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

Re: [SeattleRobotics] Kalman Filter sample code

Expand Messages
  • Jim Noetzel
    There is also sample sample code here: http://huhepl.harvard.edu/~rlee/minos/kalman/
    Message 1 of 19 , Oct 2, 2004
    • 0 Attachment
      There is also sample sample code here:

      http://huhepl.harvard.edu/~rlee/minos/kalman/



      John M. Ahlstrom wrote:

      >----- Original Message -----
      >From: "raymond melton" To: <SeattleRobotics@yahoogroups.com>
      >Sent: Friday, October 01, 2004 10:17 PM
      >Subject: Re: [SeattleRobotics] Kalman Filter sample code
      >
      >
      >
      >
      >>>OK, I've posted my Kalman filter code to:
      >>>
      >>>
      >>>
      >>>
      >>http://cylonrobot.blogspot.com/2004/10/sample-code-for-kalman-
      >>
      >>
      >>>filter.html
      >>>
      >>>
      >>Link didn't work. Maybe because I was using Netscape.
      >>
      >>
      >
      >
      >The link didn't work because the line wrapped.
      >
      >
      >Try this link. It's the same one with a shortened link.
      >
      >http://tinyurl.com/5f64u
      >
      >JohnA(Kent,WA)
      >
      >
      >
      >Visit the SRS Website at http://www.seattlerobotics.org
      >Yahoo! Groups Links
      >
      >
      >
      >
      >
      >
      >
      >
      >
      >
      >
    • Doug Kelley
      We ll certainly never try to eliminate any technology from the contest. But, the distance from the start to the destination will increase with time,
      Message 2 of 19 , Oct 2, 2004
      • 0 Attachment
        We'll certainly never try to eliminate any technology from the contest.
        But, the distance from the start to the destination will increase with
        time, particularly if bonus cones are attempted and we aren't likely to
        make the surfaces (grass, pavement, gravel, sand, etc.) easier. As
        distance increases, more turns are necessary and terrain becomes more
        difficult, pure dead reckoning will become less attractive.

        Sensor fusion is the way to go. Twenty feet (probably less with
        averaging) will look good at some point. One thing we aren't currently
        intending to do is allow some sort of beacons on the course. Been
        there, done that!

        Doug

        John McIvor wrote:

        >I have been doing a lot of thinking about RoboMagellan this past week.
        >What went wrong, what went right. In considering the use of GPS in the
        >robot, and how it should be integrated into the navigation system, at
        >what point it is more of hindrance than an aid. What is the best
        >tolerance to an ACTUAL point on the earth that a GPS can spit out, what
        >is the tolerance buildup between GPS units? Here's an example from the
        >contest. The endpoint location is described as MM.mmm 3 place fractional
        >minute. Two different people get two other readings a short time later
        >that are off that measured value by .003 and .005 I believe it was, this
        >is a difference of 15 to 25 feet, the entire circular cement area (where
        >the finish was) is about 40 feet across (a tolerance of about 12 degrees
        >over 250 feet). It is easy to picture an error circle about the size of
        >the cement circle. Now that's definitely a doable range to use a camera
        >to locate a bright orange cone. So we can't say that GPS is not a
        >reasonable device to use, but is it the best device? What if the contest
        >area was 100 feet, what about 50 feet ? When would it be no longer
        >viable?
        >
        >What tolerance would you need to maintain using dead reckoning to be on
        >the same scale of accuracy as GPS given the space ( let's say 500 feet
        >as that has been discussed ) ? If you use both dead reckoning and GPS,
        >which would you put as a higher priority? if one has a higher priority
        >why have the other, specifically if dead reckoning had a higher priority
        >than GPS why would you want to pollute your calculations with an
        >absolute measurement that had a looser error tolerance? My thought was
        >to use the GPS to record a running error circle from the calculated
        >location, but I wonder now, if that is not just more likely to induce
        >error than correct it.
        >
        >Can we build our own higher accuracy GPS type devices ? Possibly yes
        >given a local signal beacon system, not likely if we are talking about
        >getting the data from satellites.
        >
        >I believe there is more room for creative solutions to ever increasing
        >accuracy in dead reckoning.
        >
        >Please Please Please, Don't design the contest to force out dead
        >reckoning. If anything keep in mind how we could balance the rules and
        >contest area to allow a more equal ground for either or both combined.
        >In response to a suggestion to make the event more Darpa like, I think
        >the only way to make the event more Darpa like would be to turn it into
        >a money spending contest and I would surely lose in a money spending
        >contest :) so I would vote against that.
        >
        >My robot is currently "resting in pieces" while I am working on
        >building new stuff for it. The fiberglass bumper broke, I presume from
        >crashing into the wall. So that needs to be remade, and I am adding a
        >gear driven optical encoder into the center differential.
        >
        >Thanks again for all the hard work, planning and organizing!
        >
        >...John
        >
        >
        >
        >
        >Visit the SRS Website at http://www.seattlerobotics.org
        >Yahoo! Groups Links
        >
        >
        >
        >
        >
        >
        >
        >
        >.
        >
        >
        >
      • Brian Dean
        ... Agreed. But the orange cone is essentially a beacon, especially since you are shooing away people with orange shirts. I personally would feel a lot
        Message 3 of 19 , Oct 2, 2004
        • 0 Attachment
          On Sat, Oct 02, 2004 at 08:09:29AM -0700, Doug Kelley wrote:

          > Sensor fusion is the way to go. Twenty feet (probably less with
          > averaging) will look good at some point. One thing we aren't currently
          > intending to do is allow some sort of beacons on the course. Been
          > there, done that!

          Agreed. But the orange cone is essentially a beacon, especially since
          you are shooing away people with orange shirts. I personally would
          feel a lot better if you did not shoo people away who happen to be
          wearing orange shirts - that should just be part of the challenge.

          But also, I do think efforts should be made to keep the course
          relatively clear of people - folks should not be allowed to encircle
          the contestants, I like the golf analogy - there might be several
          folks on the course that are supposed to be there to enforce the rules
          and keep the course clear, etc, but most folks are kept back a bit.
          Perhaps the course personnel and judges could wear one of those
          plastic orange vests so that folks know they are one of the contest
          administrators and are easily identified :-)

          Also, while I haven't seen the latest round of rules being proposed, a
          good rule of thumb is to keep the rules as simple as possible, with
          few, if any exceptions. Generally, if you have exceptions, that's an
          indication of a problem somewhere else - better to fix that problem
          instead of making the exception.

          Also, you should rarely, if ever, have to invoke something that I've
          heard quite a bit during the past few months of discussion regarding
          this contest, and that is "enforcing the spirit of the rules". This
          basically means that the rules don't really mean what they say. It
          should be clear, up front, what the robot designer is required to do
          to get a good score. For example, I remember a discussion a few
          months ago where someone proposed building a robot to get within the
          general vicinity of the cone, and then begin swinging a long arm
          around in order to touch the cone. One of the contest admins
          indicated that that method would not count as it is against the
          "spirit of the rules". But the rules stated that the robot needs to
          navigate to and touch a cone. That's what I'm talking about here.
          The rules should be clear.

          I do see you addressing this type of thing when you said the new rules
          state something to the effect of a robot must fit within a 4'x4'x4'
          cube. That's the kind of rule that is good in the sense that it
          eliminates someone building a 12' pole on their robot that they can
          use to touch cones from a long way away. It does not eliminate a 12'
          telesoping pole, though, and IMO, that should still be allowed, unless
          you specifically state something like the robot must simulataneous
          navigate to within a 4' radius of the cone and then touch the cone.
          That makes it clear what the expectations are, and you don't have to
          resort to enforcing the spirit of the rules, which might not be clear.
          If you ask 10 people what the spirit of the rules are, you'll probably
          get at least 5 differenct answers. But the rules speak for themselves
          and anyone that can read can interpret them.

          And I'll make one last effort - why limit to 50 lbs? You previously
          said for safety reasons, but I think you also said that each
          contestant is responsible for their own robot and any damage it may
          cause. This seems to me that the robot weight should be left as a
          design decision of the builder, and they take responsibilty for it.

