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

Re: [SeattleRobotics] Updates to Loki ('bot) and new pix

Expand Messages
  • Alan KM6VV
    Hi Peter, Yes, the sensor range is limiting (I ll probably pull off the sonar, but this project made a convenient test bed to get I2C and sonar working. ... So
    Message 1 of 26 , Mar 5, 2008
    • 0 Attachment
      Hi Peter,

      Yes, the sensor range is limiting (I'll probably pull off the sonar, but
      this project made a convenient test bed to get I2C and sonar working.

      Peter Balch wrote:

      >Alan
      >
      >> Hitec HS-475 servos 5.5 kg/cm (54Ncm) 61.00 oz/in
      >>Torque above, hip-hip is 5.5". weight is 16 oz, battery adds 6 oz.
      >>
      >>
      So that's 22oz. When it lifts its left leg then the torque on the hip
      will be 22*5.5/2 which is 60.5 oz.inch.

      >Yes, your servos aren't quite strong enough.
      >
      >The rule of thumb we came up with before that, for walking, the servos need at least 150% of the max "static" torque. 200% was more than is essary
      >
      >Admittedly, that rule was deduced from just 3 examples. But it seems about right if your bot can't quite walk with the batteries.
      >
      >Without the batteries, the max static torque is 41.25 oz.inch and the servos can provide 148% of that.
      >
      >(As you say, at 6V the servos can do better. What are you running them at?)
      >
      >
      I'm running them at 6v, maybe even a little more. I have '475 servos
      that I have been reluctant to "steal" from my Shelob (octapod) project.
      I should probably test with these servos, just to see their "fit" for
      this 'bot.

      So I have to get even bigger servos for the knees? this was supposed to
      be a "cheap" project! Essentially no-cost, as I had a batch of the
      S3004 servos on hand. If I have to put in an even more expensive servo
      ('645?) at the the knees then it kind of defeats the purpose!

      HS-645 MG
      #

      # voltage = 4.8 - 6.0vdc
      # Torque = 133 oz.-in.
      #


      OR loose weight (the 'bot)!

      >
      >That was probably me. It's what I do but I don't think I'd "recommend" it. As you say, it's just easier that way.
      >
      >
      I have been running the gait cycle to always end at a particular state
      (stance), so that the feet don't get tangled up so easily. That's been
      working. I have enough servo data (all 4 for each state) so that I can
      transverse either forward or backward through my gait cycle (an array of
      SSC32 servo commands). Not terribly compact, but then I don't need a
      lot of them.

      My parser easily picks up the servo settings out of the strings.
      Anyway, it dawned on me this morning that I could "reverse" in mid gait,
      rather then running it to the end of the gait. That would essentially
      "back up" the 'bot on the when it recognizes an obstacle. I might even
      be able to segway into a turning gait. I'll have to look into that.

      >>IR about 30", Sonar about 15' !; although I've yet to get any more then
      >>about 30" out of it, even with setting up the gain and range.
      >>
      >>
      >What use is 30"? Going down the centre of a corridor? What sort of "missions" are you planning?
      >
      >
      I'm not happy with the sonar range either at this point. But with the
      limited speed the 'bot is traveling at, I probably don't need more then
      15" or so anyway. What are you getting from your IR sensors? I'm
      thinking I may just use them. Maybe add an additional pair looking
      rearward.

      >Is it good at closer distances?
      >
      >
      Ok right up close to the 'bot, as far as I can tell. I'm still testing
      both range finders. And it looks like I'll need to "calibrate" the IR
      sensors as well.

      >Mine point outward at 30deg. But I've no real justification for that. What I found was that it's easy to spot things directly ahead but harder to avoid clipping the corner of an object. So I put mine at the outer edges of the robot. Given that the body swings around so much it's hard to work out the ideal.
      >
      >Peter
      >
      >
      I've seen 30 degrees, I think for the sonar as the beam width. Is that
      30 degrees back from straight forward?. Each? I'm about 16-17
      degrees each... for now on mine.

      Perhaps I can determine the field of the IR, and adjust accordingly.

      Alan KM6VV
    • Peter Balch
      Alan ... Are you suggesting putting the more powerful servos in the knees? Aren t the hips more heavily loaded? ... I ve finally managed to find a cheap source
      Message 2 of 26 , Mar 6, 2008
      • 0 Attachment
        Alan

        > Yeah, maybe more. I need a pair of HS-475HB servos for the knees.
        > So I have to get even bigger servos for the knees?

        Are you suggesting putting the more powerful servos in the knees? Aren't the hips more heavily loaded?

        > this was supposed to be a "cheap" project!

        I've finally managed to find a cheap source of nice ESky micro-servos in the UK so I can do cheap now.

        > Anyway, it dawned on me this morning that I could "reverse" in mid gait,

        Is it easy to read the sensors in mid-gait? I have trouble doing that because some of the code interferes with the servo timing. That's the downside or a homebrew single-chip solution of course.

        > What are you getting from your IR sensors?

        Maybe 10" max. All I'm interested in is not walking into coffee mugs so anything beyond 6" doesn't matter.

        > I'm thinking I may just use them.

        That's why I was asking about your "missions". What is the sonar for? I'd seen the walker as a toy so if it just does the basics of turning away from obstacles and maybe phototaxis, I'm happy enough.

        Your Sharp GP2D12 are a lot better than my simple reflectance sensors so you can make more intelligent decisions. What are you planning?

        >>Mine point outward at 30deg.
        >I've seen 30 degrees, I think for the sonar as the beam width. Is that
        >30 degrees back from straight forward?. Each?

        Yes, 30 degrees outwards from straight forward. Both.

        Their effective beam width is about +/- 20deg.

