h a l f b a k e r y"Not baked goods, Professor; baked bads!" -- The Tick
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Please log in.
Before you can vote, you need to register.
Please log in or create an account.
|
PID Diet
Obesity: a simple control(s) problem | |
Being an electrical engineer, I was struck by the thought that one could look at weight loss as a simple controls problem with a single output variable (mass) and a single control variable, net calorie intake. (Actually, this is two control variables, gross calorie intake and calories burned, but they
can be reduced to a single variable for the purposes of this model.) So why not apply everyone's favorite control scheme to dieting? I speak, of course, of Proportional-Integral-Derivative control, or PID for short. [See link for an explanation, if interested.]
With the accompanying book, you would receive a sort of calculator that would implement the PID algorithm. At the beginning of your diet you would enter your starting weight, the amount you'd like to lose, and some reasonable timeframe in which you'd like to lose it. The calculator would tune the P, I, and D gains accordingly. Each day you would enter your new weight, and the calculator would compute what your net caloric intake for that day should be based on the "error" in your weight (i.e., the difference between your current weight and goal weight).
But rather than simply telling you what your net caloric intake should be, the calculator would offer you a range of eating and exercising choices to meet that number. In the initial setup, you would enter the minimum amount of food you were willing to eat in a day, and the maximum amount of exercise you were willing to do. Each day the calculator would plot a line on this exercise-food continuum, and you would select any point on the line, knowing that they all would result in the proper net calorie intake for that day. When you select a point on the line, it could give you a sample menu and workout routine: at the low end of the line it might suggest eating 3 rice cakes and watching TV; at the other end it might suggest eating a whole pizza and running a marathon. Hopefully you'd choose a point closer to the middle.
I think people would go wild for a diet that was advertised as "using the same technology that keeps millions of factory robots working smoothly and accurately each day!" And who wouldn't have confidence in a diet that was designed by engineers, as we are generally regarded as paragons of physical fitness?
PID Control
http://en.wikipedia.org/wiki/PID_control A Wikipedia article explaining the fundamentals of PID control. [lankybits, Mar 29 2007]
John Walker: The Hacker's Diet
http://www.fourmila...t/www/hackdiet.html "How to lose weight and hair through stress and poor nutrition." Re: engineers and dieting. [jutta, Mar 29 2007]
State-Space Control
http://en.wikipedia...te_space_(controls) Modern Control Theory! [TerranFury, Apr 03 2007]
[link]
|
|
Also, once the PID diet was discredited (possibly by me, once sales of the book and calculator started lagging), I could introduce a new controls-based diet: "Fuzzy Weight Loss". The pseudocode for the fuzzy diet would look like this: |
|
|
if(too fat){
if(getting fatter) eat much less
else if(weight not changing) eat somewhat less
else keep eating the same}
else if(weight is correct){
if(getting fatter) eat somewhat less
else if(weight not changing) keep eating the same
else eat somewhat more}
else{
if(getting fatter) keep eating the same
else if(weight not changing) eat somewhat more
else eat much more} |
|
|
Bun for the last paragraph. [+] |
|
|
Brilliant! I'll take one chicklet and a coma. |
|
|
Baked- by anyone who has ever successfully dealt with a weight problem. (+) |
|
|
If you haven't seen it yet, I recommend John Walker's write-up of his own (successful) dieting experience for a big dose of engineering 'tude applied to dieting. |
|
|
// I think people would go wild for a diet that was advertised as "using the same technology that keeps millions of factory robots working smoothly and accurately each day!" |
|
|
"Thanks for signing up with Cybernetic Diets, Inc. Here's your can of WD-40." |
|
|
you have to be careful not to set the proportional gain too high, otherwise you will have significant overshoot, and nobody wants to weigh 25 lbs |
|
|
//you have to be careful not to set the proportional gain too high, otherwise you will have significant overshoot, and nobody wants to weigh 25 lbs.lilsis, Mar 29 2007// I would have thought that integral wind up when not eating overnight would cause more overshoot!. |
|
|
[gnomethang], the new inputs are only computed daily, so the fact that you're not eating overnight wouldn't matter. |
|
|
But you guys make a good point, if you let your diet slip for a few days, the calculator would start giving you increasingly difficult (maybe impossible) calorie goals, if there were no bounds checking. The integral term should saturate at a reasonable level rather than growing out of control. Also, the lower bound for caloric intake will be the minimum food/maximum exercise point you chose at setup time. |
|
|
//Each day you would enter your new weight, and the calculator would compute// |
|
|
Just build it into the scales. |
|
|
I have to say though, that simplifying the entire complex interaction of caloric intake, nutritional value, composition, lifestyle, metabolism and biochemistry down to a single control variable system, where //eating 3 rice cakes and watching TV// is equivalent to //eating a whole pizza and running a marathon// is an abstraction so breathtakingly unrealistic as to be laughable. |
|
|
Pizza/marathon guy and rice cakes/TV guy _might_ eventually converge on the same weights (I highly doubt it) but they would be two very different animals by that time. |
|
|
ps Do I win the run-on sentence trophy? |
|
|
I like the idea of building the calculator into a scale, though you would want a handheld display module so you don't have to stoop down to read the graphs and menu suggestions at your feet. |
|
|
Assuming they both adhere to the calculator's recommendations, both rice cake/TV guy and pizza/marathon guy should meet their weight-loss goals. You're right, though, pizza/marathon guy will certainly be healthier and more muscular. The PID Diet makes no claims about the quality of body it produces, only quantity. |
|
|
Would it also work for weight gain? |
|
|
Bun for having nothing to do with Pelvic Inflammatory Disease. |
|
|
2 - Re: "Pizza/marathon guy and rice cakes/TV guy _might_ eventually converge on the same weights"[BunsenHoneydew]: One issue is that as you build muscle your resting calorie consumption goes up; this would affect system dynamics, as once you close the control loop there's now a (very slow) positive feedback. |
|
|
3 - It looks like 'The Hacker's Diet' (thanks, [jutta]) really does have prior art (But I maintain that he who re-invents the wheel has no less invented the wheel than did the first guy -- so, awesome). |
|
|
4 - More advanced control: It seems we should build a thorough state-space model. Weight then is just one input to the "observer" (state estimator), but you can enter more as well -- say, the results of blood sugar readings, or, occasionally, of percent-body-fat tests -- which provide a better estimate of the full state of your body (from which one could presumably do better control than by just looking at the one output "mass:" We could even steer towards certain body _types_). For the state estimator, one might use a Kalman or particle filter. |
|
|
Y'know, the more I think about it, the less I'm sure you need a 'D' at all. Do we humans have any kind of 'diet momentum?' I don't think we do. If anything, I thought we tended to damp things out ourselves: If we feel we're getting less food, our body kicks into "starving mode," etc. The oscillations people perceive, I would bet, arise mainly from the cycle, (1) Diet (2) Get results, be happy, and stop dieting (3) End up fat again. (4) Goto 1. |
|
| |