Has anyone considered a quadraped robot designed after a CAT? This robot would have the legs to the sides of the body, rather then extending out from it. 3DOF legs might be enough, although it would be interesting to add a foot joint as well. The front limbs would be different then the rear limbs. Might make it a little tall compared to what is being done now!
Look under a-s.Clayton.edu on this page (2nd row, last column on my screen):
I have been thinking about a centaur project that would involve leg movements similar to a cat. The hip servo would be mounted vertically to allow flexion, extension and hyperextension in a sagittal plane. The knee servo would be mounted to permit flexion and extension and the ankle servo would permit plantar flexion and dorsiflexion. I’m still in the SES part collection phase for this project.
Ummmmmmmmmm, YES, actually I have! I have been observing how my kitty (Doozi) moves. Cats have a reverse knew, which would be pretty easy to replicate with SES parts.
Those are quite complex joint movements for SES! I’d like to see what you come up with! How many DOF do you expect per limb do you expect? I see 4DOF immediately, which might be OK, but then there needs to be some yaw to the travel of the limbs as well (probably not the right way to describe it).
Cats are really quite limber! I’m wondering if there is another way to supply limited HipH movement. Maybe it can be done with a simple hinge? Otherwise, it’s another DOF, and 4 additional servos! I’d like to keep it down to 4DOF (no ankle?), although that may not be reasonable. Can a Heading change be implemented by simply shorting the stride on one side, while lengthening the stride on the other? I seem to recall something like that. So maybe 4DOF would do it!
I’m NOT looking for omni directional (crab) walking, Just a good quadraped! How hard to make the 'Bot JUMP? (I saw a frog skeleton too). ;>)
I’m in touch with the Dr. that posted the CAT skeleton pix (URL in a previous post), and he expects to email me some bone dimensions. That might give us some clues as to scaling of the limb components and body length/width (unless someone can get their cat to stand still that long). ;>)
Perhaps you could describe these joints more. I have no medical (Biology) training!
Certainly seems appropriate, and would be more “Biometric”, although I don’t know how close we could come to a simple simulation of the Cat’s legs. One thing for sure, it’s gonna make the IK’s a little rougher!
I think I was wrong about cats having a reverse knee. The femur and second bones are just about equal length, and the foot is third bone is quite long that attaches to the paw. This is true of the back legs. I will have to watch Doozi closer yet and see how her front legs work.
I would use The BiPod leg design as a starting point and add more length between each joint. There will be a minimum of 4DOF for sure. I already have a body design in mind that I believe will give the required range of motion for all the joints. I won’t add any additional DOF unless absolutely necessary.
It never ceases to amaze me the places Doozi can get in and out of! Cats are also very adaptive. I rearranged my living room the other day, and Doozi has already adapted to the new setup and figured out how to get into the window and up on the top of the computer hutch. She is very a smart kitty!
The leg design I have in mind for CatBot will be very flexible. You will have to do the testing of it though because I can’t afford to spend money on servos right now.
If the joints were driven by regular motors, jumping would definitely be doable since the motors can accelerate much faster than a servo and have much more speed. I say this when I cranked WALTER up to full speed and set him loose on the floor - he ran right into a door because he didn’t have enough time to react to sensor detections. He now runs at about 1/3 speed as I gradually tweak his controls.
Wonderful! Then we can get some proper proportions and such.
If I understand the basics of IK, there are constants that you set for leg dimensions, number of legs, and naybe proportions. Those are used in calculations of joint positions for the move.
FRONT and HIND! The front limb could be said to have a reverse-knee. Are you going to (attempt to) measure Doozi?
Well, yes and no. It’s not a parameter driven “engine” that will automatically calculate all of the angles for you; at least not that I’ve found yet. There is a Human Model that works with IKAN, what you seek MIGHT be hidden in there. IKAN is a library that supplies the trig, matrix and some other needed calcs to allow one to do the transforms, etc. needed for an elaborate arm or leg model. No, I don’t know how to use it yet. And as it is FP intensive, it might be a little too much for the uP I’d like to run on first (PIC18 family). DsPIC? ARM7??
I don’t think I will try that. She has CLAWS, even though she rarely puts them out.
Lynxmotion has Bipeds, I have The BiPod. Yes, I have all the brackets put together for The BiPod, but most of those are probably going to go in a trade soon. I just can’t afford the servos (10) to fill the brackets of 4DOF legs and a pan/tilt turret.
Well, not exactly. I have some ideas running around in my head though. I just haven’t loaded up Alibre Design to test them out yet.
I wonder if two ASB-04’s could be put together in such a way as to create a differential joint. It should be workable, but might require some additional holes (and there isn’t much room for those).
There is also the new PIC32 chips, based on a MIPS core. I already upgraded to MPLAB 8.0 and got my student licence for C32. I think IK would require a dsPIC at minium, perhaps a dsPIC33F part since they are a bit faster than the 30F series. I did just get a new LPC2148 Proto Board, but my JTAG programmer/debugger may be defective. I can flash the new board using the LPC2000 Flash Utility (2.2.3) just fine.
I have no information about it, but the main webpage of the Dallas Personal Robotics Group (DPRG) has a random-rotating image in the corner, and one of the pictures is of a quadrupedal chassis with an “inline” joint configuration:
This image has been in the rotation for quite some time, so I imagine that either progress has been made on it since then, or it’s been abandoned.
It’s not much to go on, but it might get you somewhere.
Perhaps someone within their club could provide you with more information on the robot in question…?
Okay, a crash course in kinesiological analysis of limb movement. Take a piece of paper and a pencil. Insert the pencil into the middle of the paper. The paper represent the plane of motion and the pencil represents the axis of rotation.