        What I really miss is an edge-of-table sensor - probably on the toes. I guess yours is so big it walks on the floor.

        I just bought a tip-over detector but I don't know if it's possible for the walker to get up.

        Peter


        [Non-text portions of this message have been removed]
      • Alan KM6VV
        Hi Peter, ... The knees lift the weight of the bot over the lever arm (width) of the bot, while the hips swing the weight in the plane. Even if the plane
        Message 3 of 26 , Mar 6, 2008
        • 0 Attachment
          Hi Peter,

          Peter Balch wrote:

          >Alan
          >
          >Are you suggesting putting the more powerful servos in the knees? Aren't the hips more heavily loaded?
          >
          >
          The knees lift the weight of the bot over the lever arm (width) of the
          'bot, while the hips "swing" the weight in the plane. Even if the plane
          is elevated, the angle is not as much as in the case of the knees. No?

          >I've finally managed to find a cheap source of nice ESky micro-servos in the UK so I can do cheap now.
          >
          >
          >
          That would be interesting. The little DonTronics DT106 board might be a
          better fit. Even for Loki, as the 'bot width would be reduced, and less
          torque would be required to lift the other hip. Although I DO like the
          stance of my Loki.

          >Is it easy to read the sensors in mid-gait? I have trouble doing that because some of the code interferes with the servo timing. That's the downside or a homebrew single-chip solution of course.
          >
          >
          I can read the IR's at any time. I start up an A/D cycle (system tick)
          and read one of the 8 channels each 100 mS (interrupt reads/saves
          results), and thus the result is always available. I'll add some
          filtering later. Not very fast sampling, but no waits. I could speed
          it up if necessary, but with the slow speed of the 'bot, I doubt if I
          will "over-drive" my sensors. Interrupts to handle the A/D makes this
          sensor quite easy.

          >Maybe 10" max. All I'm interested in is not walking into coffee mugs so anything beyond 6" doesn't matter.
          >
          >
          I agree.

          >That's why I was asking about your "missions". What is the sonar for? I'd seen the walker as a toy so if it just does the basics of turning away from obstacles and maybe phototaxis, I'm happy enough.
          >
          >Your Sharp GP2D12 are a lot better than my simple reflectance sensors so you can make more intelligent decisions. What are you planning?
          >
          >
          The sonar was an early "test". I had been trying to get I2R working for
          inter-board comms on my hexapod robot (comms not working yet), and the
          sonar was an excuse to try a simplified I2C example. It may not stay on
          my Loki. But it does look cool there! The I2C code will be further
          generalized to work with other sensors as well (I can probably do that
          now), and hopefully inter-board stuff too (hexapod).

          Loki will mostly be a simple demo, a toy if you like. A convenient way
          to test code. the sonar and Bluetooth (IR's too) unplug quite easily
          for use elsewhere. If it roams the floor and teases the cats, I'm
          happy. It should allow me to test some simple avoidance behaviors. I
          certainly don't need absolute distances, but I was happy to observe all
          three sensors reporting the same distance to a sheet of copy paper in
          some simple experiments yesterday. The IR's are quite directional! I
          thought maybe they weren't "calibrated" too well before, but I think
          they're just seeing different objects sitting next to my keyboard on my
          desk.

          If the parts cost could be minimized (mainly servo), I think it would be
          a good robot club intro project, as the body parts can be cut out of
          cheap PCB stock.

          But what to do with the programming? A 'C' compiler is probably not a
          good choice. One could use a "BASIC module" (like LM uses) to implement
          the 'bot, but that raises the price! I can enter individual moves for
          the servos and save them in EEPROM, but I'm more interested in making
          behavior editing available. Thoughts?

          >Yes, 30 degrees outwards from straight forward. Both.
          >
          >Their effective beam width is about +/- 20deg.
          >
          >
          That seems like a lot for IR's.

          >What I really miss is an edge-of-table sensor - probably on the toes. I guess yours is so big it walks on the floor.
          >
          >I just bought a tip-over detector but I don't know if it's possible for the walker to get up.
          >
          >Peter
          >
          Loki doesn't seem to mind the table, but he Tends to ketch his feet on
          the carpet. I have put off designing a foot switch; however I do have a
          small hole centered near each toe that could be used for a pivoting toe
          piece. I could probably mount a small micro switch there as well.

          The original Loki has a spring wire going through a piece of tubing with
          a bent end to sense the ground. I have several good detailed pix from
          David Buckley (not on his website) that illustrates this if you're
          interested.

          Alan KM6VV
        • Phazer
          Hi Alan, The battery pack, while it can be secured below the controller board ... You could try splitting up the battery pack and placing it in the robots
          Message 4 of 26 , Mar 6, 2008
          • 0 Attachment
            Hi Alan,

            The battery pack, while it can be secured below the controller board
            > adequately, is still a little to heavy for the S3004 servos, so I leave
            > the battery back connected via a cable.
            >

            You could try splitting up the battery pack and placing it in the robots
            feet. I'm not sure how much of a difference that really makes, but it should
            at least cut the weight that needs to be lifted at any given time in half.

            ~Patrick


            [Non-text portions of this message have been removed]
          • Randy M. Dumse
            Phazer said: Thursday, March 06, 2008 4:42 PM ... You know Patrick, that s clever. Even more clever might be a robot design where the feet weigh very little.
            Message 5 of 26 , Mar 6, 2008
            • 0 Attachment
              Phazer said: Thursday, March 06, 2008 4:42 PM
              > You could try splitting up the battery pack and placing it in
              > the robots feet. I'm not sure how much of a difference that
              > really makes, but it should at least cut the weight that
              > needs to be lifted at any given time in half.

              You know Patrick, that's clever.

