MPG Cruise Control Algorithm

[theholycow]

Originally Posted by maximilian

hills give you the opportunity to pulse longer up them. Of course the downhill would interrupt a pulse if it occurred then, but you get more coast out of it too. Do I have this right?

The descent is not a repenting enemy, it is an ally. The pulse is a price you pay for the glide. By climbing a hill, you store energy that was produced more efficiently during the pulse. Interrupting a pulse is fine, it's interrupting a glide that stinks. But, that minor bit aside, you are correct.

[GasSavers_maximilian]

Wasting energy gliding at very high speeds seems counterproductive because of air resistance. Bummer.

[theholycow]

If you're descending and gaining speed, you have nothing to lose by the free acceleration. If you gain too much speed, you didn't plan well enough and continued your pulse for too long. When you know you'll have a long steep descent, you quit your pulse before cresting the hill.

[GasSavers_maximilian]

Originally Posted by theholycow

If you gain too much speed, you didn't plan well enough and continued your pulse for too long.

Right. That means a pure OBDII pulse and glide is not going to work acceptably. Sorry; I wasn't clear. A GPS enabled one would be interesting, but that's way more complicated.

So how come I've seen suggestions to lose speed when climbing hills? I've been crawling up them. Wait a sec, since this is getting off topic, I'll just do a search. Thanks for the info.

[R.I.D.E.]

In my scenario I have found that when climbing hills after descending a previous hill, I can keep my mileage at the highest (based on instant bar graph) by loosing speed while climbing. If I try to maintain the same speed my instantaneous mileage drops dramatically.

This is probably specific to my car and the very slight grades I face in my daily drive, as well as the fact that in the areas where I encounter the steeper grades are on a 4 lane road that runs parallel to the Interstate with little traffic to cause problems with other drivers.

In my situation my speed varies by about 20 MPH (50-60-40) in negotiating those grades. Not really what I would call cruise control in the classic sense of a constant speed. Also not applicable in longer sustained grades, which I do not encounter with elevation changes of 100 foot extremes in a 18.5 mile route.

I have also found that even on flat surfaces I can do mini pulse and glides. Very gradual acceleration, in the range of several seconds per MPH increase, allow me to achieve instantaneous readings of over 100 MPH for a similar distance as used for acceleration. This improves my overall mileage compared to constant speeds. In may daily trip that is relatively rare, when I get close to the areas where the elevation changes are measured by a few feet per mile.

regards
gary

[GasSavers_maximilian]

Once I get my new mpg gauge I will have to experiment some more. I'd read about losing speed when going up hills so was already biased in that direction. I'll try to be a bit more adventurous. I discovered a fairly long level stretch on the other side of the mountain I live next to. I had forgotten it was there and was unable to find a spot to test the best level travel speed for my vehicle. Not much reason to go over there, but one trip won't hurt.

[GasSavers_maximilian]

Thinking about it on my drive this weekend, and hill climbing speed is pretty straightforward. The best level cruise speed is an compromise between running your engine at partial load and the losses that increase with speed. When you are climbing a hill, the extra power required changes the equilibrium point, making the best speed lower. When going downhill the best speed would be higher (assuming it's not steep enough to just coast). All pretty easy to figure, so I'll rewrite my original algorithm to have slightly more refined hill handling.

[GasSavers_maximilian]

Redid the algorithm. A lot of the complexity is hidden in the "compute speed for maximum efficiency" statement. A look up table could get the system close and then a control system examining the instantaneous mpg could refine it. I considered a maximum speed case, but since braking is under the control of the driver and also that a steep grade would lead to coasting anyway, I don't think it will be needed. This differs slightly from what I do now in that I do a limited form of pulse and glide when already coasting anyway: I just let my speed go down really far. That behaviour seems less appropriate in a cruise control so I left it out. Once I have my new car computer and get some other testing done I will drive this algorithm to gauge its efficiency. Now I can think about how a GPS equipped algorithm could work. That gives a lot more possibilities!

Repeat {

if( grade <= MIN_GRADE_FOR_COASTING) {

// on steep downhill
coast;

} else if( speed <= MIN_SPEED ) {

// very steep uphill, economy is not the concern
adjust throttle as needed to maintain MIN_SPEED;

} else {

// slight uphill, level, or slight downhill
compute speed for maximum efficiency;
// for a level, this will be a cruise speed of around 50-55
// for slight uphill, this will be slower than the cruise speed
// for slight downhill, this will be faster than the cruise speed

}

}

[GasSavers_maximilian]

Just for kicks I wrote up the algorithm I currently follow when driving. It's very similar to the last one. Just stays in coast mode once it enters it until the speed reaches the minimum allowable. Obviously I adjust it to account for stops and sharp curves. The GPS algorithm is for that info.

Repeat {

if( grade <= MIN_GRADE_FOR_COASTING) {

// on steep downhill
while( speed >= MIN_SPEED ) {

coast;

}

} else if( speed <= MIN_SPEED ) {

// very steep uphill, economy is not the concern
adjust throttle as needed to maintain MIN_SPEED;

} else {

// slight uphill, level, or slight downhill
compute speed for maximum efficiency;
// for a level, this will be a cruise speed of around 50-55
// for slight uphill, this will be slower than the cruise speed
// for slight downhill, this will be faster than the cruise speed

}

}

[dkjones96]

Computing vs table depends on how much memory you have and what resolution you want. If you want it to the tenths or even to +-2 mph you will want it to calculate, but that takes processor time and can make the response sluggish depending on how involved the calculation is. If you want it to, say 5mph, but only in a window of 45-75 mph you can do that without needing too much memory.

I only mention memory because it is easy to fill 4Kb of flash memory on an MC. The mock program I put on the first page is already 3.29Kb. A simple 8-bit MC with 8K is probably what you would want to use but you'll still need to take some things into account (processor speed).