Team-BHP > Technical Stuff
Register New Topics New Posts Top Thanked Team-BHP FAQ


Reply
  Search this Thread
50,677 views
Old 14th February 2013, 10:30   #31
Senior - BHPian
 
sarathlal's Avatar
 
Join Date: Apr 2012
Location: Trivandrum
Posts: 1,831
Thanked: 6,239 Times
Re: How fast ECU learns the driving style?

Reviving an old thread in the hope that there could be little more of an information by this time.

During my last servive, the SA's had to unplug the battery and because of which the ECU got reset. Noticable differences are;
1. average mileage values went blank
2. Brakes are lot harder now
3. Accelerations are little bit jerky
4. Downshifts happen quickly, it is very difficult to control the highly responsive car now. The moment I press the accelerator, the car is darting away!

Am confused mainly because I am surprise that a single battery disconnection will cause the car to lose these much information.
Can you please guide to let know, whether these behaviour is normal or I should be worried?

EDIT: The car in question is the Civic.

Last edited by sarathlal : 14th February 2013 at 10:56.
sarathlal is offline   (1) Thanks
Old 14th February 2013, 10:42   #32
Senior - BHPian
 
Gansan's Avatar
 
Join Date: May 2009
Location: Chennai
Posts: 4,625
Thanked: 5,917 Times
Re: How fast ECU learns the driving style?

My car's battery was also kept disconnected for a couple of days last May when they were fixing a bang-up that happened in the MASS. The only difference I noticed was the HU of the audio system went blank and I had to tune / store the FM channels once again! There was absolutely no other difference in the driving.

May be it is because Altos are not so intelligent to begin with, perhaps?!
Gansan is offline  
Old 14th February 2013, 11:59   #33
Senior - BHPian
 
alpha1's Avatar
 
Join Date: Apr 2007
Location: LandOfNoWinters
Posts: 2,178
Thanked: 3,008 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by sarathlal View Post
Reviving an old thread in the hope that there could be little more of an information by this time.

During my last servive, the SA's had to unplug the battery and because of which the ECU got reset. Noticable differences are;
1. average mileage values went blank
2. Brakes are lot harder now
3. Accelerations are little bit jerky
4. Downshifts happen quickly, it is very difficult to control the highly responsive car now. The moment I press the accelerator, the car is darting away!

Am confused mainly because I am surprise that a single battery disconnection will cause the car to lose these much information.
Can you please guide to let know, whether these behaviour is normal or I should be worried?

EDIT: The car in question is the Civic.
As far as I know the ECU doesn't control the brakes or the automatic transmission (does your Civic has auto?)

Looks more like a case of placebo.
alpha1 is offline   (2) Thanks
Old 14th February 2013, 12:23   #34
BHPian
 
Prowler's Avatar
 
Join Date: Jul 2008
Location: Madras
Posts: 798
Thanked: 1,403 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by sarathlal View Post
1. average mileage values went blank
2. Brakes are lot harder now
3. Accelerations are little bit jerky
4. Downshifts happen quickly, it is very difficult to control the highly responsive car now. The moment I press the accelerator, the car is darting away!


Can you please guide to let know, whether these behaviour is normal or I should be worried?

EDIT: The car in question is the Civic.
Other than the brakes, everything else points to the ECU being reset. It is nothing abnormal. It takes a few days to 'learn'. When they updated my ECU, I had similar symptoms except that mine is a manual. The Engine was 'eager' for a few kms and then settled down.
Prowler is offline   (2) Thanks
Old 15th February 2013, 21:54   #35
Senior - BHPian
 
Join Date: Jul 2009
Location: Calcutta
Posts: 4,668
Thanked: 6,233 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by sarathlal View Post
2. Brakes are lot harder now
Has the idling rpm gone up?

Regards
Sutripta
Sutripta is offline  
Old 15th February 2013, 22:27   #36
BHPian
 
Join Date: Apr 2009
Location: .
Posts: 489
Thanked: 132 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by Gansan View Post
My car's battery was also kept disconnected for a couple of days last May when they were fixing a bang-up that happened in the MASS. The only difference I noticed was the HU of the audio system went blank and I had to tune / store the FM channels once again! There was absolutely no other difference in the driving.

May be it is because Altos are not so intelligent to begin with, perhaps?!

Less the better in most cases .
YaeJay is offline  
Old 15th February 2013, 22:57   #37
Senior - BHPian
 
sarathlal's Avatar
 
Join Date: Apr 2012
Location: Trivandrum
Posts: 1,831
Thanked: 6,239 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by Sutripta View Post
Has the idling rpm gone up?

