Loading ...
Sorry, an error occurred while loading the content.

Re: [SeattleRobotics] Re: 1 DOF legs

Expand Messages
  • Peter Balch
    John ... In which case, in your mind, absolutely every robot has at least 7 DOF because it can be tossed up into the air. I expect the rest of us will go on
    Message 1 of 28 , Apr 1, 2008
    • 0 Attachment
      John

      >> I am determining DOF in terms of the robot motion possible,

      In which case, in your mind, absolutely every robot has at least 7 DOF
      because it can be tossed up into the air. I expect the rest of us will go on
      using the term in its more usual sense.

      > You were confused how a robot with only 4 actuators can get motion in 7
      > DOF.

      No I wasn't. My question was:

      "I've searched the web but have found no analysis of how it achieves all
      those gaits".

      I am impressed by Roboquad's abilities and would like to understand how Mark
      Tilden did it.

      Peter
    • John Palmisano
      Peter, I think you are being over-conflictive here ;) Just incubate what I said for a bit. Remember, I am making a distinction between DOF of joints (ie
      Message 2 of 28 , Apr 1, 2008
      • 0 Attachment
        Peter, I think you are being over-conflictive here ;)

        Just incubate what I said for a bit. Remember, I am making a distinction
        between DOF of joints (ie actuator count) and DOF of a robot as a whole
        (such as all those motions that the roboquad can do that impresses you). I
        am not saying they are the same thing! =P

        Sorry for badly wording that in previous emails . . .

        Anyway, why don't you hack it open to check out the mechanics? I found a few
        images on google, but the images weren't the best . . . share for all of us
        to see!

        John
        societyofrobots.com



        2008/4/1, Peter Balch <peterbalch@...>:
        >
        > John
        >
        > >> I am determining DOF in terms of the robot motion possible,
        >
        > In which case, in your mind, absolutely every robot has at least 7 DOF
        > because it can be tossed up into the air. I expect the rest of us will go
        > on
        > using the term in its more usual sense.
        >
        > > You were confused how a robot with only 4 actuators can get motion in 7
        > > DOF.
        >
        > No I wasn't. My question was:
        >
        > "I've searched the web but have found no analysis of how it achieves all
        > those gaits".
        >
        > I am impressed by Roboquad's abilities and would like to understand how
        > Mark
        > Tilden did it.
        >
        > Peter
        >
        >
        >



        --
        John Palmisano

        Robotics Specialist
        Naval Research Laboratory, Washington, DC
        B.S. Mech. Eng., Robotics, Carnegie Mellon University
        www.societyofrobots.com


        [Non-text portions of this message have been removed]
      • Peter Balch
        John ... If you will (knowingly) say such contentious things ... ... I don t expect you ll ever get me to call if DOF but I d be interested if you have some
        Message 3 of 28 , Apr 1, 2008
        • 0 Attachment
          John

          > I think you are being over-conflictive here ;)

          If you will (knowingly) say such contentious things ...


          > Just incubate what I said for a bit. Remember, I am making a distinction
          > between DOF of joints (ie actuator count) and DOF of a robot as a whole

          I don't expect you'll ever get me to call if "DOF" but I'd be interested if
          you have some new theories about how timing gives a robot extra flexibility.

          Clearly, it does. But I've always considered each robot as a special case.
          Are you suggesting that there are general rules that can apply to all or
          many dynamic robots? Perhaps dynamic is the wrong word. It's more that they
          have a history.

          > Anyway, why don't you hack it open to check out the mechanics? I found a
          > few
          > images on google, but the images weren't the best . . . share for all of
          > us
          > to see!

          Will do but the mechanics are extremely simple. The pix on the web show it
          all. It's the design of the gaits that gives roboquad its range of
          behaviours. I'm surprised no-one else has been intriqued. But, given the
          total absence on replies, it looks like no-one has.

          Peter
        • Randy M. Dumse
          Peter Balch said: Tuesday, April 01, 2008 2:59 PM ... As far as I know, you re the only one who has one. I ve only seen one video of it, and wasn t that
          Message 4 of 28 , Apr 1, 2008
          • 0 Attachment
            Peter Balch said: Tuesday, April 01, 2008 2:59 PM
            > the web show it all. It's the design of the gaits that gives
            > roboquad its range of behaviours. I'm surprised no-one else
            > has been intriqued.