If the paper is oriented so that it has a top and bottom side, it represents the horizontal plane. The hip in the Lynxmotion 2DOF quadraped leg moves in this plane. Turn the page so that it has a front and a back, this is the frontal plane. Movements in this plane are usually called abduction (moving away from the mid-line of the body) and adduction (moving towards the mid-line; mnemonic is “adding to the body”).
If you turn the piece of paper so that it has a left and right side, then it represents the sagittal plane. The axis of rotation is oriented frontal/horizontal. Typical movements in the sagittal plane are flexion, extension and hyperextension. From observations of my cat, it appears that the primary movements of a cat’s limb joints during normal locomotion appear to be flexion, extension, and hyperextension. As was noted, this is different than the Lynxmotion quadrapeds and is more similar to the bipeds.
Unfortunately, many movements that our joints make are more complicated than can be described by a single plane or axis of rotation. The shoulder and hip, being ball and socket joints, are prime examples of this. However, since we are using servos which have a single axis of rotation, using a piece of paper and a pencil can be a helpful way to visualize how you will need to orient your servos to get the desired motions.
Once you can describe the sequence of joint movements of single limb during locomotion, then you can move on to describing the pattern of footfalls needs for a various gaits. In turn, this may lead to a need to shorten and lengthen various components of the joint movements just as we change the range of motion of our joints to achieve different gaits.
Unfortunately, the random-rotating image is just that - an image, with no description or link to anything more informative. The image I posted earlier is r6.jpg. If you keep hitting in your browser (reload page), you should eventually get the exact same image that I posted earlier (you may have to hit it a bunch of times - it’s random, and there are 120 images). Or if you really like it, you could just save the one that I posted, since it’s exactly the same, and referencing the DPRG webpage anyway.
If you’re clever, you can just go to the directory that the images are pulled from and look at the images sequentially, but you still won’t find any more pictures of the quadruped. (A number of the bots’ chassis should look familiar, though)
Thanks for the coursework! I and I’m sure others appreciated it. I’m familiar with planes and axis in CNC, and also for aircraft and submarines, but in Biology I’m sadly lacking! I had a sneak peek at Wiki to discover some of the terms shortly after you posted.
Flexion, extension, and hyperextension I’ll have to look up, although I can just about guess as to their meaning. Are you in a medical field? You might be well positioned then for studies into Neurons and CPGs then.
When your cat turns, does he/she “lean” into the turn, and extend a front paw somewhat laterally in the desired direction?
I agree, off hand, I can’t think of an easy way to implement a driven ball joint with R/C servos. The hip for a cat might be implemented by a servo positioned fore-aft (anterior-posterior), which could be used to allow some limited sideways motion (frontal plane?). Attach a 2nd servo positioned to allow movement in the sagittal plane. From then on, a Biped leg should suffice. I think I’ve got the legs worked out! It will, I think, necessitate some additional C-brackets of different lengths. I’d also like to fabricate a somewhat curved (like the spine) body.
Only efforts for now might be to build up one leg and photograph it before I commit all my SES brackets to my CH3-R 'Bot (unless I come into a windfall of parts).
I got a set of cat skeleton dimensions from a doctor in GA. I’m lacking two dimensions (width across hips and shoulders), but these should give some ideas as to proportions.
I’ve also found a paper on CPGs that has some useful information on cat Gaits.
I saw the photo on the Texas Robotics club, but didn’t find any additional information, or bigger pix. Too bad, it could have been useful!
I’m in search the paper “A Computerized look at Cat Locomotion, or One Way to Scan a Cat” by Glen Speckert (MIT '76). I’ve emailed him, no reply yet. The paper is on the internet, but it’s a scan of a photocopy or something, no detail in the photos.
Alan - my original career was exercise physiology and human performance training. I taught kinesiology for many years.
I guess the key here is whether you are trying for a robot that moves exactly like a cat or a robot that moves more like a cat than the current Lynxmotion quadrapeds. I think the first might be near impossible. One of the things that playing with robots has done for me and that is deepen my appreciation for the wonderful abilities of our bodies.
Some more definitions:
Flexion of the knee joint - starting in a straight leg position and moving your foot towards your buttocks
Extension of the knee joint - returning to a normal straight knee position (0 or sometimes referred to as 180 degrees).
Hyperextension of the knee joint - moving past the normal knee extension (past 0 or past 180 degrees) - more common in women
Flexion of the hip joint - starting in a straight leg position and moving the thigh towards your stomach
Extension of the hip joint - returning to a normal straight leg position
Hyperextension of the hip joint - moving the thigh past the normal straight leg position toward the buttocks
Plantar flexion of the ankle joint (talotibial) - pointing your toes down, away from your head
Dorsiflexion of the ankle joint - pulling your toes up, towards your head
Well, that explains the medical terminology! Quite apropos of course for a cat’s anatomy!
A limited “interpretation” of a cat would satisfy me. The arched back can be imitated, the differing leg segment lengths scaled as much as possible, and leg flexion in the proper plane would go a long way towards giving a realistic imitation of a cat in motion.
Initial efforts can ignore the hip joint with it’s side to side action, and also any true ankle motion. This leaves at least 3DOF, in a leg similar (I’m told) to that of a biped. That’s probably the easy part. Actually generating the appropriate gaits with realistic joint action will be the real task!
Thanks for the additional definitions! Now I think I’ll look around for some diagrams showing the motions.
Yes, I have all the brackets put together for The BiPod, but most of those are probably going to go in a trade soon. I just can’t afford the servos (10) to fill the brackets of 4DOF legs and a pan/tilt turret.