Regards
Sutripta

Nopes, they are all the same. Initially at 1.2k and once the engine warms up, it rests at .8, which was how it was earlier as well.
sarathlal is offline  
Old 15th February 2013, 23:41   #38
A M
BHPian
 
A M's Avatar
 
Join Date: Aug 2010
Location: Panchkula
Posts: 558
Thanked: 660 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by sarathlal View Post
Reviving an old thread in the hope that there could be little more of an information by this time.

During my last servive, the SA's had to unplug the battery and because of which the ECU got reset. Noticable differences are;
1. average mileage values went blank
2. Brakes are lot harder now
3. Accelerations are little bit jerky
4. Downshifts happen quickly, it is very difficult to control the highly responsive car now. The moment I press the accelerator, the car is darting away!

Am confused mainly because I am surprise that a single battery disconnection will cause the car to lose these much information.
Can you please guide to let know, whether these behaviour is normal or I should be worried?

EDIT: The car in question is the Civic.
Congrats!! It just saved you 20000/- you were thinking to spend for a remap.
Now you only need a FFE and revv away to glory.
A M is offline  
Old 16th February 2013, 08:17   #39
Distinguished - BHPian
 
Join Date: Oct 2012
Location: Delhi
Posts: 8,906
Thanked: 61,493 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by sarathlal View Post
Reviving an old thread in the hope that there could be little more of an information by this time.

During my last servive, the SA's had to unplug the battery and because of which the ECU got reset. Noticable differences are;
1. average mileage values went blank
2. Brakes are lot harder now
3. Accelerations are little bit jerky
4. Downshifts happen quickly, it is very difficult to control the highly responsive car now. The moment I press the accelerator, the car is darting away!

Am confused mainly because I am surprise that a single battery disconnection will cause the car to lose these much information.
Can you please guide to let know, whether these behaviour is normal or I should be worried?

EDIT: The car in question is the Civic.
When you disconnect the battery there are a few things that could happen.
First, on some, in particular older, cars nothing would happen. The older electronic systems did not reset at all. If you want to change engine behaviour you would have to reprogram the chip. (sometimes also known as flashing, remapping etc.)

On most modern cars the ECU once resetted by disconnecting the battery needs to run through a number of test cycles. These test cycles verify that the engine, all of its sensors, and its programs run correctly. On most modern cars during the test cycles the ECU also optimizes/calibrates the performance of the engine, mostly from an ignition/emission perspective.

There are also a few other things that are unique to each car and it's respective ECU needs to adjust to it. Mainly tolerances in sensors and actuators. For instance, the automatic choke coming in or out. Will be slightly different on each cars and it varies over time due to wear and tear. The ECU will figure it all out automatically!

Depending on the car / ECU, it could take up to 60 - 80 kilometers to run through the test cycles. Many people report that the car won't run smoothly during these test cycles. Could be as the ECU needs to calibrate for instance the fuel trim.

The only way to verify that the test cycles have been completed successfully is to hook up an OBD code reader and check. Most ECU will throw out a P-code, for instance P-1111 on Jaguar means all test completed successfully.

So here's another anorak fact; On a car that runs perfect, you will always have a few codes. Those that show the test cycles were concluded successfully.

After the test cycles have been completed the ECU will continue to optimize the various parameters. It never stops. Its just that disconnecting the battery will reset a number of input parameters stored in the ECU to default values. It needs a bit of time to get them optimized.

Disconnecting the batteries will as far as I know always reset the trip computer, e.g. average miles, fuel used etc. The only value that remains is the odometer. There are some cars / ECU where disconnecting the battery might not fully reset the ECU. I seem to recall reading that for instance on Toyata this is the case. Means, when the battery is reconnect, it doesn't need to go through the test cycles

Now, next to to test cycles, that could optimize engine performance a bit, modern ECU also are able to learn.

What your typical ECU is able to adjust for personal driving style is mainly throttle response and gear shift points. So it tends to be most noticeable when you drive in heavy traffic. I.e. on the motorway won't make a blind bit of difference. There is endless debates on cars forum the world over on how big a difference these 'adaptive/learning" ECU's make.

Not sure how quickly it adapts, but I would think it will take several hundreds of kilometers as running average, so adjusting is very slow.

My own thoughts are that the more powerful your engine is the less likely you are to notice it at all in the first place.

As I said earlier, when the ECU has been reset you might notice a bit of rough running at first. But within 100 kilometers that should all be back to normal. If it doesn't something is wrong, get somebody to read the codes! Everything else is guessing or trial and error at best.

