Subject:
|
Re: Simultaneous Control of Multiple Servos
|
Newsgroups:
|
lugnet.robotics.handyboard
|
Date:
|
Fri, 17 Jul 1998 08:50:48 GMT
|
Original-From:
|
MICHAELJ@WT.NETstopspam
|
Viewed:
|
1164 times
|
| |
|
|
I would suggest that just getting a walking robot to walk with coodination and
then being able to turn in place and in an arc will be challenging enough to
begin with. Just adding enough intelligence to adapt to rough terrain and the
change gaits based on sensory perception ans speed will no doubt keep you busy
for quite awhile. :)
I am in process of researching and designing a hexapod myself. I want it to
be able to move at about 2 mph (human walking speed)in its fast pace. I have
not found any hobby servos that can move in a 90 degree arc at 6 hz (6 inches
times 6 hz is about 2 mph). Well, let me correct myself...I haven't found any
I could afford! Keep in touch and we'll compare notes as we go.
I am actually considering using an Atmel 90S1200 16Mhz microcontroller for each
leg (3 motors per leg - only 2 for the first prototype) tied together with a
central microcontroller via an SPI bus.
Mike
> That's the way I thought it had to be done.
>
> What I was considering was routines that moved each leg the same short
> distance (for straight line walking). Walking would consist of very short
> increments of movement. Now, this seems like it would move the robot in a
> jerkey motion. Depends on how short each movement was.
>
> So, the way it might work would be to have variables for each leg
> position (or pulse width for each leg position). Starting position. To
> that you would add a delta (same for each leg). So, that each time all
> servos moved they would all move the same amount but since each might have
> a different starting position, all legs would move in sync and the same
> distance.
>
> The routine would do something like:
> 1) Set start position (initial pulse; all might be different)
> 2) Update each leg position register with delta
> 3) Wait 30-60ms (or whatever the update rate should be)
> 4) Update the new leg position (all would move together and the same
> length of travel.
>
> You would need one interrupt routine controlling the PRF. I'm not sure
> the best way to control the pulse width for each of the possible 12 legs.
> Maybe a tight routine that simply started each pulse and output the value
> of each of the leg position registers. Then it would decrement each every
> x microseconds, When the count reached zero the pulse would end.
>
> thanks for your help.
>
> Side comment: When I first started thinking about this, I was trying to
> think of something interesting and challenging. I thought it might be
> interseting to see if one could develop a four-legged robot that could
> run (move fast) and detect and jump over an obstacle.
>
>
>
> On Fri, 17
> Jul 1998 michaelj@wt.net wrote:
>
> > You are correct to assume that all legs should be in sync with each other
> > and that all the legs on the ground should be moving at the same rate in • the
> > same direction to move straight (sometimes you don't want to move
> > straight!). BUT now think of the speed of the processor and how many
> > different pins can be turned on and off in a VERY short amount time. The
> > servo pulse for each motor is about 50 hz but much of that time is spent
> > waiting for the time to end the pulse and then waiting for the delay to end.
> > Interlacing the signal control allows you to compress the time spent handing
> > the signal generation because the processor can start and stop the pulse • on
> > several different motors in a much shorter time frame. This would make it
> > appear as if the motor pulses were being done in parallel when in fact the
> > signals were handled sequentially. See my crude example:
> >
> > S = start pulse
> > T = terminate pulse
> > - = active pulse
> > . = inter-pulse delay
> >
> > S--------T..................................................................
> > ..................
> > .S----T.....................................................................
> > ...................
> > ..S----------T..............................................................
> > ..................
> > ...S-----T..................................................................
> > ..................
> > ....S--------T..............................................................
> > ..................
> > .....S--T...................................................................
> > ....................
> >
> > Note that the time delay between the start of one pulse and the start of • the
> > next is only the amount of time the processor needs to setup for it which
> > may be only a couple of instructions. By ordering which motors are adjacent
> > to each other in the sequence, you can minimize the effect of being "out • of
> > sync" (all swing motors on one side are adjacent followed by swing motors • on
> > the other side followed by lift motors on the first side, etc).
> >
> > Computers multitask the same way but because the switching happens at such • a
> > fast rate, it appears that everything is happening simultaneously. If you
> > had a hundred legs you would probably need a faster CPU but you could still
> > handle them in basicly the same way. Using a well designed interrupt-based
> > timer you can even reduce the time spent is handling the leg motor motions
> > and perform some sensing. If you reduce the flexibility of your walking
> > gaits and only use the alternating tripod gait then two legs on each side
> > can receive exactly the same signal since they always move exactly together.
> > That would reduce the number of individual motor sequences you would have • to
> > manage.
> >
> > I hope I didn't ramble on too long and that this helps answer your
> > questions.
> >
> > Mike
> >
> >
> >
> > -----Original Message-----
> > From: Mike Davis <ipscone@halcyon.com>
> > To: handyboard@media.mit.edu <handyboard@media.mit.edu>
> > Date: Thursday, July 16, 1998 10:46 PM
> > Subject: Simultaneous Control of Multiple Servos
> >
> >
> > I have asked this before and not received much regarding this question. • So,
> > here goes again.
> >
> > In building a walking robot, how important is it to control the motion of
> > all legs
> > simultaneously?
> >
> > Seems like that pulsing one at a time creates the situation where one leg
> > drags
> > the others. I know there is a delay between the time an mcu command is
> > issued and the time the pulse goes out, but still it seems like one leg
> > would
> > drag the others without some kind of parallel controller. I also wonder
> > about the
> > ability of the serial controllers to perform simultaneous servo control.
> >
> > Seems like each leg should be programmed to walk identically, and that each
> > would have a phase relationship with each other. But all have to be moving
> > in
> > sync and not one at a time. Is the HC11 a poor choice for this kind of
> > motion?
> > Or is it just fine? Or does it require some kind of servo controller?
> > Mostly I'm
> > interested in the walking algorithm and synchronization control.
> >
> > Is this at all an important issue? How is it addressed? Any web info on
> > this? I
> > have looked at many walking robot pages but have not seen this addressed.
> >
> > Thanks
> > --
> > -----------------------------------------------------------------
> > ipscone@halcyon.com Shoot-To-Win!
> >
> > Protect the 2nd Amendment! Your rights may be next!
> > -----------------------------------------------------------------
> >
> >
> >
>
> ----------------------------------------------------
> Shoot-to-Win
>
> Protect the 2nd Amendment
> ----------------------------------------------------
>
>
|
|
1 Message in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