            As far as I know, you're the only one who has one. I've only
            seen one video of it, and wasn't that impressed at that time.
            I'm only curious now because of your comments. So I'm not
            surprized you're not getting feedback. You're ahead of the pack.


            Randy
            www.newmicros.com
          • John Palmisano
            ... As KJohn suggested, perhaps we need a new term to distinguish the two . . . ideas? ... flexibility Well, it basically requires taking advantage of
            Message 5 of 28 , Apr 1, 2008
            • 0 Attachment
              > I don't expect you'll ever get me to call if "DOF"
              As KJohn suggested, perhaps we need a new term to distinguish the two . . .
              ideas?

              > I'd be interested if
              > you have some new theories about how timing gives a robot extra
              flexibility
              Well, it basically requires taking advantage of time-dependent physical
              properties of the system. For example, a walking robot can do a lot with
              friction - if you control it just right. Swimming bots - viscosity. Winged
              flying and swimming bots can take advantage of fluid wakes - time it
              properly and your bot can get huge efficiency boosts. Or playing with
              inertia/friction and getting your robot car to 'drift' sideways, despite not
              having any actuator that would directly cause horizontal motion. Dozens more
              examples . . .

              John
              societyofrobots.com


              2008/4/1, Peter Balch <peterbalch@...>:
              >
              > John
              >
              > > I think you are being over-conflictive here ;)
              >
              > If you will (knowingly) say such contentious things ...
              >
              > > Just incubate what I said for a bit. Remember, I am making a distinction
              > > between DOF of joints (ie actuator count) and DOF of a robot as a whole
              >
              > I don't expect you'll ever get me to call if "DOF" but I'd be interested
              > if
              > you have some new theories about how timing gives a robot extra
              > flexibility.
              >
              > Clearly, it does. But I've always considered each robot as a special case.
              >
              > Are you suggesting that there are general rules that can apply to all or
              > many dynamic robots? Perhaps dynamic is the wrong word. It's more that
              > they
              > have a history.
              >
              > > Anyway, why don't you hack it open to check out the mechanics? I found a
              >
              > > few
              > > images on google, but the images weren't the best . . . share for all of
              >
              > > us
              > > to see!
              >
              > Will do but the mechanics are extremely simple. The pix on the web show it
              >
              > all. It's the design of the gaits that gives roboquad its range of
              > behaviours. I'm surprised no-one else has been intriqued. But, given the
              > total absence on replies, it looks like no-one has.
              >
              > Peter
              >
              >
              >



              --
              John Palmisano

              Robotics Specialist
              Naval Research Laboratory, Washington, DC
              B.S. Mech. Eng., Robotics, Carnegie Mellon University
              www.societyofrobots.com


              [Non-text portions of this message have been removed]
            • gmitsuoka
              Not that far ahead. From my Roboquad review published in Robot Magazine back in October, 2007: Roboquad is surprisingly mobile. Although each of its four legs
              Message 6 of 28 , Apr 2, 2008
              • 0 Attachment
                Not that far ahead.

                From my Roboquad review published in Robot Magazine back in October, 2007:

                "Roboquad is surprisingly mobile. Although each of its four legs has
                only a single degree of freedom, through clever geometry and
                coordinated motion, it can walk quite well in any direction, rotate in
                place, raise itself, bow and dance."

                Roboquad is a great toy and is cleverly designed. An easy way to
                explain its mobility is to think of each of the legs as an asymmetric
                wheel. If each leg were replaced with a round wheel, Roboquad would
                actually be quite a bit more mobile, but would lose its ability to
                raise and tilt, and wouldn't look nearly as cool.