In this particular case I would say that it is normal to lose the average mileage.
Throttle response and shift points are likely to be adjusted during the test cycles. I have no immediate thought about the brakes being different. I have no idea if there is anything that is controlled, brake wise, through the ECU.

Jeroen
Jeroen is offline   (2) Thanks
Old 16th February 2013, 14:12   #40
Senior - BHPian
 
Gansan's Avatar
 
Join Date: May 2009
Location: Chennai
Posts: 4,625
Thanked: 5,917 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by YaeJay View Post
Less the better in most cases .

There is a saying which goes "The most advanced technology should also be the most simple!"
Gansan is offline   (1) Thanks
Old 18th February 2013, 12:34   #41
Distinguished - BHPian
 
Join Date: Oct 2012
Location: Delhi
Posts: 8,906
Thanked: 61,493 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by Gansan View Post
There is a saying which goes "The most advanced technology should also be the most simple!"
Too true. And with cars that has been the case. But you do need to look at it over a longer period of time.

Not to long ago every car would have a manual choke. That has all been automated. No manual choking anymore

Shifting gears; Thirty years agao you would be hard pressed to find an autobox in aunything other than a US Car. These days they are available on just about any car, from the very small and basic to the super and luxirious segment.

On my car I have auto lights and auto wipers. Never even have to think of switching on anything. Electric windows, I don't need to wind a crank, just push a button. Same for the open roof. touch of the button versus prolonged turning an ugly handle poking out of the roof

Think about ignition advance. Used to be manual on the very first cars. Technology made it automatic, throuogh various steps, mechanical, pneumtatic and these days electronic advance.

The list goes on and on.

Although many would argue different, but I do believe cars have become much more simple and more reliable over time. We just tend to forget how it was twenty years ago. Unless, like me you own a few classic cars and you can still experience all this unreliabillity as a hobby!

Jeroen
Jeroen is offline  
Old 19th February 2013, 14:26   #42
BHPian
 
Join Date: Mar 2011
Location: Bangalore
Posts: 35
Thanked: 7 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by sgiitk View Post
An adaptive ECU! this is news to me as well. An ECU will keep track of things like knock (adjust the timing), filter (max. revs to give you), etc. but driving style ??
i agree. no ecu learns your driving style or anything. simple placebo effect. if any change at all the reset itself must be making a small difference
jayeshrc is offline  
Old 19th February 2013, 19:22   #43
Distinguished - BHPian
 
Join Date: Oct 2012
Location: Delhi
Posts: 8,906
Thanked: 61,493 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by jayeshrc View Post
i agree. no ecu learns your driving style or anything. simple placebo effect. if any change at all the reset itself must be making a small difference
I dont want to be caught in semantics so bear with me whilst I try and explain. It does "learn" from your driving style. It for instance, monitors how agressively or calmly you use your right foot. Throttle response is definitely one of the parameters that in some cars is being adjusted, based on how you drive. Same thing for the shift pattern. If you drive agressively, on some cars the computer will adjust the shift pattern to accomodate that. Ie shift a little later, keep up the rpm a bit more, throttle up more quickly etc.

Whether you call that learning is semantics. But it does (for some cars) adjust the cars/engine/autobox response to a certain driving style.

Jeroen
Jeroen is offline   (2) Thanks
Old 19th February 2013, 19:30   #44
BHPian
 
Join Date: Mar 2011
Location: Bangalore
Posts: 35
Thanked: 7 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by Jeroen View Post
I dont want to be caught in semantics so bear with me whilst I try and explain. It does "learn" from your driving style. It for instance, monitors how agressively or calmly you use your right foot. Throttle response is definitely one of the parameters that in some cars is being adjusted, based on how you drive. Same thing for the shift pattern. If you drive agressively, on some cars the computer will adjust the shift pattern to accomodate that. Ie shift a little later, keep up the rpm a bit more, throttle up more quickly etc.

Whether you call that learning is semantics. But it does (for some cars) adjust the cars/engine/autobox response to a certain driving style.

Jeroen
No. This is not true. For example ill use personal experienxe. I've worked with a motec m800, which is right up there when you're talking about ecus and it definitely follows exactly what you program and nothing more. There is a main table and then compensations based on variables like the environment and so on and so forth, all programmed again, by the company.

So far there hasn't been too much development in terms of neural networks or adaptive processing for ecus that we see.
jayeshrc is offline  
Old 19th February 2013, 22:07   #45
Distinguished - BHPian
 
Join Date: Oct 2012
Location: Delhi
Posts: 8,906
Thanked: 61,493 Times
Re: How fast ECU learns the driving style?

