TCS, etc. things that help a driver but requires them to still actively drive the car.

Where I get concerned is all these techs that allow the driver to not have to really pay attention to driving, autopilots, etc. particularly when the argument is given that if the autonomy fails for whatever reason, the driver can just take back manual control. Ok sure that’s somewhat reasonable IF: 1) the driver is experienced (new drivers don’t have this luxury and if they are only ever driving cars that auto drive they will never build this skill set) and 2) are paying attention and ready to step in, which over time they will trust the autonomy so much that they WONT be paying attention.

Case in point - look what happened to that Air France flight that went down over the Atlantic back in the early 2000s. Very experienced flight crew, flying on AP, were having a long conversation in the cockpit and not really paying attention to the instruments for a while. Pitot tube froze over causing the instruments to show bad data. Autopilot disengaged because it was getting bad data. Pilots were totally confused by what was happening and kept pitching the nose down trying to gain airspeed because the airspeed indicator was showing low airspeed when in fact it was frozen over and they put the plane into a nose dive all the way down from 40k ft. You can’t expect people to quickly reengage and reorient themselves in an emergency situation when they have been totally disengaged.

Dont even get me started on concerns over software updates, software security, or maintenance of all these sensors (what happens when somebody replaces a faulty sensor with a cheaper non-OEM one? Does the software know it’s potentially getting crappier data?) what happens when a sensor fails while autonomy is driving? Does it just pull over? Does the car stop suddenly in the middle Of traffic? How does it know HOW to “pull over” if it’s a sensor critical to doing that maneuver?

Yeah, no, I’ll stick to driving myself thank you. Machine learning is impressive but it is NOT foolproof. You will NEVER get me in a Tesla driving under Autopilot…. More like auto coffin.

AMEN - I agree 10000% with everything you just wrote.....I have no where near the technical background that you do.......but based on the little research I have done.....I'm with you.....

 

[John G]

yeah… as stated previously I do work in robotics research and autonomy. A few years back work sent me to an AUVSI Autonomous Vehicle Symposium. I walked out of there deathly afraid of autonomous cars. I know WAY too much about how they do and don’t work and where the pitfalls are and limitations.

I’m all for driving aids (things like lane departure warnings, emergency braking, ABS, TCS, etc. things that help a driver but requires them to still actively drive the car.

Where I get concerned is all these techs that allow the driver to not have to really pay attention to driving, autopilots, etc. particularly when the argument is given that if the autonomy fails for whatever reason, the driver can just take back manual control. Ok sure that’s somewhat reasonable IF: 1) the driver is experienced (new drivers don’t have this luxury and if they are only ever driving cars that auto drive they will never build this skill set) and 2) are paying attention and ready to step in, which over time they will trust the autonomy so much that they WONT be paying attention.

Case in point - look what happened to that Air France flight that went down over the Atlantic back in the early 2000s. Very experienced flight crew, flying on AP, were having a long conversation in the cockpit and not really paying attention to the instruments for a while. Pitot tube froze over causing the instruments to show bad data. Autopilot disengaged because it was getting bad data. Pilots were totally confused by what was happening and kept pitching the nose down trying to gain airspeed because the airspeed indicator was showing low airspeed when in fact it was frozen over and they put the plane into a nose dive all the way down from 40k ft. You can’t expect people to quickly reengage and reorient themselves in an emergency situation when they have been totally disengaged.

Dont even get me started on concerns over software updates, software security, or maintenance of all these sensors (what happens when somebody replaces a faulty sensor with a cheaper non-OEM one? Does the software know it’s potentially getting crappier data?) what happens when a sensor fails while autonomy is driving? Does it just pull over? Does the car stop suddenly in the middle Of traffic? How does it know HOW to “pull over” if it’s a sensor critical to doing that maneuver?

Yeah, no, I’ll stick to driving myself thank you. Machine learning is impressive but it is NOT foolproof. You will NEVER get me in a Tesla driving under Autopilot…. More like auto coffin.

You brought up some very good points. Thank you for sharing them.

I don't use these features all of the time, and often I turn them off such as when I am entering city limits, heavy traffic situations, or going around curves. Those are instances where I want to be in control. Most of the time I use this technology when I am traveling long distances in open areas.

Sponsored