              Even more clever might be a robot design where the feet weigh
              very little. You take and move them out where you want to place
              them, then you drag the whole weight of the robot over the top
              of them, while placing another foot ahead.

              It's a different way of thinking about the problem, and I find
              myself surprised by how different it makes things looks.

              Randy
              www.newmicros.com
            • Alan KM6VV
              Hi Patrick, That s an interesting idea! There is almost a shelf on each foot. That should cut the weight that has to be lifted down. I ll give it some
              Message 6 of 26 , Mar 6, 2008
              • 0 Attachment
                Hi Patrick,

                That's an interesting idea! There is almost a shelf on each foot. That
                should cut the weight that has to be lifted down. I'll give it some
                thought.

                Alan KM6VV

                Phazer wrote:

                >Hi Alan,
                >
                >The battery pack, while it can be secured below the controller board
                >
                >
                >>adequately, is still a little to heavy for the S3004 servos, so I leave
                >>the battery back connected via a cable.
                >>
                >>
                >>
                >
                >You could try splitting up the battery pack and placing it in the robots
                >feet. I'm not sure how much of a difference that really makes, but it should
                >at least cut the weight that needs to be lifted at any given time in half.
                >
                >~Patrick
                >
                >
                >
              • Dale Weber
                ... I have found that the sonar tends to be more accurate and at a closer distance than IR. I usually pair a Sharp IR with a PING, but am looking forward to
                Message 7 of 26 , Mar 7, 2008
                • 0 Attachment
                  On Wednesday 05 March 2008 16:00:12 Alan KM6VV wrote:
                  > Hi Peter,
                  >
                  > Yes, the sensor range is limiting (I'll probably pull off the sonar, but
                  > this project made a convenient test bed to get I2C and sonar working.

                  I have found that the sonar tends to be more accurate and at a closer
                  distance than IR. I usually pair a Sharp IR with a PING, but am looking
                  forward to using some of the I2C capable sensors (I have an I2C compass and
                  RTC module).

                  > I'm running them at 6v, maybe even a little more. I have '475 servos
                  > that I have been reluctant to "steal" from my Shelob (octapod) project.
                  > I should probably test with these servos, just to see their "fit" for
                  > this 'bot.

                  When I finally build my Octapod, I won't use less than HS-645MG servos, even
                  though they are more expensive. I have to operate on MsBRAT to replace her
                  left knee joint, and did get a replacement for it.

                  MsBRAT has all HS-475HB servos, which seem to be plenty good enough for this
                  project. I just wish I didn't have to use HSERVO on the Atom PRO. I am
                  finding it quite difficult to manually program a good gait.

                  > So I have to get even bigger servos for the knees? this was supposed to
                  > be a "cheap" project! Essentially no-cost, as I had a batch of the
                  > S3004 servos on hand. If I have to put in an even more expensive servo
                  > ('645?) at the the knees then it kind of defeats the purpose!

                  The knees, and maybe the hipV servos are the heavy lifting joints in a
                  Hexapod or Octapod, so they may need stronger servos. The knees can be
                  especially strained.

                  > My parser easily picks up the servo settings out of the strings.
                  > Anyway, it dawned on me this morning that I could "reverse" in mid gait,
                  > rather then running it to the end of the gait. That would essentially
                  > "back up" the 'bot on the when it recognizes an obstacle. I might even
                  > be able to segway into a turning gait. I'll have to look into that.

                  My normal robot behavior on detecting an obstacle is to backup a bit before
                  it takes any action. This often helps get it unstuck from things like table
                  and chair legs or low floor type obstacles. This should work for any kind of
                  robot. This is where the programmers of the RoboQuad failed miserably in my
                  opinion - not having the robot backup before taking action after an obstacle
                  detection.

                  > Ok right up close to the 'bot, as far as I can tell. I'm still testing
                  > both range finders. And it looks like I'll need to "calibrate" the IR
                  > sensors as well.

                  The Sharp IR rangers, if that's what you are using, have always been a bit of
                  a problem for me. One of these days I am going to have to calibrate one and
                  get a better set of value/distance readings. The data I have is not as
                  accurate as I would like. Somewhere, I saw a calculation that can be done,
                  but it still requires some calibration for each IR sensor.

                  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
                • Dale Weber
                  Hi Peter, ... Which IR sensor are you using? On the Sharp IR Rangers I get from Lynxmotion, I get up to about 64 cm (approximately 25 inches) for reliable
                  Message 8 of 26 , Mar 7, 2008
                  • 0 Attachment
                    Hi Peter,

                    On Thursday 06 March 2008 12:49:58 Peter Balch wrote:

                    > Maybe 10" max. All I'm interested in is not walking into coffee mugs so
                    > anything beyond 6" doesn't matter.

                    Which IR sensor are you using? On the Sharp IR Rangers I get from Lynxmotion,
                    I get up to about 64 cm (approximately 25 inches) for reliable readings. The
                    IR sensors don't seem to react as fast as a PING though.

                    > What I really miss is an edge-of-table sensor - probably on the toes. I
                    > guess yours is so big it walks on the floor.

                    I'm thinking about putting some SES holes in the feet of MsBRAT so I can
                    mount an IR ranger on each foot. I'd rather use a PING, but it has been
                    strongly suggested I use an IR sensor instead.