                --George

                --- In SeattleRobotics@yahoogroups.com, "Randy M. Dumse" <rmd@...> wrote:
                >
                > Peter Balch said: Tuesday, April 01, 2008 2:59 PM
                > > the web show it all. It's the design of the gaits that gives
                > > roboquad its range of behaviours. I'm surprised no-one else
                > > has been intriqued.
                >
                > As far as I know, you're the only one who has one. I've only
                > seen one video of it, and wasn't that impressed at that time.
                > I'm only curious now because of your comments. So I'm not
                > surprized you're not getting feedback. You're ahead of the pack.
                >
                >
                > Randy
                > www.newmicros.com
                >
              • Peter Balch
                Randy ... (Sorry about the delay in replying - I was away on holiday.) George ... That s a very interesting way of looking at it. It makes a lot of sense. If
                Message 7 of 28 , Apr 11, 2008
                • 0 Attachment
                  Randy

                  > From: "Randy M. Dumse" <rmd@...>
                  > As far as I know, you're the only one who has one. I've only
                  > seen one video of it, and wasn't that impressed at that time.
                  > m only curious now because of your comments. So I'm not
                  > surprized you're not getting feedback. You're ahead of the pack.

                  (Sorry about the delay in replying - I was away on holiday.)


                  George

                  > From: "gmitsuoka" <georgemitsuoka@...>
                  > Roboquad is a great toy and is cleverly designed. An easy way to
                  > explain its mobility is to think of each of the legs as an asymmetric
                  > wheel. If each leg were replaced with a round wheel, Roboquad would
                  > actually be quite a bit more mobile, but would lose its ability to
                  > raise and tilt, and wouldn't look nearly as cool.

                  That's a very interesting way of looking at it. It makes a lot of sense.

                  If one had 4 wheels angled so their axles formed a square-base pyramid then
                  you would indeed get all the translations and rotation of Roboquad. To, say,
                  move forwards, you'd rotate the left wheels cw and the right wheels ccw. If
                  the wheels were highly asymmetric (i.e. just a foot) then the bot would
                  appear to "walk" rather than roll forwards. Of course, wheels rotate
                  continuously in the same direction.

                  Such an asymmetric wheel is off the ground when the foot is 'distal' and on
                  the ground when the foot is 'proximal'. That's the crux of the problem with
                  a 1DOF leg - you can't separate the 'drive' part of the step from the 'lift'
                  part. If the foot alternately rotates cw and ccw then it will alternately
                  drive forwards and backwards. If the foot continuously rotates cw then it
                  will always drive in the same direction. Your explanation made perfect
                  sense.

                  So I got Roboquad to walk forwards fully expecting the legs to rotate
                  continuously in the same direction.

                  They don't. They all alternately rotate cw and ccw to form a "gait".

                  They're not just acting as asymmetric wheels. There's something else going
                  on.

                  Somehow, the "backward drive" part of each step doesn't happen.

                  I guess I'll have to make a movie and analyse the frames.

                  Peter
                • Peter Balch
                  I still haven t got very far understanding how RoboQuad walks. I made a video and analysed the frames but it doesn t help much. The design looks like it s
                  Message 8 of 28 , Apr 13, 2008
                  • 0 Attachment
                    I still haven't got very far understanding how RoboQuad walks. I made a
                    video and analysed the frames but it doesn't help much.

                    The design looks like it's based on BEAM walker techniques.

                    2-motor BEAM walkers have the back leg axis running front-back and the front
                    leg axis running up-down. That's easy to understand because it's
                    asymmetrical front/back. (Many have the front leg axis running at 45 deg to
                    make the legs step over obstacles.) I've built one of those myself and its
                    very simple to program. The motors are 90deg out of phase so it walks
                    forward because of the mechanical asymmetry.

                    Rotating the back legs determines whether the weight is carried on the
                    Front-left and Back-right feet or the Front-right and Back-left feet. If the
                    weight is on the Front-left and Back-right legs feet then the front motor
                    rotates so the Front-left foot moves backwards and the body is propelled
                    forwards; the Back-right foot skids along the floor. (The Front-right and
                    Back-left feet have little weight on them so they also skid.) RoboQuad
                    doesn't seem to do much skidding with the 'down' feet.

                    RoboQuad is (almost) symmetrical front/back. It can crab left/right and it's
                    certainly symmetrical left/right. So the direction of movement is determined
                    by the stepping pattern - not the mechanical asymmetry. Maybe.

                    Or maybe not. The gait for walking backwards is very different from the gait
                    for walking forwards. Is that because it's asymmetrical front/back? The
                    front is slightly higher than the back. I'd initially assumed that the
                    asymmetry was just aesthetic.