          Anyway, there aren't many robot contests that really get me
          interested. This one has some serious potential, in the same way that
          the actual DARPA Grand Challenge which inspired this contest is very
          interesting. The reason is that it doesn't seem too contrived - folks
          need to solve real world problems in order to compete. Solving line
          mazes just don't push the technology forward. But solving real world
          problems does. While drafting the rules, please try hard to retain
          the "real world" aspect, and minimize _any_ allowances that are
          designed to make the goal easier. It is the "real world" aspect, I
          think, that has generated so much interest in this contest in the
          first place.

          -Brian
          --
          Brian Dean
          BDMICRO - ATmega128 Based MAVRIC Controllers
          http://www.bdmicro.com/
        • Doug Kelley
          I wouldn t have as big a problem with orange shirts if the same person stood in the same place for the entire contest. It s not really an obstacle thing; it s
          Message 4 of 19 , Oct 2, 2004
          • 0 Attachment
            I wouldn't have as big a problem with orange shirts if the same person
            stood in the same place for the entire contest. It's not really an
            obstacle thing; it's a fairness thing. But, I do like the idea of
            orange shirts for the judges! ;-)

            I think you'll be pleased with the rule changes. Nothing is really
            changed but there are a lot of clarifications to help us with the
            "spirit of the rules" issue you brought up. Robot builders are a clever
            sort and we want to keep the focus of the contest on navigation, not
            simple finding a cone (which I agree is a type of beacon).

            See my previous post about the 50 pound limit; there are several reasons
            other than safety to keep the limit where it is.

            Glad you are motivated to build a Magellan bot!

            Doug

            Brian Dean wrote:

            >On Sat, Oct 02, 2004 at 08:09:29AM -0700, Doug Kelley wrote:
            >
            >
            >
            >>Sensor fusion is the way to go. Twenty feet (probably less with
            >>averaging) will look good at some point. One thing we aren't currently
            >>intending to do is allow some sort of beacons on the course. Been
            >>there, done that!
            >>
            >>
            >
            >Agreed. But the orange cone is essentially a beacon, especially since
            >you are shooing away people with orange shirts. I personally would
            >feel a lot better if you did not shoo people away who happen to be
            >wearing orange shirts - that should just be part of the challenge.
            >
            >But also, I do think efforts should be made to keep the course
            >relatively clear of people - folks should not be allowed to encircle
            >the contestants, I like the golf analogy - there might be several
            >folks on the course that are supposed to be there to enforce the rules
            >and keep the course clear, etc, but most folks are kept back a bit.
            >Perhaps the course personnel and judges could wear one of those
            >plastic orange vests so that folks know they are one of the contest
            >administrators and are easily identified :-)
            >
            >Also, while I haven't seen the latest round of rules being proposed, a
            >good rule of thumb is to keep the rules as simple as possible, with
            >few, if any exceptions. Generally, if you have exceptions, that's an
            >indication of a problem somewhere else - better to fix that problem
            >instead of making the exception.
            >
            >Also, you should rarely, if ever, have to invoke something that I've
            >heard quite a bit during the past few months of discussion regarding