                    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
                  • dan michaels
                    ... testing both range finders. And it looks like I ll need to calibrate the IR sensors as well. ... always been a bit of a problem for me. One of these
                    Message 9 of 26 , Mar 7, 2008
                    • 0 Attachment
                      --- In SeattleRobotics@yahoogroups.com, Dale Weber <robotguy@...>
                      wrote:
                      >

                      >
                      > > Ok right up close to the 'bot, as far as I can tell. I'm still
                      testing both range finders. And it looks like I'll need
                      to "calibrate" the IR sensors as well.
                      >
                      > The Sharp IR rangers, if that's what you are using, have
                      always been a bit of a problem for me. One of these days I am going
                      to have to calibrate one and get a better set of value/distance
                      readings. The data I have is not as accurate as I would like.
                      Somewhere, I saw a calculation that can be done, but it still
                      requires some calibration for each IR sensor.
                      >
                      > 8-Dale



                      There can be some problems with the Sharp IR Rangers, depending on
                      how you mount them, and also how fast your bot is moving. They
                      require about 50-msec to produce an output, by averaging several
                      samples, so if your bot is moving or turning, they can give false
                      readings. Since they work by triangulation, you can see that small
                      changes in geometry between sensor and target can be a problem. I had
                      this trouble with a GP2D12 mounted on a servo.

                      Also, Sharp has an appnote about mounting the sensors. You can get it
                      here among other places.

                      http://www.oricomtech.com/download.htm#Misc2
                      Sharp AppNote on Distance Measuring Sensors (GP2D12, etc) -
                      rangers.pdf (370 Kbytes)
                    • Alan KM6VV
                      I m reading the sensors at a fairly slow rate, and I use a conversion equation derived from this document:
                      Message 10 of 26 , Mar 7, 2008
                      • 0 Attachment
                        I'm reading the sensors at a fairly slow rate, and I use a conversion
                        equation derived from this document:

                        http://maslab.csail.mit.edu/2002/wiki/SharpGP2D12_E.html

                        dan michaels wrote:

                        >There can be some problems with the Sharp IR Rangers, depending on
                        >how you mount them, and also how fast your bot is moving. They
                        >require about 50-msec to produce an output, by averaging several
                        >samples, so if your bot is moving or turning, they can give false
                        >readings. Since they work by triangulation, you can see that small
                        >changes in geometry between sensor and target can be a problem. I had
                        >this trouble with a GP2D12 mounted on a servo.
                        >
                        >Also, Sharp has an appnote about mounting the sensors. You can get it
                        >here among other places.
                        >
                        >http://www.oricomtech.com/download.htm#Misc2
                        >Sharp AppNote on Distance Measuring Sensors (GP2D12, etc) -
                        >rangers.pdf (370 Kbytes)
                        >
                        >
                        I have this document as well, found it after the one I sited above. I
                        haven't used the sensors enough to get a feel for them in use, but off a
                        white reflector, they seem to be close, even agreeing with my sonar.

                        I can see where the mounting might be critical. I have them fairly out
                        in front on my 'bot.

                        I would like to work out a "clustering" algorithm, or some sort of a
                        filter to eliminate the spikes that are reported to be a common artifact.

                        Alan KM6VV
                      • dan michaels
                        ... above. I haven t used the sensors enough to get a feel for them in use, but off a white reflector, they seem to be close, even agreeing with my sonar. ...
                        Message 11 of 26 , Mar 7, 2008
                        • 0 Attachment
                          --- In SeattleRobotics@yahoogroups.com, Alan KM6VV <KM6VV@...> wrote:
                          >

                          > >Sharp AppNote on Distance Measuring Sensors (GP2D12, etc) -
                          > >rangers.pdf (370 Kbytes)
                          > >
                          > >
                          > I have this document as well, found it after the one I sited
                          above. I haven't used the sensors enough to get a feel for them in
                          use, but off a white reflector, they seem to be close, even agreeing
                          with my sonar.
                          >
                          > I can see where the mounting might be critical. I have them fairly
                          out in front on my 'bot.
                          >


                          This is probably a mistake, espcially with the GP2D12. In the MIT
                          link you gave, you'll notice 2 different drawings of the GP2 response
                          curve. The "second" one shows that, for ranges less than 4" for the
                          GP2D12, there is a "foldback" in the curve, and you will not be able
                          to determine the correct distance, close in or far away.

                          So, better to mount the sensors so they are never < 4" from any
                          target, meaning on the rear of the robot.

                          Also, as Loki sweeps the sensor by rotating the feet forwards and
                          backwards, ie producing a horizontal pan, the Sharp document
                          indicates you will do better with the sensors mounted "vertically".


                          >
                          > I would like to work out a "clustering" algorithm, or some sort of
                          a filter to eliminate the spikes that are reported to be a common
                          artifact.
                          >


                          Use a 10-uF capacitor, or other lowpass filter, on the analog output
                          pin instead.
                        • Alan KM6VV
                          ... Yeah, I realized that. Might be a good reason to retain the sonar. I ve seen the foldback. At first I thought I could simply use logic to exclude the
                          Message 12 of 26 , Mar 7, 2008
                          • 0 Attachment
                            dan michaels wrote:

                            >>I can see where the mounting might be critical. I have them fairly out in front on my 'bot.
                            >>
                            >>
                            >
                            >This is probably a mistake, espcially with the GP2D12. In the MIT
                            >link you gave, you'll notice 2 different drawings of the GP2 response
                            >curve. The "second" one shows that, for ranges less than 4" for the
                            >GP2D12, there is a "foldback" in the curve, and you will not be able
                            >to determine the correct distance, close in or far away.
                            >
                            >So, better to mount the sensors so they are never < 4" from any
                            >target, meaning on the rear of the robot.
                            >
                            >Also, as Loki sweeps the sensor by rotating the feet forwards and
                            >backwards, ie producing a horizontal pan, the Sharp document
                            >indicates you will do better with the sensors mounted "vertically".
                            >
                            >
                            >
                            Yeah, I realized that. Might be a good reason to retain the sonar.
                            I've seen the foldback. At first I thought I could simply use logic to
                            exclude the < 4" range, but obviously that doesn't work! The sonar data
                            can be used to rule out that foldback.