                    There are very few 4-motor BEAM walkers around that have the motor axes
                    almost vertical. I've found two:
                    http://www.geocities.com/SiliconValley/Lakes/5081/MECI4.html
                    http://www.solarbotics.net/bestiary/pix/2504_walkers_4/2504_stryderstyle.jpg
                    Neither of which are explained. Neither have videos.

                    (As an aside, if you make Roboquad walk forwards while suspended in the air,
                    the front legs gradually move to the wrong position - they're pointing
                    backwards instead of forwards. Roboquad's motors aren't servos so all
                    positioning is done via blind timing. The motors have a sensor switch which
                    allows the controller to zero the motor position - you can see them in
                    http://www.evosapien.com/robosapien-hack/nocturnal2/node/752. But apparently
                    the algorithms doesn't keep re-zeroing the motors during the walk. It relies
                    on the forces generated to keep the motors in the right position. Weird.)

                    I've yet to look closely at the backwards, sideways and rotating gaits.

                    It's a very subtle machine.

                    Back to analysing video frames ...

                    Peter
                  • Peter Balch
                    I made some videos of roboquad, analysed the frames, wrote a simulation and made some animations from the simulation. I ve uploaded the results as Roboquad.zip
                    Message 9 of 28 , Apr 19, 2008
                    • 0 Attachment
                      I made some videos of roboquad, analysed the frames, wrote a simulation and
                      made some animations from the simulation.

                      I've uploaded the results as Roboquad.zip to:

                      http://tech.groups.yahoo.com/group/SeattleRobotics/files/

                      Even though I can see that the physical Roboquad walks and the simulation
                      walks in much the same way, I still don't really understand why.

                      I think it's best to imagine it as a "2-motor BEAM walker" (Google for that
                      and you'll find lots). In a 2-motor BEAM walker, one pair of legs (usually
                      the rear) transfers the weight between one or other pair of diagonally
                      opposite legs. The other pair of legs (usually the front) provides
                      propulsion while the other legs skid. I think Roboquad does something
                      similar but acts a a "double-ended" 2-motor walker. But a 2-motor walker
                      walks because the front is different from the back but Roboquad is
                      symmetrical.

                      Weird.

                      Peter
                    • Alan KM6VV
                      Hi Peter, Interesting study! I m not sure I understand any more about the gait either! I m not so much interested in BEAM walkers, or this apparent
                      Message 10 of 28 , Apr 19, 2008
                      • 0 Attachment
                        Hi Peter,

                        Interesting study! I'm not sure I understand any more about the gait
                        either!

                        I'm not so much interested in BEAM walkers, or this apparent derivative,
                        but eventually I'd like to get a 2DOF quad working.

                        Alan KM6VV

                        Peter Balch wrote:

                        >I made some videos of roboquad, analysed the frames, wrote a simulation and
                        >made some animations from the simulation.
                        >
                        >I've uploaded the results as Roboquad.zip to:
                        >
                        >http://tech.groups.yahoo.com/group/SeattleRobotics/files/
                        >
                        >Even though I can see that the physical Roboquad walks and the simulation
                        >walks in much the same way, I still don't really understand why.
                        >
                        >I think it's best to imagine it as a "2-motor BEAM walker" (Google for that
                        >and you'll find lots). In a 2-motor BEAM walker, one pair of legs (usually
                        >the rear) transfers the weight between one or other pair of diagonally
                        >opposite legs. The other pair of legs (usually the front) provides
                        >propulsion while the other legs skid. I think Roboquad does something
                        >similar but acts a a "double-ended" 2-motor walker. But a 2-motor walker
                        >walks because the front is different from the back but Roboquad is
                        >symmetrical.
                        >
                        >Weird.
                        >
                        >Peter
                        >
                        >
                      • Peter Balch
                        Alan ... Do you mean 2 DOF/leg or 2 DOF in total - a 2-motor walker? As you know, there are lots of 2-motor walkers on the web. My experience was that it s
                        Message 11 of 28 , Apr 20, 2008
                        • 0 Attachment
                          Alan

                          > I'm not so much interested in BEAM walkers, or this apparent derivative,
                          > but eventually I'd like to get a 2DOF quad working.

