Slashdot Mirror
Ford Engineers Test 'Predictive Logic' To Improve Cruise Control
cartechboy writes "Sometimes what we think of as 'car tech' is colored by sensational coverage of things like autonomous cars. But real engineers are working behind the scenes every day to make existing auto technologies more efficient. Take cruise control: Today, even adaptive cruise systems just throttle up when the car's speed drops and ease off when speed rises or a car gets too close. Today's cruise-control systems aren't predictive--meaning they don't plan ahead. At all. Now, engineers at Ford are adding predictive algorithms and more sophisticated powertain mapping to reduce the built-in overcompensation that ends up wasting fuel. Ford has mapped each vehicle's powertrain in much greater detail, and their prototype control systems look at grade steepness, load on the engine, and other variables every few seconds to predict what's likely coming up. Will the hill level off soon? Will the driver ask for more gas, or let up on the accelerator? Down the road, connected-vehicle and cloud-based data will build on these predictive developments--as will those autonomous vehicles you hear so much about. Think of this as a building block to the future."
I wonder if Bezos will be kind?
First coast!
I see. Is this a way to get around patents by Kosko?
Really? The cruise control in a 10 year old TDI has a D component. Unless I forgot all my control theory that is predictive.
Let's integrate the cruise control and ABS logic into the robot driver!
Some might call that an architecturally flawed design, but those people have been proven to be fucking morons in similar technical discussions.
... if it means the legions of cars who drive 35 on a 75 MPH highway whenever the road goes up a slight grade because they haven't worked out that they've slowed down and are creating a huge bunched up traffic jam for miles behind them, when there road ahead of them is empty.
This is basic control theory. They haven't been doing this all along? Fire the lot of 'em, I say!
... after it has been found, that the cause for a crash of an Embraer 195 in Africa has been linked directly to the captain being alone on the flight deck, putting the engines to idle and descend rapidly... http://www.aeroinside.com/item/3416/lam-e190-over-botswana-namibia-on-nov-29th-2013-captain-intentionally-crashed-aircraft
Still trying to unnecessarily save fuel. This all goes along with stopping global warming that doesn't exist, finding the origin of man which was God.
Why model the power train. Why not instead ship cars with a system designed to learn the characteristics of your power train? It needs to be adaptive anyway to deal with variables like vehicle loading, towing, altitude and wind.
Make them lighter too.
"Honestly, officer, the car thinks I'm my wife going to a salon appointment, and the cruise control was trying to get me there on time!"
"The car sped up just as some blonde bimbo passed me in a Corvette convertible. I have no idea why..."
Feature bloat in cars. Solutions to problems that don't exist.
I give it 2 years before some security researcher is showing how to override the new "adaptive cruise control" and peg the throttle.
Seems like we all use the same roads... if we just log with altitude and accelerometer readings, we can make a 3d model of all the road surfaces, and layer this into the road database. Problem solved.
Slashdot's rate-of-post filter: Preventing you from posting too many great ideas at once.
Because the variability is minimal in a EFI system. You can just code a LUT. This is how your engine works today anyway.
Slashdot's rate-of-post filter: Preventing you from posting too many great ideas at once.
You must have because no 10 year old car has a predictive cruise. Dynamic is not predictive. Even a brand new European Touareg (not US) has adaptive cruise, which is still not predictive.
Ford really isn't this far behind. From the results I've seen in the heavy trucking industry, 'predictive logic' isn't all that. Creating a responsive powertrain with a few set conditions will probably lead to a product a customer wants to drive, not something which will make a customer second guess the system.
I haven't driven an American car in awhile, but I always remember the CC systems to have terrible wind up in the system. This wind up would give the prius a run for the money.
Down the road, connected-vehicle and cloud-based data will build on these predictive developments
It's cruse control for frack sakes, not global weather predicting. Cars/trains/planes/boats have had mechanical/electronic cruise control devices since the first road trip made holding the pedal down boring and they've worked successfully for decades. Why does it now need the power of 'the cloud' to do this job???
After reading this article (yes, I actually read the article and the deposition) I think automobile software should be reviewed by anybody that wants to review it. Let's face it, the software may be somewhat sophisticated, but it shouldn't be rocket science. Certain algorithms could even be patented for all I care, but the code quality must be reviewed. For those of you that haven't read the deposition from the link above, the upshot is that the expert witness saw horrible software practices being performed in a vacuum - as it were - he couldn't even take a pen and notepad in the room where he could view the software. He had to exit the room, make notes, then come back in after a security screening. This is the worst kind of software, and people are driving with it every day. Until software that has my life in its hands is peer/public reviewed, I'm going to buy only older cars for as long as I can. I sold my Prius after reading the above article.
