Polerskating
Polerskating
by beermat
Mon Aug 21, 2006 3:11 am
So, finally getting back to my rollerskating project. Having some success generating forward movement, but it's a little inelegant.
Anyway, in a break from coding, I decided to cheat a bit. Here's the video. Apologies for the quality, but I present the first EVER polerskating robot:
Robonova-1 polerskating
Anyway, in a break from coding, I decided to cheat a bit. Here's the video. Apologies for the quality, but I present the first EVER polerskating robot:
Robonova-1 polerskating
by hivemind
Mon Aug 21, 2006 4:07 am
Ha, great job. What are you using for poles there? they seem to be very thin with those nice grips on them, almost the size of chopsticks. What was your fortune?
So i was thinking of a way to solve the skating problem a while back. And something occurred to me as a way of cheating the additional DOF with some momentum. You have proven that given a push the RN can glide (and the pole push seems to work nicely here too, so not much force is required). Perhaps, you could lean the RN back, extend a leg, and have him snap back(highspeed) onto the reverse skate(like a quick forward step), so instead of a humanlike skate, its more of a chicken walk on ice. Essentially you could count on friction to hold you in place while you lean back, and then you set highspeed on to go forward, shift the weight, and by the time friction is overcome you have nothing but forward motion as the RN shifts to the alternate foot.
Perhaps a beggining way to try this is to start with the poles and basically walk using the poles until you find the right gait and it can generate enough force. Kind of like training wheels. But Poles.
Of course I could certainly be wrong(and am the first to admit that), but it may be something to try while you are waiting for additional independant DOF. But with those new grippers on there your center of gravity is raised quite a bit.
I'll try running some sims in the next couple of days as I have been getting my act together to get my RN model into a physics engine. But the reality is that you seem to be doing great here.
Nice work, and btw, have you tried using one of those objects as a turning mechanism? A couple of weeks ago i had a plastic sword and my front/back and side to side gyros on and was holding a sword (thanks to matts grippers). The arm was out perfectly straight forward (like a human elbow locked). I rotated only the shoulder servo back so the tip of the sword went 180 degrees to facing down, and it bowed a little on my hard mat. When i brought it back it was stuck, and so instead of snapping the sword or hurting the servo i was able to get a perfect 90 degree turn. Consistantly. Your poles reminded me very much of this.
I look forward to your progress, keep up the good work.
So i was thinking of a way to solve the skating problem a while back. And something occurred to me as a way of cheating the additional DOF with some momentum. You have proven that given a push the RN can glide (and the pole push seems to work nicely here too, so not much force is required). Perhaps, you could lean the RN back, extend a leg, and have him snap back(highspeed) onto the reverse skate(like a quick forward step), so instead of a humanlike skate, its more of a chicken walk on ice. Essentially you could count on friction to hold you in place while you lean back, and then you set highspeed on to go forward, shift the weight, and by the time friction is overcome you have nothing but forward motion as the RN shifts to the alternate foot.
Perhaps a beggining way to try this is to start with the poles and basically walk using the poles until you find the right gait and it can generate enough force. Kind of like training wheels. But Poles.
Of course I could certainly be wrong(and am the first to admit that), but it may be something to try while you are waiting for additional independant DOF. But with those new grippers on there your center of gravity is raised quite a bit.
I'll try running some sims in the next couple of days as I have been getting my act together to get my RN model into a physics engine. But the reality is that you seem to be doing great here.
Nice work, and btw, have you tried using one of those objects as a turning mechanism? A couple of weeks ago i had a plastic sword and my front/back and side to side gyros on and was holding a sword (thanks to matts grippers). The arm was out perfectly straight forward (like a human elbow locked). I rotated only the shoulder servo back so the tip of the sword went 180 degrees to facing down, and it bowed a little on my hard mat. When i brought it back it was stuck, and so instead of snapping the sword or hurting the servo i was able to get a perfect 90 degree turn. Consistantly. Your poles reminded me very much of this.
I look forward to your progress, keep up the good work.
by beermat
Mon Aug 21, 2006 12:13 pm
Hive - i feel like you've been spying on me! *Searches for hidden webcam* ...... yup, I can use the poles for turning, I have routines for left_pole_turn and right_pole_turn, which are basically the same as the forward_pole_skate routine, but only using one arm / pole, and the little bugger executes a turn of around 70 degrees. Can be refined a little, too....
As for your thoughts of cheating the extra DOF for rollerskating with a 'chicken walk on ice' motion, that's where I'm heading - my first line of my post said I'm having some success, but it is inelegant, referring to just that! After attaching the new skates, I basically went through the stock motions to see what effect they have - left/right_shift, left/right_turn, forward_walk and fast_walk. The last one generated some forward momentum on the skates, so I'm experimenting with that one, adjusting the moves and inserting short delays to see where the forward momentum comes in; basically some trial and error to see what works well.
It's made a lot harder by the fact that the floors in my 100 year old house are pretty wavy, so some sections are flat, some are uphill, some are downhill. I might have to create a contour map of my floors! Anyway, the uneveness makes the timing hard - sometimes the 'next' move comes too soon and stops the momentum, sometimes too late and no cumulative moment is generated.....
The poles are 3mm carbon fibre rods which work great because they are pretty stiff but still slightly flexible, which makes the difference between them jamming into the ground and toppling him - or not. I wrapped insulating foam and electrical tape in two places at the top of each pole to allow the grippers to have something bigger to grip onto.
The forward momentum by the way is initially generated by rocking forward on his knees and using that to push on the poles. The arm motions add to that.