                          Do you mean 2 DOF/leg or 2 DOF in total - a 2-motor walker?

                          As you know, there are lots of 2-motor walkers on the web. My experience was
                          that it's easy to get them walking forwards/backwards but much harder to
                          turn. Mine turned rather crudely: it adjusted the front legs until the feet
                          were horizontal. That put the weight equally on all 4 feet. Then it turned
                          the back legs. The result was that the front legs skidded sideways and the
                          robot turned. To get the back legs to grip and the front legs to skid, I had
                          to put little rubber socks on the back feet.

                          My 2-motor walker had the front and back legs the opposite way round from
                          most people's. See it on http://www.peterbalch.btinternet.co.uk/walker2.htm

                          If you mean a 2 DOF/leg walker then that's a very different matter.
                          Previously I'd have assumed that a 2 DOF/leg walker would effectively turn
                          as a skid-steer. It would be very clumsy and inefficient. It would have the
                          motor axes all left-right. But having seen how good Roboquad is, I now
                          believe there's going to be a clever and elegant solution involving strange
                          motor axis angles. But I've not idea what that solution would be - I'd never
                          have
                          designed Roboquad because I wouldn't expect it to work.


                          Peter
                        • Dale Weber
                          ... The RoboQuad is a 1DOF quadrapod - it has only a single DOF in each leg. It moves remarkably well too, including crab walking, turning, etc. 8-Dale -- I
                          Message 12 of 28 , Apr 20, 2008
                          • 0 Attachment
                            On Sunday 20 April 2008 03:11:35 Peter Balch wrote:
                            > Alan
                            >
                            > > I'm not so much interested in BEAM walkers, or this apparent derivative,
                            > > but eventually I'd like to get a 2DOF quad working.
                            >
                            > Do you mean 2 DOF/leg or 2 DOF in total - a 2-motor walker?

                            The RoboQuad is a 1DOF quadrapod - it has only a single DOF in each leg. It
                            moves remarkably well too, including crab walking, turning, etc.

                            8-Dale
                            --
                            I can handle complexity. It's the simple things that confound me.
                            The Dynaplex Network - http://www.thedynaplex.org and
                            Hybrid Robotics - http://www.hybotics.com
                          • Alan KM6VV
                            Hi Peter, I was thinking of 2DOF per leg. Maybe even 3DOF. I was thinking of something like G-Dog. It s to be a kit from HPI Racing. It’s 14cm tall and 40cm
                            Message 13 of 28 , Apr 20, 2008
                            • 0 Attachment
                              Hi Peter,

                              I was thinking of 2DOF per leg. Maybe even 3DOF. I was thinking of
                              something like G-Dog. It's to be a kit from HPI Racing. It’s 14cm tall
                              and 40cm long, has 9 servos. About $1000 this July. Video:

                              http://www.botjunkie.com/

                              OK, I don't want to spend $1000, but I think we could learn from it! But
                              that project is a ways off for now.

                              Alan KM6VV

                              Peter Balch wrote:

                              >Alan
                              >
                              >Do you mean 2 DOF/leg or 2 DOF in total - a 2-motor walker?
                              >
                              >As you know, there are lots of 2-motor walkers on the web. My experience was
                              >that it's easy to get them walking forwards/backwards but much harder to
                              >turn. Mine turned rather crudely: it adjusted the front legs until the feet
                              >were horizontal. That put the weight equally on all 4 feet. Then it turned
                              >the back legs. The result was that the front legs skidded sideways and the
                              >robot turned. To get the back legs to grip and the front legs to skid, I had
                              >to put little rubber socks on the back feet.
                              >
                              >My 2-motor walker had the front and back legs the opposite way round from
                              >most people's. See it on http://www.peterbalch.btinternet.co.uk/walker2.htm
                              >
                              >If you mean a 2 DOF/leg walker then that's a very different matter.
                              >Previously I'd have assumed that a 2 DOF/leg walker would effectively turn
                              >as a skid-steer. It would be very clumsy and inefficient. It would have the
                              >motor axes all left-right. But having seen how good Roboquad is, I now
                              >believe there's going to be a clever and elegant solution involving strange
                              >motor axis angles. But I've not idea what that solution would be - I'd never
                              >have
                              >designed Roboquad because I wouldn't expect it to work.
                              >
                              >
                              >Peter
                              >
                              >
                            • Dick Curtiss
                              How about strapping on stilts and observe how you manage to walk (or fall on your nose). It is not exactly 1 DOF, but might shed some light on the subject.
                              Message 14 of 28 , Apr 20, 2008
                              • 0 Attachment
                                How about strapping on stilts and observe how you manage to walk (or fall on
                                your nose).
                                It is not exactly 1 DOF, but might shed some light on the subject.