It bad for fuel economy to let the cruise control slam on the gas to keep the speed up
The derivative term in a P[I]D control loop is predictive.
Car manufacturers misusing standard terminology is not my problem.
That's the great thing about "standard terminology", there are so many standards to choose from. One industry's standard may mean something different in another. Cruise control PIDs aren't really predictive anyway. They are simply responding to the rate of change of error.
On my recent model F150 there's a very handy feature: "Tow/Haul" mode. It changes the transmission shift behavior so that touching the brakes makes it downshift and STAY downshifted (and doing it again downshifts another step on the many-gear-ratios transmission), while touching the accelerator lets it shift back up. This is GREAT for long downgrades in mountains, even if you're NOT towing.
But it doesn't interact well with the cruise control. The speed control raises the throttle setting to keep you from going under the setpoint. But when you're over it just goes down to idle and lets your speed runaway. Touch the brakes to enable engine braking and the cruise control disengages. No automatic speed control for YOU on the downgrade. When the grade starts to level out the speed drops, and even before that you're back to watching the speedometer and doing manual adjustments,
Result: On mountain roads you're constantly disengaging and re-engaging the speed control.
They should integrate the two: When tow/haul mode is engaged, the speed control should send downshift signals to the transmission to control too-fast as well as too-slow conditions.
When tow-haul is on the speed control should signal the transmission to downshift when necessary, to keep the speed from running away and requiring the driver to brake. (The speed control's acceleration when too-slow will handle the upshift correctly.)
You'll still have to touch the brakes or tap-down the speed setting for curves and other locally slow zones. But then you'd just hit "resume" or tap-up the setting. Meanwhile the automation would handle the non-exceptional condition of preventing overspeed and runaway on downgrades.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
I just want a cruise control that doesn't go to 100% throttle when I want to resume 70MPH after slowing down.
Since I gave up hope, I feel much better.
Traffic would be less of a problem if more people learned how to use cruise control. It's an easy way to maintain a constant speed and increase fuel efficiency.
I mostly drive American cars but have noticed that Euro cruise control has long been quite smooth. In my second last American car the cruise control was so twitchy that my wife would regularly ask what the hell I was doing. My last American car was still jerky. Typically the event that concerned her was the stupid car not gassing it enough on a hill climb and then stomping on the gas and dropping a gear to compensate for the great loss of speed. My other complaint was that for some stupid reason the cruise control would still leave a little gas on during the decent resulting in the car going way too fast. I see the RPMs still up a bit and then would turn the cruise off and see a 50% drop in the power. Lucky to not get a speeding ticket with that gem of a feature.
So while it is good ford is making it sound like they are leaping into the future, step one should be catching up with 12 year old Euro technology.
My 2002 Honda CRV (still kicking with 200K+ miles and 12 years on her) has a sensor that measures the incline of the hill I'm on. It adjusts the cruise control in response to the hill and doesn't upshift while still on the incline, the way so many other cruise controls do. Way smarter, 12 years ago.
I will say that I continue to give the car plenty of help with predictive stuff (like turning off overdrive, forcing it to downshift, before it loses speed on the hill), but I really, REALLY don't want the car trying to do that stuff for me. I've worked in computers for 20 years, on everything from medical systems to web and mobile applications, and I don't trust computer software with anything that can kill me. Frankly, those of us who program aren't nearly rigorous enough to make sure we don't get it wrong with catastrophic effects. Just ask the guys who program Toyota's accelerators.
My Toyota Sienna (2006) and Camry (2007) also downshift when you brake on a grade (though I don't know about how long they stay in the lower gear - it could only be until you let go of the brake) and they also don't do this with the cruise control. I would love it if they did. It's not just pickup trucks and/or towing that this would be great for.
"Will the driver ask for more gas, or let up on the accelerator? "
If he's using the cruise control....no...and, no.
If they know the current:
- Previous Orientation of vehicle
- Current Orientation of vehicle
- Previous speed
- Current Speed
Simple Method =
Use the previous speed against current speed. Predict what the new thottle level will need to be to reach their target (how cruise control should work)
But theres some hills =
They can simply make an additional function to increase/decrease throttle based on a angle prediction using the previous and current orientations.
This is simple logic, i just dont understand their need to overcomplicate this for sake of mentioning the "cloud".
I remember years ago when you wanted a drivers license you actually had to learn how to drive a car. Now all you have to do is learn how to point it in the direction you want to go.