Ok, I have to go to work now! I'll see if I can make more videos later.
As for your thoughts of cheating the extra DOF for rollerskating with a 'chicken walk on ice' motion, that's where I'm heading - my first line of my post said I'm having some success, but it is inelegant, referring to just that! After attaching the new skates, I basically went through the stock motions to see what effect they have - left/right_shift, left/right_turn, forward_walk and fast_walk. The last one generated some forward momentum on the skates, so I'm experimenting with that one, adjusting the moves and inserting short delays to see where the forward momentum comes in; basically some trial and error to see what works well.
It's made a lot harder by the fact that the floors in my 100 year old house are pretty wavy, so some sections are flat, some are uphill, some are downhill. I might have to create a contour map of my floors! Anyway, the uneveness makes the timing hard - sometimes the 'next' move comes too soon and stops the momentum, sometimes too late and no cumulative moment is generated.....
The poles are 3mm carbon fibre rods which work great because they are pretty stiff but still slightly flexible, which makes the difference between them jamming into the ground and toppling him - or not. I wrapped insulating foam and electrical tape in two places at the top of each pole to allow the grippers to have something bigger to grip onto.
The forward momentum by the way is initially generated by rocking forward on his knees and using that to push on the poles. The arm motions add to that.
Ok, I have to go to work now! I'll see if I can make more videos later.
by hivemind
Mon Aug 21, 2006 1:21 pm
Awesome, carbonfiber poles sound like a great idea.
Perhaps a way to time it better is to do the fast step(as you had already stated-although perhaps modified), but each time you are about to step again you check an accelerometer(3d preferably, albiet you only need 2d, and in this case 1 axis...). So you put in a loop that waits for it to be above a certain value (aka: leaning/tipping forward) before you kick into the next step. Or if you dont have an accelerometer you could try gyros and see how that works. Just a thought. I have had success in some expirements using both, but its usually a lot of hassle to get something that works only decent. A lot of the time the ports seem to be slow reading especially with motors on (which is rather important...) but the gyros feed directly and thus are plenty fast, so it may be as easy as just turning them on with the correct gain...
Have you attempted having the RN stepping onto the skates and attaching them himself? They look to be just about the height of the robostairs that i believe ive seen Pev using. And with matts grippers, well ok, maybe its a bit much to tie stuff. velcro? hehe
Very cool how you also have the turn with poles. I can't help thinking that there must be a better way to turn though, without an extra object or servos. Hmm. *ponders fate of the universe*
Well, I look forward to the new videos, great.
Perhaps a way to time it better is to do the fast step(as you had already stated-although perhaps modified), but each time you are about to step again you check an accelerometer(3d preferably, albiet you only need 2d, and in this case 1 axis...). So you put in a loop that waits for it to be above a certain value (aka: leaning/tipping forward) before you kick into the next step. Or if you dont have an accelerometer you could try gyros and see how that works. Just a thought. I have had success in some expirements using both, but its usually a lot of hassle to get something that works only decent. A lot of the time the ports seem to be slow reading especially with motors on (which is rather important...) but the gyros feed directly and thus are plenty fast, so it may be as easy as just turning them on with the correct gain...
Have you attempted having the RN stepping onto the skates and attaching them himself? They look to be just about the height of the robostairs that i believe ive seen Pev using. And with matts grippers, well ok, maybe its a bit much to tie stuff. velcro? hehe
Very cool how you also have the turn with poles
. I can't help thinking that there must be a better way to turn though, without an extra object or servos. Hmm. *ponders fate of the universe*
Well, I look forward to the new videos, great.
by beermat
Tue Aug 22, 2006 5:18 pm
Thanks Matt! Couldn't have done it without those grippers!
Bullit - based on a tip-off a while ago from (I think) Gman, I looked into the Tech Deck Handboards. Aparently, there is a whole world out there of people who 'skateboard' with their hands on mini-skateboards! Odd...anyway, the wheels and wheel assembly came from a couple of 27cm handboards I picked up for just $6.99 each here:
Tech Deck
...which is a bargain, IMHO. So the 'professional looking' skates are really made by someone else I simply unscrewed the wheel assemblies from the handboards and, after drilling the appropriate holes in Robonovas feet, re-attached them. A simple mod. You'll need a handboard per foot (duh!).
The handboard themselves can accomate one RN foot pretty easily, but thats about it - they're not wide enough for both feet side-by-side and, even if RN had the DOF required to put one leg behind the other, the front and back slope up and the length could not accomodate both feet easily. That being said, I think I can still get a RN to balance with one foot on the board and the other propelling him along.....
Bullit - based on a tip-off a while ago from (I think) Gman, I looked into the Tech Deck Handboards. Aparently, there is a whole world out there of people who 'skateboard' with their hands on mini-skateboards! Odd...anyway, the wheels and wheel assembly came from a couple of 27cm handboards I picked up for just $6.99 each here:
Tech Deck
...which is a bargain, IMHO. So the 'professional looking' skates are really made by someone else
I simply unscrewed the wheel assemblies from the handboards and, after drilling the appropriate holes in Robonovas feet, re-attached them. A simple mod. You'll need a handboard per foot (duh!).
The handboard themselves can accomate one RN foot pretty easily, but thats about it - they're not wide enough for both feet side-by-side and, even if RN had the DOF required to put one leg behind the other, the front and back slope up and the length could not accomodate both feet easily. That being said, I think I can still get a RN to balance with one foot on the board and the other propelling him along.....