                            Keeping the IR sensors > 4" away from an obstacle is not really feasible
                            on this small 'bot. I don't see any way of mounting them on the back of
                            the 'bot, unless they are on a tower above the 'bot.

                            Loki does sweep the body (after inclining it) a limited amount while
                            rotating the feet, but I'm currently only checking the distance at the
                            end of a gait cycle. Maybe this is incorrect. Maybe I need to keep a
                            "history" (buffer) of sensor distances, and go back over them to
                            determine ranging to an obstacle. It would be interesting to study a
                            buffer of data as Loki walks towards an obstacle.

                            >
                            >
                            >>I would like to work out a "clustering" algorithm, or some sort of a filter to eliminate the spikes that are reported to be a common
                            >>artifact.
                            >>
                            >>
                            >
                            >Use a 10-uF capacitor, or other lowpass filter, on the analog output
                            >pin instead.
                            >
                            >
                            A cap? That would certainly "smooth" out the data! But I'm thinking
                            I'd rather do it in software.

                            Alan KM6VV
                          • Jon Hylands
                            ... Look at using ProxDots... Very small, very low power, use it as a close truth sensor to compliment the Sharp sensors... Later, Jon ... Jon Hylands
                            Message 13 of 26 , Mar 7, 2008
                            • 0 Attachment
                              On Fri, 07 Mar 2008 13:53:10 -0800, Alan KM6VV <KM6VV@...> wrote:

                              > Loki does sweep the body (after inclining it) a limited amount while
                              > rotating the feet, but I'm currently only checking the distance at the
                              > end of a gait cycle. Maybe this is incorrect. Maybe I need to keep a
                              > "history" (buffer) of sensor distances, and go back over them to
                              > determine ranging to an obstacle. It would be interesting to study a
                              > buffer of data as Loki walks towards an obstacle.

                              Look at using ProxDots...

                              Very small, very low power, use it as a close "truth" sensor to compliment
                              the Sharp sensors...

                              Later,
                              Jon

                              --------------------------------------------------------------
                              Jon Hylands Jon@... http://www.huv.com/jon

                              Project: Micro Raptor (Small Biped Velociraptor Robot)
                              http://www.huv.com/blog
                            • Dale Weber
                              Hi Dan, ... My robot, W.A.L.T.E.R. ( www.thedynaplex.org/robotics/rovers/walter ) is always in motion. I ve considered modifying his behavior to stop and do a
                              Message 14 of 26 , Mar 7, 2008
                              • 0 Attachment
                                Hi Dan,

                                On Friday 07 March 2008 08:12:12 dan michaels wrote:
                                > There can be some problems with the Sharp IR Rangers, depending on
                                > how you mount them, and also how fast your bot is moving. They
                                > require about 50-msec to produce an output, by averaging several
                                > samples, so if your bot is moving or turning, they can give false
                                > readings. Since they work by triangulation, you can see that small
                                > changes in geometry between sensor and target can be a problem. I had
                                > this trouble with a GP2D12 mounted on a servo.

                                My robot, W.A.L.T.E.R. ( www.thedynaplex.org/robotics/rovers/walter ) is
                                always in motion. I've considered modifying his behavior to stop and do a
                                scan every so often though. It would also make it easier to use the pan/tilt
                                that has two sensors on it at present. The latest info is in the Version 2.0
                                section. I just mounted my Hammer Carrier Board and an SSC-32 on him, but
                                have not converted the carrier board to battery power yet - I may just wait
                                for the Hammer Bot Board.

                                Hmmm, from the examples in that app note, it looks like the preference is to
                                mount the Sharp IR Rangers vertically instead of horizontally. I have three
                                of them on W.A.L.T.E.R. now, and the front one can't be mounted vertically
                                since it is right below a ping on a pan/tilt platform.

                                I could mount the two on either side of the pan/tilt vertically by adding an
                                SES "L" bracket. I am starting to wonder if the IR Rangers are worth the
                                effort though. They don't seem to do as well as the PING, and the PING
                                returns reliable distances down to 1 or 2 cm without having to use a look up
                                table.

                                I'm considering converting W.A.L.T.E.R. to just use sonar and other non-IR
                                sensors. Do all IR sensors work the same way? Are there other IR Ranging
                                sensors available to us?

                                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
                                Thanks Jon They re small! I might try them. but I ve currently got a pair of Sharp GP2D12 sensors designed and programmed in, and I d like to get some mileage
                                Message 15 of 26 , Mar 7, 2008
                                • 0 Attachment
                                  Thanks Jon

                                  They're small! I might try them.

                                  but I've currently got a pair of Sharp GP2D12 sensors designed and
                                  programmed in, and I'd like to get some mileage out of them first.
                                  Don't your Sumo bots (or Dave's) use these? And right out in front too
                                  as I recall.

                                  OK, read your comments again, and you're suggesting using the ProxDots
                                  as a qualification for other sensors. Yeah, that could work.

