PDA

View Full Version : Center of Gravity



Bill Chennault
11-29-2006, 09:37 AM
All--

Is there a sensor, such as the Memsic 2125, that will allow one to determine the "center of gravity" by determining the tilt or inclination of the component from 0 degrees? (I am only interested in left/right.)

My application is an 8-legged walker with an articulated spine (a spine like yours and mine). I know I could ignore the center of gravity when one of a pair of legs is lifted, but that would result in a more rigid, ie., less manuverable structure. Since I am still in the concept stage, it is easy to conceptualize a walker that pulls a leg towards the spine when the opposite leg lifts, thereby positioning the center of gravity over the supporting leg. Doubtless, it will be far more difficult to implement the concept since it will require another actuator on each of the eight legs . . . and I am not even considering the processing speed necessary at this time.

Anyway, now that you might have a clue as to what I want to do, is the Memsic 2125 the proper device? I am thinking that I need something like 1 degree resolution to make this work (conceptually).

--Bill

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
You are what you write.

kelvin james
11-30-2006, 01:53 PM
The site link i supplied has some detailed info that is worth looking at, i am pretty sure this is what you are trying to do. This was from the gyro page at sparkfun.

kelvin

www.dprg.org/projects/2003-01a/ (http://www.dprg.org/projects/2003-01a/ )

Bill Chennault
11-30-2006, 08:55 PM
kelvin--

Thank you very much for that link! Now, the quest begins!

--Bill
ps I only had time to read it very quickly this morning, but I am amazed by the sensitivity of the various devices.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
You are what you write.

LSB
12-01-2006, 06:02 AM
Bill,
Curious newbie here. Isn't tilt information too late? Seems like pressure pads on the bottoms of 'feet' to compare weight distribution would avoid the need for tilt and resolve the issue before tip? What am I missing? ...Just in case I ever progress to the point where I can actually make something that works the way I intended!

Bill Chennault
12-01-2006, 06:53 AM
LSB--

You may well be right. It is my plan to correct the tilt as soon as it is detected . . . or as soon as feasible if the detection method is extremely sensitive. The device will have force sensors, as well. I just never thought of using them in the manner you suggest.

I would like to detect and correct tilts in the range of 1 to 5 degrees (or so) from the horizontal. The devices I have been reading about lately seem to be able to detect even smaller tilts. This sounds perfect to me. On the other hand, force sensors on the feet sound (to me) like good input devices for other reasons.

Mainly, I suppose, it is fun to think of these things and to see what others--like you--think of the same things. It certainly stimulates the imagination.

By the way, my project--which is an 8-legged walker with an articulated spine--probably needs neither the tilt sensors (mounted atop each hip)·or the force sensors, but with them a more natural gait should be attainable.

You missed nothing: You added something. Thank you.

--Bill

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
You are what you write.

LSB
12-01-2006, 08:14 PM
Thank you.
I've been playing with the Mesmic sensor recently in regard to measuring wind speed and direction. Also pressure sensors as I try to figure out a way to weigh rainfall accurately. My experiments aren't particularly applicable to your application, but similar enough that it piques my interest--and fosters hopes that some facet of your discovery will aid me. Just to put the matter to rest in my head--
In a tripod leg configuration, pressure on each foot would be proportional to COG distance from center, correct? Say, 60% of the weight was on one foot and the remaining 40% split between the other two. This would place the COG at directly between the two light legs (Y axis), roughly 2/3 of the way toward the heavy leg (X axis). This method would provide a roughly linear calculation of COG anywhere INSIDE the tripod of legs where no tip is possible. Outside the tripod the formula would 'go negative' because the light legs would weigh 0 (or less!) and tip would be imminent as only one leg bears weight, yes?
I see the value of the Mesmic here, because no further calculation of weight or pressure would be possible--one leg would bear 100% weight until the tripod landed on its side--at which point the weight supported by all legs would be 0. I also understand that any leg in the "up" position will bear no weight and therefore provide no relative pressure information.
Please continue posting your progress, I find walking robots facinating. The mechanical complexities are beyond me, but thinking about the problems is cheap and /nearly/ as rewarding.