                                Dick


                                -----Original Message-----
                                From: SeattleRobotics@yahoogroups.com
                                [mailto:SeattleRobotics@yahoogroups.com] On Behalf Of Alan KM6VV
                                Sent: Sunday, April 20, 2008 3:40 PM
                                To: SeattleRobotics@yahoogroups.com
                                Subject: Re: [SeattleRobotics] Re: 1 DOF legs

                                Hi Peter,

                                I was thinking of 2DOF per leg. Maybe even 3DOF. I was thinking of
                                something like G-Dog. It's to be a kit from HPI Racing. It's 14cm tall
                                and 40cm long, has 9 servos. About $1000 this July. Video:

                                http://www.botjunkie.com/

                                OK, I don't want to spend $1000, but I think we could learn from it! But
                                that project is a ways off for now.

                                Alan KM6VV

                                Peter Balch wrote:

                                >Alan
                                >
                                >Do you mean 2 DOF/leg or 2 DOF in total - a 2-motor walker?
                                >
                                >As you know, there are lots of 2-motor walkers on the web. My experience
                                was
                                >that it's easy to get them walking forwards/backwards but much harder to
                                >turn. Mine turned rather crudely: it adjusted the front legs until the feet
                                >were horizontal. That put the weight equally on all 4 feet. Then it turned
                                >the back legs. The result was that the front legs skidded sideways and the
                                >robot turned. To get the back legs to grip and the front legs to skid, I
                                had
                                >to put little rubber socks on the back feet.
                                >
                                >My 2-motor walker had the front and back legs the opposite way round from
                                >most people's. See it on
                                http://www.peterbalch.btinternet.co.uk/walker2.htm
                                >
                                >If you mean a 2 DOF/leg walker then that's a very different matter.
                                >Previously I'd have assumed that a 2 DOF/leg walker would effectively turn
                                >as a skid-steer. It would be very clumsy and inefficient. It would have the
                                >motor axes all left-right. But having seen how good Roboquad is, I now
                                >believe there's going to be a clever and elegant solution involving strange
                                >motor axis angles. But I've not idea what that solution would be - I'd
                                never
                                >have
                                >designed Roboquad because I wouldn't expect it to work.
                                >
                                >
                                >Peter
                                >
                                >



                                ------------------------------------

                                Visit the SRS Website at http://www.seattlerobotics.orgYahoo! Groups Links
                              • Peter Balch
                                Alan ... Yes, it does seem rather overpriced. I was unimpressed by its gait in the video. As I said before I d have assumed that a 2 DOF/leg walker would
                                Message 15 of 28 , Apr 21, 2008
                                • 0 Attachment
                                  Alan

                                  > I was thinking of
                                  > something like G-Dog. It's to be a kit from HPI Racing. It’s 14cm tall
                                  > and 40cm long, has 9 servos.
                                  > OK, I don't want to spend $1000, but I think we could learn from it! But
                                  > that project is a ways off for now.

                                  Yes, it does seem rather overpriced. I was unimpressed by its gait in the
                                  video. As I said before "I'd have assumed that a 2 DOF/leg walker would
                                  effectively turn as a skid-steer. It would be very clumsy and inefficient.
                                  It would have the motor axes all left-right." All of which is depressingly
                                  true of GDog.

                                  > I was thinking of 2DOF per leg. Maybe even 3DOF.

                                  3DOF is the classic route. I assumed that was the only reasonable way to go
                                  for bipeds until I saw a 2DOF ankle-twister. I assumed 1DOF legs were
                                  pointless until I saw Roboquad. I guess I just lack imagination. Sigh.

                                  Peter
                                Your message has been successfully submitted and would be delivered to recipients shortly.