                                  Alan KM6VV

                                  Jon Hylands wrote:

                                  >On Fri, 07 Mar 2008 13:53:10 -0800, Alan KM6VV <KM6VV@...> wrote:
                                  >
                                  >
                                  >
                                  >>Loki does sweep the body (after inclining it) a limited amount while
                                  >>rotating the feet, but I'm currently only checking the distance at the
                                  >>end of a gait cycle. Maybe this is incorrect. Maybe I need to keep a
                                  >>"history" (buffer) of sensor distances, and go back over them to
                                  >>determine ranging to an obstacle. It would be interesting to study a
                                  >>buffer of data as Loki walks towards an obstacle.
                                  >>
                                  >>
                                  >
                                  >Look at using ProxDots...
                                  >
                                  >Very small, very low power, use it as a close "truth" sensor to compliment
                                  >the Sharp sensors...
                                  >
                                  >Later,
                                  >Jon
                                  >
                                • dan michaels
                                  ... had this trouble with a GP2D12 mounted on a servo. ... www.thedynaplex.org/robotics/rovers/walter ) is ... and do a scan every so often though. It would
                                  Message 16 of 26 , Mar 7, 2008
                                  • 0 Attachment
                                    --- In SeattleRobotics@yahoogroups.com, Dale Weber <robotguy@...>
                                    wrote:
                                    >
                                    > Hi Dan,
                                    >
                                    > On Friday 07 March 2008 08:12:12 dan michaels wrote:
                                    > > There can be some problems with the Sharp IR Rangers, depending on
                                    > > how you mount them, and also how fast your bot is moving. They
                                    > > require about 50-msec to produce an output, by averaging several
                                    > > samples, so if your bot is moving or turning, they can give false
                                    > > readings. Since they work by triangulation, you can see that small
                                    > > changes in geometry between sensor and target can be a problem. I
                                    had this trouble with a GP2D12 mounted on a servo.
                                    >
                                    > My robot, W.A.L.T.E.R. (
                                    www.thedynaplex.org/robotics/rovers/walter ) is
                                    > always in motion. I've considered modifying his behavior to stop
                                    and do a scan every so often though. It would also make it easier to
                                    use the pan/tilt that has two sensors on it at present. The latest
                                    info is in the Version 2.0 section. I just mounted my Hammer Carrier
                                    Board and an SSC-32 on him, but have not converted the carrier board
                                    to battery power yet - I may just wait for the Hammer Bot Board.
                                    >
                                    > Hmmm, from the examples in that app note, it looks like the
                                    preference is to mount the Sharp IR Rangers vertically instead of
                                    > horizontally.
                                    >


                                    Right, if you're scanning horizontally.


                                    >
                                    >I have three of them on W.A.L.T.E.R. now, and the front one can't be
                                    mounted vertically
                                    > since it is right below a ping on a pan/tilt platform.
                                    >
                                    > I could mount the two on either side of the pan/tilt
                                    vertically by adding an SES "L" bracket. I am starting to wonder if
                                    the IR Rangers are worth the effort though. They don't seem to do as
                                    well as the PING, and the PING returns reliable distances down to 1
                                    or 2 cm without having to use a look up table.
                                    >
                                    > I'm considering converting W.A.L.T.E.R. to just use sonar and
                                    other non-IR sensors. Do all IR sensors work the same way? Are there
                                    other IR Ranging sensors available to us?
                                    >
                                    > 8-Dale


                                    I also use the Ping sonars, and prefer them to the Sharp IR rangers,
                                    and also very much over the SRF0x sonars, which I found to have bad
                                    sidelobe pickup. There is also the GP2D120 [IIRC, ie, the one also
                                    having analog output], with max range 2-3 less than the GP2D12, and
                                    the foldback occurs at a much lower distance, 1" IIRC.

                                    The Rangers are nice in having very narrow beams, but I think you
                                    have to move them slowly for reliable measurements. A moment's
                                    consideration makes it clear that triangulation geometry is very
                                    sensitive to movement and angle changes. I think they might be more
                                    useful for determining the basic outline of an object, than ranging
                                    on a moving bot.

                                    Like Jon mentioned, I also use IR proximity detectors [similar to the
                                    proxdots] for close-in detection, but of course, they don't do
                                    ranging. However, the IRPDs are extremely sensitive to different
                                    reflective surfaces. I made some measurements, once upon a time.

                                    http://www.oricomtech.com/projects/irpd-an.htm#Obj2

                                    Apparently the GP2 devices are not as sensitive to surface
                                    reflectivity, as they use triangulation on whatever reflected pickup
                                    they receive.
                                  • dan michaels
                                    ... feasible on this small bot. I don t see any way of mounting them on the back of the bot, unless they are on a tower above the bot. ... How about on
                                    Message 17 of 26 , Mar 7, 2008
                                    • 0 Attachment
                                      --- In SeattleRobotics@yahoogroups.com, Alan KM6VV <KM6VV@...> wrote:
                                      >

                                      >
                                      > Keeping the IR sensors > 4" away from an obstacle is not really
                                      feasible on this small 'bot. I don't see any way of mounting them on
                                      the back of the 'bot, unless they are on a tower above the 'bot.
                                      >


                                      How about on standoffs above each rotator hip servo?


                                      > >
                                      > >
                                      > >>I would like to work out a "clustering" algorithm, or some sort
                                      of a filter to eliminate the spikes that are reported to be a common
                                      > >>artifact.
                                      > >>
                                      > >>
                                      > >
                                      > >Use a 10-uF capacitor, or other lowpass filter, on the analog
                                      output pin instead.
                                      > >
                                      > >
                                      > A cap? That would certainly "smooth" out the data! But I'm
                                      thinking I'd rather do it in software.
                                      >
                                      > Alan KM6VV
                                      >


                                      Dogbert advice: "Never do in software what can be done more properly
                                      [and easily] in hardware".

                                      http://www.google.com/custom?q=GP2D12+output+filtering
                                      http://oopic.com/gp2d12.htm

                                      There's a reason you need a "hardware" anti-aliasing filter ahead of
                                      your A/D converter. Ie, once the signal is aliased, there is no
                                      recovery.
                                    • Alan KM6VV
                                      ... That s a thought. Although it might make it hard to see right in front of the bot. I m going to have to play with things a little. ... Interesting
                                      Message 18 of 26 , Mar 7, 2008
                                      • 0 Attachment
                                        dan michaels wrote:

                                        >How about on standoffs above each rotator hip servo?
                                        >
                                        >
                                        That's a thought. Although it might make it hard to see right in front
                                        of the bot. I'm going to have to play with things a little.