Bill Chennault
12-01-2006, 11:40 PM
LSB--
I will keep posting my progress, although it will be very slow. The reason I "like" the Mesmic 2125 is because Parallax sells it and they provide much valuable support and advice for their products.·I really have no idea if it is as sensitive as I need, or not. I will have to buy it and experiment with it to find out, I guess. If you already know, please tell me. (I would like to get a signal from it with AT LEAST every one degree of tilt). My plan is to use that feedback to eliminate the tilt with a stepper by pulling the center of mass back to the center of gravity, thus achieving balance (basically by returning the Mesmic 2125 to 0 degrees). Since it is--will be--an 8-leggged walker, I do not care about fore/aft tilt.
Since you seem to have some interest in what I am thinking: Here is another piece of trivia. I intend to network 8 BS2p24s and dedicate each one of them to a leg. A ninth Stamp of some kind will exist in the network and its sole function will be to build communication packets--based on data from the "leg-Stamps"--for a tablet PC, which will also be in the RS-232 network. The leg-Stamps will receive communication packets from the tablet PC and be controlled, accordingly. In turn, the tablet PC will communicate wirelessly with a network of desktop machines. These desktops will be dedicated to various functions, including the construction of successful movements (consisting of joint positions combined with sensor readings) over time. (In other words, this thing will learn by doing via the ever-evolving, off-board joint position-sensor reading table: It SHOULD get smarter.)
There will be some degree of intelligence located within the on-board tablet PC . . . plenty to keep the device out of trouble. MOST of the intelligence will reside in the off-board network with each host PC contributing its commands to an ever-evolving "joint position table" located on the server which has wireless communication with the on-board tablet PC. During normal operation, the server will transmit only the changes to the tablet PC, sort of like an incremental backup procedure. Upon startup, once the server decides what "state" the robot is in, it will transmit the desired joint position table to the tablet PC to place the machine in the desired state.
If implemented, the project will be neither small nor inexpensive. I am about to buy a mill to begin making articulated cervical vertebra and hip joints. I'll make prototypes of the hip joints first and afix some kind of hoaked up but powered·femur/knee/tibula-fibula/ankle/foot to a pair of them (a set of hips) and experiment with motors: steppers, servo, PWM DC, and just plain low geared DC. (Does it sound like a cat? It should, since I am trying to create a robotic "baanth" which was a creation of Edgar Rice Burroughs some time around 1917. He was a genius and created a VERY stable platform on paper for me to try to put in metal and silicon.
I will spend $50k to get something that will move freely around the house and put its head in my wife's lap.· It may have the added benefit of scaring her cats to death. One can only hope.
I will not apologize for the length of the post, because--as you may have heard--it iced and snowed a bit here in The Land Of Oz (Kansas) and we cancelled classes at the college and I love to write! (Dorothy says hi.)
--Bill

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
You are what you write.

LSB
12-02-2006, 09:53 AM
There’s a tutorial in the “Stamps in Class” forum on this part (look for it in the ‘Sticky’ section titled mini-projects)--it's how I got started. The tutorial describes the range of readings you can expect (it works out to near 8 tics per degree. I'd agree, but I don't have any specific data as I reduce the raw readings to a byte value for wind speed and 16 compass points for direction) it also explains how to increase the accuracy as the part is somewhat temperature sensitive in extremes.

As to the rest, remember the old addage: "inch by inch it's a cinch, mile by mile it's a trial"

...and keep posting your progress!

Mark

Bill Chennault
12-02-2006, 10:40 PM
Mark--

Thanks for the reference to Mesmic 2125. I will certainly study it. And, the advice is great, too! (There is no doubt that this will be a very slow project, since I have to learn as I go. But, I have approached many things with success in that manner . . . hopefully, this will prove to be one of them.)

I document everything I build, from firearms to tile-work, so when I start, everyone here will be tortured with pictures. Fortunately, it will be after Christmas (so Merry Christmas, everyone!), because the mill not be here until sometime in January.

--Bill

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
You are what you write.