            >this contest, and that is "enforcing the spirit of the rules". This
            >basically means that the rules don't really mean what they say. It
            >should be clear, up front, what the robot designer is required to do
            >to get a good score. For example, I remember a discussion a few
            >months ago where someone proposed building a robot to get within the
            >general vicinity of the cone, and then begin swinging a long arm
            >around in order to touch the cone. One of the contest admins
            >indicated that that method would not count as it is against the
            >"spirit of the rules". But the rules stated that the robot needs to
            >navigate to and touch a cone. That's what I'm talking about here.
            >The rules should be clear.
            >
            >I do see you addressing this type of thing when you said the new rules
            >state something to the effect of a robot must fit within a 4'x4'x4'
            >cube. That's the kind of rule that is good in the sense that it
            >eliminates someone building a 12' pole on their robot that they can
            >use to touch cones from a long way away. It does not eliminate a 12'
            >telesoping pole, though, and IMO, that should still be allowed, unless
            >you specifically state something like the robot must simulataneous
            >navigate to within a 4' radius of the cone and then touch the cone.
            >That makes it clear what the expectations are, and you don't have to
            >resort to enforcing the spirit of the rules, which might not be clear.
            >If you ask 10 people what the spirit of the rules are, you'll probably
            >get at least 5 differenct answers. But the rules speak for themselves
            >and anyone that can read can interpret them.
            >
            >And I'll make one last effort - why limit to 50 lbs? You previously
            >said for safety reasons, but I think you also said that each
            >contestant is responsible for their own robot and any damage it may
            >cause. This seems to me that the robot weight should be left as a
            >design decision of the builder, and they take responsibilty for it.
            >
            >Anyway, there aren't many robot contests that really get me
            >interested. This one has some serious potential, in the same way that
            >the actual DARPA Grand Challenge which inspired this contest is very
            >interesting. The reason is that it doesn't seem too contrived - folks
            >need to solve real world problems in order to compete. Solving line
            >mazes just don't push the technology forward. But solving real world
            >problems does. While drafting the rules, please try hard to retain
            >the "real world" aspect, and minimize _any_ allowances that are
            >designed to make the goal easier. It is the "real world" aspect, I
            >think, that has generated so much interest in this contest in the
            >first place.
            >
            >-Brian
            >
            >
          • Michael
            I haven t chimed in yet as much of what I have heard sounds good. I am a little bit concerned about the distance increasing. As everyone knows, LevelHeaded
            Message 5 of 19 , Oct 2, 2004
            • 0 Attachment
              I haven't chimed in yet as much of what I have heard sounds good. I
              am a little bit concerned about the distance increasing. As
              everyone knows, LevelHeaded is based on a small chasis, my battery
              carry capacitity is limited, and thus my distance is limited. The
              current distances should be no problem, but by increasing the
              distances over time you have and are specifically effecting a core
              design and this could require significant expense to upgrade.