Wuddooeyeno? IITYWYBMAD? Like nuts? eclecticallyincorrect.com
Maybe you should try a Dodge truck. My Charger downshifts without me doing anything if idle throttle results in not enough braking force down a grade. It always holds the speed to +/- 1mph. It also has adaptive cruise - the one option I appreciate the most, times 10. It really shines on a rainy night, when the guy in front hits his brakes enough to light the brake lights but not enough to slow his speed. It doesn't slow down unless he really does slow down, and without it I'd be tapping my brakes whenever I see his light up. And it's collision warning has saved me once from a fender bender, or maybe worse. I can't say enough about the software in my Charger; everything it does is just what I'd want it to do.
If I didn't have absolutely NOTHING to do, I wouldn't be here.
I'm still waiting for one that sets the cruise-control's speed limit automatically from my GPS-Navigator instead of just beeping when I drive too fast.
How hard can that be when it's all in the same computer. Why do we still have to do that manually every couple of miles?
Why bring up that point.
If the road database is not correct, the only result is that some extra fuel is burned because the predections were wrong. En even then i think that not a lot is won this way. (compare the combusion engine efficiency of a Volt and a prius and you can see ford still has a huge gap to fill)
The driver still is fully responsible.
The fully automated prototypes still have lot of scanners and cameras to do the 100% automatic driving.
...that this story was followed with "can you trust your router" -- my router can't kill me!
Everyone else using cruise control is tailgating down a grade, forcing them to brake and disengage theirs.
...works better than is described by the story blurb.
Acceleration when resuming or when a vehicle going slower than the setpoint moves or exits my lane is pretty smooth as is deceleration when overtaking a slower vehicle.
I only have two complaints. I wish there was an audible indicator when driving slower than the setpoint. On long interstate drives I occasionally find myself behind someone driving slower than my setpoint but because the deceleration is so subtle it's easy to not notice.
The other is icing on the radar panel in bad weather causing the cruise to not work at all. It'd be great to have it heated so that it resists icing and possibly a way to run it in manual mode like normal cruise control and bypass the adaptive aspect.
On my recent model F150 there's a very handy feature: "Tow/Haul" mode.
Welcome to the 1990s, when that started to show up on vehicles. Our 2000 Astro LS has the same feature. It changes the shift points, that's about it.
Result: On mountain roads you're constantly disengaging and re-engaging the speed control.
Sigh. It says right in the owner's manual that the cruise control isn't for that.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
That's not "Adaptive" in the sense that they're talking about here in my opinion. That's just an offshoot of what some companies will call radar cruise, to where the car maintains a set position behind the car in front of you. I like the premise though, I'm sure it works rather well - and on long highway drives, I wish I had it.
In my personal opinion, I would think an adaptive cruise would work similar to what I do to attempt to get a leg up on milage. I have a car with a pretty healthy v6 in it - and it also has an instantaneous MPG gauge. The hills around here aren't many, but if I control my speed +/- 10 mph to the limit (up to -10 on a long uphill, +10 on a downhill) I can keep my milage in a much better range than the cruise control attempting to limit me to +/-2 mph when set. Over the course of a tank, I've seen a 2-3mpg gain, which translates into real (albeit a small amount of) money back into my pocket.
Karnal
So you're one of THOSE guys, eh? :)
If I didn't have absolutely NOTHING to do, I wouldn't be here.
But real engineers are working behind the scenes every day to make existing auto technologies more efficient. ...
Think of this as a building block to the future.
Did Ford marketing write this?
Real engineers...working behind the behind the scenes every day... creating building blocks of the future.
That could be my wordy, but non-informative job description.
Adding, some "Airplane",
Real engineers bust their buns every day creating building blocks of the future. You tell your old man that.
Everytime I hear the cloud knowing anything about my car, or my car using anything in the cloud, I shudder.
Cruise control helps with the biggest expense in trucking- fuel. However, such a heavy vehicle behaves very differently in hills, and cruise is a detriment there. As far back as 2000, Freightliner and International have been working on predictive cruise control that can adjust engine speed, gear (in automatics) and such to the load of the truck and the grade it faces. likewise it will select the best gear to descend that hill safely and lock in to it. The last case I saw on it was trying to integrate GPS and Magellan maps into the computer control to accomplish this. thus far, nothing has come of it- technology changes/improves before they have a stable prototype so they keep starting over.
control theory that is predictive
control theory (assuming you mean transfer functions and bode plots etc) isn't predictive; its reactive (based on continuous feedback)
i think what TFA is talking about (who bothers reading it) is using GPS+grade contour maps and triple-axis accelerometers to help the system plan in advance for changes in grade, because most with a bit of cruise control experience in hilly terrain can attest that often when you reach the crest of a hill the cruise will overcompensate a bit (sometimes requiring a bit of braking so as not to go over the speed limit too much)