                                        >http://www.google.com/custom?q=GP2D12+output+filtering
                                        >http://oopic.com/gp2d12.htm
                                        >
                                        >
                                        Interesting search. Guess I'll be doing some additional reading.

                                        >There's a reason you need a "hardware" anti-aliasing filter ahead of
                                        >your A/D converter. Ie, once the signal is aliased, there is no
                                        >recovery.
                                        >
                                        >
                                        I don't think it's an "alias" problem of the A/D, but erroneous data.
                                        Probably why it's called a fold-back.

                                        Alan KM6VV
                                      • Peter Balch
                                        Hi Dale (8-Dale) ... They re homebrews that only cost a few cents each: a bright red LED and an IR sensor in a clear package (fairly short wavelength IR). The
                                        Message 19 of 26 , Mar 9, 2008
                                        • 0 Attachment
                                          Hi Dale (8-Dale)

                                          > > Maybe 10" max. All I'm interested in is not walking into coffee mugs so
                                          > > anything beyond 6" doesn't matter.
                                          >
                                          > Which IR sensor are you using?

                                          They're homebrews that only cost a few cents each: a bright red LED and an IR sensor in a clear package (fairly short wavelength IR). The idea is: measure the light-level, switch on the LED, measure the light-level again. The tricky part is getting the right sort of (approx logarithmic) response out of the IR sensor because it sould give the same response to a reflection whether it's in a bright room or under my desk. But that logarithmic response is all done in s/w so having written it years ago, I can put the code into any new bot.

                                          The advantage of an IR sensor in a clear package is that it's also reasonably sensitive to visible light so I can do photo-taxis for free.

                                          The advantage of a red LED (rather than IR) is that two glowing red "eyes" look cool.

                                          > I'm thinking about putting some SES holes in the feet of MsBRAT so I can
                                          >mount an IR ranger on each foot. I'd rather use a PING, but it has been
                                          >strongly suggested I use an IR sensor instead.

                                          I guess some small cheap reflectance sensors would do.

                                          Peter

                                          [Non-text portions of this message have been removed]
                                        • Peter Balch
                                          Alan ... Yes, sorry. My mistake. I thought your servos were the other way round. ... I like the way you ve used double-sided pcb with no solder resist. I guess
                                          Message 20 of 26 , Mar 9, 2008
                                          • 0 Attachment
                                            Alan

                                            > The knees lift the weight of the bot over the lever arm (width) of the
                                            > 'bot, while the hips "swing" the weight in the plane. Even if the plane
                                            > is elevated, the angle is not as much as in the case of the knees. No?

                                            Yes, sorry. My mistake. I thought your servos were the other way round.

                                            > If the parts cost could be minimized (mainly servo), I think it would be
                                            > a good robot club intro project, as the body parts can be cut out of
                                            > cheap PCB stock.

                                            I like the way you've used double-sided pcb with no solder resist. I guess you just slot them together and then solder the corners.

                                            I'd looked at that sort of technology before I built mine (although my drawings weren't quite as elegant).

                                            My main difficulty was how to make the right-angle "thigh" bracket. I see you've used a (store bought?) metal bracket. No doubt one could make a slotted-pcb bracket.

                                            Anyway, someone suggested it was easy and cheap to get aluminium sheet cut, drilled and bent in 100off quantities so I made the aluminium prototype. I wasn't expecting to turn the biped into a product but I thought it would be good to test the technology and the manufacturers.

                                            It turns out the suggestion was wrong. Such companies in the English Midlands were desparate for business five years ago but now they've all gone bust. The few remaining ones charge huge setup costs and the Chinese are only interested in 10,000off.

                                            But I've also found that funny shaped pcbs are expensive to have made. With the pcb company I use, a rectangular pcb, single-sided with legend, solder-resist and lots of holes, costs half a funny shped pcb with no-copper, no legend, no solder-resist and no holes. I think they just don't like making them. So pcb material might not be cheap for a good robot club intro project. It depends what you call cheap.

                                            > But what to do with the programming? A 'C' compiler is probably not a
                                            > good choice. One could use a "BASIC module" (like LM uses) to implement
                                            > the 'bot, but that raises the price! I can enter individual moves for
                                            > the servos and save them in EEPROM, but I'm more interested in making
                                            > behavior editing available. Thoughts?

                                            As you know, I like to roll my own languages. I've just remembered, I promised to send you some examples about a month ago. I'll do so.

                                            > >Their effective beam width is about +/- 20deg.
                                            > That seems like a lot for IR's.

                                            They're just IR LEDs and IR photo-transistors. So it's average.

                                            It's a good width for obstacle detection if all you're going to do is turn away. But wouldn't be much use for finding an object to pick up.

                                            >Loki doesn't seem to mind the table, but he Tends to ketch his feet on
                                            >the carpet. I have put off designing a foot switch; however I do have a
                                            >small hole centered near each toe that could be used for a pivoting toe
                                            >piece.

                                            How do you mean? As a sensor?

                                            > The original Loki has a spring wire going through a piece of tubing with
                                            > a bent end to sense the ground. I have several good detailed pix from
                                            > David Buckley (not on his website) that illustrates this if you're
                                            > interested.

                                            Please. A spring wire? To a microswitch?