              You can achieve similiar goals by just selecting a challenging
              path. I do applaud the selection of the finish cone this time
              around, the obstacles around it were complex (pillars, garbage cans,
              light poles, and a significant number of overhangs created by park
              benches). Complex enough to catch all entrants that got near.

              Further, what about the thought that there be at least one other
              cone you must hit, and thus eliminating the straight shot approach.

              Michael


              --- In SeattleRobotics@yahoogroups.com, Doug Kelley <yahoo@k...>
              wrote:
              > We'll certainly never try to eliminate any technology from the
              contest.
              > But, the distance from the start to the destination will increase
              with
              > time, particularly if bonus cones are attempted and we aren't
              likely to
              > make the surfaces (grass, pavement, gravel, sand, etc.) easier.
              As
              > distance increases, more turns are necessary and terrain becomes
              more
              > difficult, pure dead reckoning will become less attractive.
              >
              > Sensor fusion is the way to go. Twenty feet (probably less with
              > averaging) will look good at some point. One thing we aren't
              currently
              > intending to do is allow some sort of beacons on the course. Been
              > there, done that!
              >
              > Doug
              >
              > John McIvor wrote:
              >
              > >I have been doing a lot of thinking about RoboMagellan this past
              week.
              > >What went wrong, what went right. In considering the use of GPS
              in the
              > >robot, and how it should be integrated into the navigation
              system, at
              > >what point it is more of hindrance than an aid. What is the best
              > >tolerance to an ACTUAL point on the earth that a GPS can spit
              out, what
              > >is the tolerance buildup between GPS units? Here's an example
              from the
              > >contest. The endpoint location is described as MM.mmm 3 place
              fractional
              > >minute. Two different people get two other readings a short time
              later
              > >that are off that measured value by .003 and .005 I believe it
              was, this
              > >is a difference of 15 to 25 feet, the entire circular cement area
              (where
              > >the finish was) is about 40 feet across (a tolerance of about 12
              degrees
              > >over 250 feet). It is easy to picture an error circle about the
              size of
              > >the cement circle. Now that's definitely a doable range to use a
              camera
              > >to locate a bright orange cone. So we can't say that GPS is not a
              > >reasonable device to use, but is it the best device? What if the
              contest
              > >area was 100 feet, what about 50 feet ? When would it be no longer
              > >viable?
              > >
              > >What tolerance would you need to maintain using dead reckoning to
              be on
              > >the same scale of accuracy as GPS given the space ( let's say 500
              feet
              > >as that has been discussed ) ? If you use both dead reckoning
              and GPS,
              > >which would you put as a higher priority? if one has a higher
              priority
              > >why have the other, specifically if dead reckoning had a higher
              priority
              > >than GPS why would you want to pollute your calculations with an
              > >absolute measurement that had a looser error tolerance? My
              thought was
              > >to use the GPS to record a running error circle from the
              calculated
              > >location, but I wonder now, if that is not just more likely to
              induce
              > >error than correct it.
              > >
              > >Can we build our own higher accuracy GPS type devices ? Possibly
              yes
              > >given a local signal beacon system, not likely if we are talking
              about
              > >getting the data from satellites.
              > >
              > >I believe there is more room for creative solutions to ever
              increasing
              > >accuracy in dead reckoning.
              > >
              > >Please Please Please, Don't design the contest to force out dead
              > >reckoning. If anything keep in mind how we could balance the
              rules and
              > >contest area to allow a more equal ground for either or both
              combined.
              > >In response to a suggestion to make the event more Darpa like, I
              think
              > >the only way to make the event more Darpa like would be to turn
              it into
              > >a money spending contest and I would surely lose in a money
              spending
              > >contest :) so I would vote against that.
              > >
              > >My robot is currently "resting in pieces" while I am working on
              > >building new stuff for it. The fiberglass bumper broke, I presume
              from
              > >crashing into the wall. So that needs to be remade, and I am
              adding a
              > >gear driven optical encoder into the center differential.
              > >
              > >Thanks again for all the hard work, planning and organizing!
              > >
              > >...John
              > >
              > >
              > >
              > >
              > >Visit the SRS Website at http://www.seattlerobotics.org
              > >Yahoo! Groups Links
              > >
              > >
              > >
              > >
              > >
              > >
              > >
              > >
              > >.
              > >
              > >
              > >
            • Ted Larson
              I think this suggestion is a fantastic one! In fact, I thought that this would be the case originally, and was surprised that ALL the cones were considered
              Message 6 of 19 , Oct 2, 2004
              • 0 Attachment
                I think this suggestion is a fantastic one! In fact, I thought that this
                would be the case originally, and was surprised that ALL the cones were
                considered bonuses except for one, and that you really only needed to touch
                only one cone to complete the course. I think there should be several cones
                on the basic course, with a couple of others considered bonuses.

                So...I 2nd this suggestion.

                - Ted


                -----Original Message-----
                From: Michael [mailto:BotMaster@...]
                Sent: Saturday, October 02, 2004 10:48 AM
                To: SeattleRobotics@yahoogroups.com
                Subject: [SeattleRobotics] Re: Robomegellan - NOT a FLOP



                ......
                Further, what about the thought that there be at least one other cone you
                must hit, and thus eliminating the straight shot approach.

                .......
              • raymond melton
                Someone recently posted a url here that linked to a little birdy robot walker. I (foolishly) deleted the eletter that had the url. It looked for all the
                Message 7 of 19 , Oct 3, 2004
                • 0 Attachment
                  Someone recently posted a url here that linked to a
                  little birdy robot walker. I (foolishly) deleted the
                  eletter that had the url. It looked for all the world
                  like a robotic Malard duck, though the colors were off
                  a bit. Does anyone recall this, and have the url?

                  Regards, Ray.