Quote:
Originally Posted by jayeshrc View Post
No. This is not true. For example ill use personal experienxe. I've worked with a motec m800, which is right up there when you're talking about ecus and it definitely follows exactly what you program and nothing more. There is a main table and then compensations based on variables like the environment and so on and so forth, all programmed again, by the company.

So far there hasn't been too much development in terms of neural networks or adaptive processing for ecus that we see.
As I said before, lets not get caught in semantics. You're taking this way to far. Nothing to do with neural networks. What I described is very straightforward mathematics with some clever algorithms thrown in. Maybe even some simple fuzzy logic, which has been available since the late 80s. Just about every modern ECU has fuzzy logic in. I doubt your Motec products have it. I had a quick look at their site and it seems to me these are pretty basic ECU's that use only fixed mapping as you point out, because they are used for specific cars and or drivers. No need to adapt, you want to optimize it once. You tell me, you're the expert on their products.

To boldly go where you have not gone before I suggest you google:

"adaptive shift logic".

You will get 2.400.000 hits all explaining to you from various Car manufactures, transmission manufactures, media articles and a lot of car forums that there are lots of gearboxes with their associated electronics, indeed, remarkably, do adapt to particular driver style.

E.g. from Wikipedia on the DSG autobox:

Quote:
The actual shift points are determined by the DSG's transmission ECU, which commands a hydro-mechanical unit.[2] The transmission ECU, combined with the hydro-mechanical unit, are collectively called a "mechatronics"[2] unit or module. Because the DSG's ECU uses "fuzzy logic", the operation of the DSG is said to be "adaptive"; that is, the DSG will "learn" how the user drives the car, and will progressively tailor the shift points accordingly to suit the habits of the driver.
Unquote
See: http://en.wikipedia.org/wiki/Direct-Shift_Gearbox

So all VW's, Audi and Skoda that use the DSG box have this! I know the mechatronics sounds a bit SciFi, but that can't be helped. Even a very reputable manufacturer as ZF uses this term! Has been available on the ZF 6HP autobox as provided in many BMW's, Mercedes and Jaguars from the late 80's onward.

This is just on the shift pattern. And the same can be found for advance timing, throttle response etc.

Now, I'd be the first one to agree you should not believe anything you find on the internet. However, given that many of the google results point to very reputable parties, I'd say that the chance that this is something akin to the Nasa Moon landing hoax, are less then remote. If you believe the Nasa Moon Landing was indeed a hoax I have a bridge for sale you should have a look at!

Now, if you want to argue that the real advantage of all these electronic gadgety is difficult to quantify, I'd be the first to agree and have already said so. In those 2.400.000 hits you will find various test reports, often from car magazines that write about the lack of finding real practical advantage of all this stuff. So be it. From a technical point of view car manufacturers and their various OEM part manufacturers have been at it for more than twenty years. The first patents on adaptive shift logic are over twenty years old, for those of a more Anorak disposition.

By the way, combining adaptive shift logic with neural networks is pretty old school as well. The industry has been producing papers on this for more than a decade:

See http://www.sciencedirect.com/science...8943049500040E

In all honesty, I don't know if it ever went into real life production. But I suspect it might be down to semantics again.

On my Alfa Romeo Spider with it Bosch L-tronic there is nothing adaptive. Its all hard coded. If I want to change anything, I either swap the chip, or put the car on a rolling road and have the various curves adjusted/optimized. Again, all hard coded. Nothing gets adjusted automatically. There is no "learning' or "adapting".

On my jaguar I have two settings. Normal and Sport mode. In sport mode the suspension stiffens up, the throttle response increases and the shift patterns get adjusted. In essence, pushing that S-mode button, off sets the normal curves, maybe even alters the shape a little. But whilst in normal mode, the various curves do get adjusted. No neural network is needed to very simply determine how fast I typically accelerate. The ECU knows speeds, throttle position, and based on time can calculate acceleration. Over time that gets averaged out and if I'm more heavy on the old right foot than average, the shift curves get adjusted a little bit. Simple as that. Very straightforward math.

I'm not sure, but your Motec units are probably capable of storing multiple curves and shifting between based on external inputs.

Enjoy and make sure you don't get stuck in your personal experience. There's a whole big world out there you can learn from; Adapt is the name of the game!

Jeroen
Jeroen is offline   (5) Thanks
Reply

Most Viewed


Copyright ©2000 - 2025, Team-BHP.com
Proudly powered by E2E Networks