                                            Peter



                                            [Non-text portions of this message have been removed]
                                          • Alan KM6VV
                                            Hi Peter, ... Yes, I slot the joints, and solder. Seems quite strong. I suppose un-clad board stock could be used, and then epoxied together instead. I like
                                            Message 21 of 26 , Mar 9, 2008
                                            • 0 Attachment
                                              Hi Peter,

                                              Peter Balch wrote:

                                              >Alan
                                              >
                                              >>If the parts cost could be minimized (mainly servo), I think it would be
                                              >>a good robot club intro project, as the body parts can be cut out of
                                              >>cheap PCB stock.
                                              >>
                                              >>
                                              >I like the way you've used double-sided pcb with no solder resist. I guess you just slot them together and then solder the corners.
                                              >
                                              >
                                              Yes, I slot the joints, and solder. Seems quite strong. I suppose
                                              un-clad board stock could be used, and then epoxied together instead. I
                                              like soldering. Probably also makes a good "electrical" frame as well.

                                              >I'd looked at that sort of technology before I built mine (although my drawings weren't quite as elegant).
                                              >
                                              >My main difficulty was how to make the right-angle "thigh" bracket. I see you've used a (store bought?) metal bracket. No doubt one could make a slotted-pcb bracket.
                                              >
                                              >
                                              I designed and made a simple L-shaped aluminum bracket. I've bought a
                                              small desk-brake that seems sufficient to do the job of bending small
                                              'bot parts. But yeah, the L-bracket could be made up from notched PCB
                                              stock as well. I'd probably add a pair of triangular gussets to each
                                              for added strength.

                                              >Anyway, someone suggested it was easy and cheap to get aluminium sheet cut, drilled and bent in 100off quantities so I made the aluminium prototype. I wasn't expecting to turn the biped into a product but I thought it would be good to test the technology and the manufacturers.
                                              >
                                              >
                                              I actually wanted to use aluminum (and would have), but short of
                                              MIG-welding (or possibly epoxy), I couldn't figure out how the feet were
                                              put together. Turns out, they're WOOD! (Thin model plywood).

                                              >It turns out the suggestion was wrong. Such companies in the English Midlands were desparate for business five years ago but now they've all gone bust. The few remaining ones charge huge setup costs and the Chinese are only interested in 10,000off.
                                              >
                                              >
                                              Probably expensive here as well. I was thinking students could cut the
                                              pieces out of PCB stock with a scroll saw. I'd hate to do that work
                                              (I'm spoiled w/ CNC), but I think it's reasonable.

                                              >But I've also found that funny shaped pcbs are expensive to have made. With the pcb company I use, a rectangular pcb, single-sided with legend, solder-resist and lots of holes, costs half a funny shped pcb with no-copper, no legend, no solder-resist and no holes. I think they just don't like making them. So pcb material might not be cheap for a good robot club intro project. It depends what you call cheap.
                                              >
                                              >
                                              Could be. Sounds logical to send them out to a PCB house, but I haven't
                                              tried it.

                                              >>But what to do with the programming? A 'C' compiler is probably not a
                                              >>good choice. One could use a "BASIC module" (like LM uses) to implement
                                              >>the 'bot, but that raises the price! I can enter individual moves for
                                              >>the servos and save them in EEPROM, but I'm more interested in making
                                              >>behavior editing available. Thoughts?
                                              >>
                                              >>
                                              >
                                              >As you know, I like to roll my own languages. I've just remembered, I promised to send you some examples about a month ago. I'll do so.
                                              >
                                              >
                                              Thanks, I'd appreciate it. I probably mentioned that my first robot ran
                                              a little 5K BASIC on an 8085 board. I added my own commands to it for
                                              LEFT, RIGHT, etc. for navigation. I downloaded the basic from a TRS-80
                                              notebook computer over a serial link. I think something like that may
                                              be OK for kids. LOGO?

                                              On second thought, if an '877 were used, I believe there is a free
                                              compiler (limited?). But I'm not sure C is the best language for
                                              youngsters!

                                              >>>Their effective beam width is about +/- 20deg.
                                              >>>
                                              >>>
                                              >>That seems like a lot for IR's.
                                              >>
                                              >>
                                              >
                                              >They're just IR LEDs and IR photo-transistors. So it's average.
                                              >
                                              >
                                              That's your sensors then? I believe the GP2D12 sensors are fairly sharp.

                                              >It's a good width for obstacle detection if all you're going to do is turn away. But wouldn't be much use for finding an object to pick up.
                                              >
                                              >
                                              That's all I'm looking for.

                                              http://www.mil.ufl.edu/imdl/papers/IMDL_Report_Summer_05/padilla-joel/bi-mode.pdf

                                              This a paper that talks about an "autonomous multi behavior robot. The
                                              robot will have two modes
                                              of operation: 1) run away (emulating a prey) and 2) follow (emulating a
                                              prey)."

                                              >>Loki doesn't seem to mind the table, but he Tends to ketch his feet on
                                              >>the carpet. I have put off designing a foot switch; however I do have a
                                              >>small hole centered near each toe that could be used for a pivoting toe
                                              >>piece.
                                              >>
                                              >>
                                              >
                                              >How do you mean? As a sensor?
                                              >
                                              >
                                              A short "toe tip" would be added on the pivot, and a spring would force
                                              it down below the fixed (existing toe) by a small distance. When the
                                              moving toe segment came into contact with the floor, it would move up,
                                              and work a mechanical switch contact, Hall Effect sw or similar.

                                              >>The original Loki has a spring wire going through a piece of tubing with
                                              >>a bent end to sense the ground. I have several good detailed pix from
                                              >>David Buckley (not on his website) that illustrates this if you're
                                              >>interested.
                                              >>
                                              >>

                                              >Please. A spring wire? To a microswitch?
                                              >
                                              >Peter
                                              >

                                              David's had 4 spring wires (toes and heels) going through small tubing.

                                              Alan KM6VV






                                              [Non-text portions of this message have been removed]
                                            Your message has been successfully submitted and would be delivered to recipients shortly.