                  __________________________________
                  Do you Yahoo!?
                  New and Improved Yahoo! Mail - 100MB free storage!
                  http://promotions.yahoo.com/new_mail
                • dan michaels
                  ... There s one shown here .... may not be what you re after ... http://vakuumtv.c3.hu/collection/robottortenet/robot.html - dan michaels www.oricomtech.com
                  Message 8 of 19 , Oct 3, 2004
                  • 0 Attachment
                    --- In SeattleRobotics@yahoogroups.com, raymond melton
                    <rtmelton@y...> wrote:
                    > Someone recently posted a url here that linked to a
                    > little birdy robot walker. I (foolishly) deleted the
                    > eletter that had the url. It looked for all the world
                    > like a robotic Malard duck, though the colors were off
                    > a bit. Does anyone recall this, and have the url?
                    >
                    > Regards, Ray.


                    There's one shown here .... may not be what you're after ...

                    http://vakuumtv.c3.hu/collection/robottortenet/robot.html


                    - dan michaels
                    www.oricomtech.com
                    =========================
                  • Ted Larson
                    Was it this guy? http://kaduhi.com/weird-7/ It kinda looks like a bird. = Ted ... From: dan michaels [mailto:dan@oricomtech.com] Sent: Sunday, October 03, 2004
                    Message 9 of 19 , Oct 3, 2004
                    • 0 Attachment
                      Was it this guy?
                      http://kaduhi.com/weird-7/

                      It kinda looks like a bird.

                      = Ted


                      -----Original Message-----
                      From: dan michaels [mailto:dan@...]
                      Sent: Sunday, October 03, 2004 4:55 PM
                      To: SeattleRobotics@yahoogroups.com
                      Subject: [SeattleRobotics] Re: little birdy robot



                      --- In SeattleRobotics@yahoogroups.com, raymond melton <rtmelton@y...>
                      wrote:
                      > Someone recently posted a url here that linked to a little birdy robot
                      > walker. I (foolishly) deleted the eletter that had the url. It
                      > looked for all the world like a robotic Malard duck, though the colors
                      > were off a bit. Does anyone recall this, and have the url?
                      >
                      > Regards, Ray.


                      There's one shown here .... may not be what you're after ...

                      http://vakuumtv.c3.hu/collection/robottortenet/robot.html


                      - dan michaels
                      www.oricomtech.com
                      =========================







                      Visit the SRS Website at http://www.seattlerobotics.org Yahoo! Groups Links
                    • dan michaels
                      ... That s great! He, he. Much better than the one I found. Looks like 2-servo legs, one for femur, one for tibia, should walk good. Better than toddler.
                      Message 10 of 19 , Oct 3, 2004
                      • 0 Attachment
                        --- In SeattleRobotics@yahoogroups.com, Ted Larson <ted@l...> wrote:
                        >
                        > Was it this guy?
                        > http://kaduhi.com/weird-7/
                        >
                        > It kinda looks like a bird.
                        >
                        > = Ted
                        >


                        That's great! He, he. Much better than the one I found. Looks like
                        2-servo legs, one for femur, one for tibia, should walk good. Better
                        than toddler.
                      • Tom Capon
                        I think this one came up recently (looking through my emails) http://www.ai.mit.edu/people/chunks/chunks.html It is an absolutely incredible robotic dinosaur
                        Message 11 of 19 , Oct 4, 2004
                        • 0 Attachment
                          I think this one came up recently (looking through my emails)

                          http://www.ai.mit.edu/people/chunks/chunks.html

                          It is an absolutely incredible robotic dinosaur that walks on two legs, balances with a tail, has quite a number of degrees of freedom in each leg, and has a mock-skeletal dinosaur head on it.

                          dan michaels <dan@...> wrote:


                          --- In SeattleRobotics@yahoogroups.com, Ted Larson wrote:
                          >
                          > Was it this guy?
                          > http://kaduhi.com/weird-7/
                          >
                          > It kinda looks like a bird.
                          >
                          > = Ted
                          >


                          That's great! He, he. Much better than the one I found. Looks like
                          2-servo legs, one for femur, one for tibia, should walk good. Better
                          than toddler.









                          Visit the SRS Website at http://www.seattlerobotics.org
                          Yahoo! Groups Links









                          ---------------------------------
                          Do you Yahoo!?
                          vote.yahoo.com - Register online to vote today!

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