Announcement

Collapse
No announcement yet.

Creative ways to "glue" an axel without violating R01 (LEGO-made building parts)?

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    I've wondered about using stickers as a light shroud. I thought it might be a good idea to place 4 stickers around the color sensor to reduce the amount of ambient light hitting the sensor (which might affect accuracy). Such stickers wouldn't be "construction elements". The stickers wouldn't tape elements together, but they also wouldn't be used "as intended" or "per instructions" since they would be serving a useful function (light shroud) rather than merely decorative.

    In reality, ambient light doesn't seem to be a big problem. The color sensor is mounted as close to the mat as possible and calibrated at the start of the match. Nonetheless, stickers might help particularly if a light sensor was mounted far from the mat.

    Comment


    • #17
      Stickers as shields, I would say no on that.
      FIRST LEGO League Mentor and Referee/Head Referee since 2011.

      Comment


      • #18
        Originally posted by WilliamFrantz View Post
        The color sensor is mounted as close to the mat as possible
        You really should not do that. There is a sweet spot of how far the color sensor should be away from the mat. It is somewhere around 1 hole (8mm). Putting the color sensor too close to the mat is bad.

        Comment


        • #19
          Originally posted by OllyLi View Post
          You really should not do that. There is a sweet spot of how far the color sensor should be away from the mat. It is somewhere around 1 hole (8mm). Putting the color sensor too close to the mat is bad.
          If you look at the geometry of how the light comes out of the LED and is reflected back into the sensing element, when the color sensor is close enough to the mat, no light can make it to the sensing element and it is "always in the dark".

          Comment


          • #20
            Originally posted by OllyLi View Post
            You really should not do that. There is a sweet spot of how far the color sensor should be away from the mat. It is somewhere around 1 hole (8mm). Putting the color sensor too close to the mat is bad.
            There really is no "sweet spot".

            I was doing a lot of line following tests and built a robot where I could quickly change the standoff from zero up to 40 mm. Zero and close to zero is bad because it is so easy to block out the reflected light and get a false dark reading. From 4 mm to 24 mm I got pretty good sensor readings. For line following the larger the sensing area the better, so I got the best results at 24 mm. When set the height ad 40 mm my sensing area was pretty wide, but the feedback range was 0 to 16. I could calibrate the sensor to return 0..100, but plots revealed there were only 16 different intensity levels from black to white.

            Using the light sensor for line squaring I might want a smaller sensor spot. A smaller sensor spot gives you increased sensitivity. At 8 mm standoff 100% sensitivity change might be 8 mm, at 16 mm standoff 100% sensitivity change might be 12 mm. Of course it is unlikely you need all that sensitivity because you cannot move the robot that precisely. For line finding I don't know if a bigger or smaller sensor area is better. A big sensor area might make it easier to ignore small specks, but maybe you want to see those specs.

            Sensor standoff is not some cookbook thing where 8mm is what everyone should use. Different standoffs work better for different tasks.

            Comment


            • #21
              Originally posted by Dean Hystad View Post

              There really is no "sweet spot".

              I was doing a lot of line following tests and built a robot where I could quickly change the standoff from zero up to 40 mm. Zero and close to zero is bad because it is so easy to block out the reflected light and get a false dark reading. From 4 mm to 24 mm I got pretty good sensor readings. For line following the larger the sensing area the better, so I got the best results at 24 mm. When set the height ad 40 mm my sensing area was pretty wide, but the feedback range was 0 to 16. I could calibrate the sensor to return 0..100, but plots revealed there were only 16 different intensity levels from black to white.

              Using the light sensor for line squaring I might want a smaller sensor spot. A smaller sensor spot gives you increased sensitivity. At 8 mm standoff 100% sensitivity change might be 8 mm, at 16 mm standoff 100% sensitivity change might be 12 mm. Of course it is unlikely you need all that sensitivity because you cannot move the robot that precisely. For line finding I don't know if a bigger or smaller sensor area is better. A big sensor area might make it easier to ignore small specks, but maybe you want to see those specs.

              Sensor standoff is not some cookbook thing where 8mm is what everyone should use. Different standoffs work better for different tasks.
              That is good to know. I cannot remember where I saw the data saying color sensor should be places around 8mm (more precisely, I think it said 4mm-12mm) above the mat anymore.
              One thing I think makes sense is that if the sensor spot is too big, it would be hard to find a black/white spot to calibrate the sensor. Currently our robot has the sensor around 10mm above, and the diameter of the red circle is pretty close to the width of black/white strip now. I am not sure where would we calibrate the color sensor if it is moved higher.

              Comment


              • #22
                Originally posted by WilliamFrantz View Post
                I've wondered about using stickers as a light shroud. I thought it might be a good idea to place 4 stickers around the color sensor to reduce the amount of ambient light hitting the sensor (which might affect accuracy).
                Some minifigs come with accessories that might be useful for shielding color sensors. Robot construction parts can also be used. I suspect either would be more practical than stickers, many of which aren't totally opaque anyway.
                Kansas City Region Head Ref 2014-present
                KC Region coaches and teams can ask FLL robot game rules questions at kcfllref@gmail.com

                Comment


                • #23
                  Do you folks think that maybe shielding the light sensors might be unnecessary? My team spent a whole meeting last year shining 1000 watt halogen lights, LED flashlights, fluorescent tubes, etc. at our unshielded sensors, and never moved the reading more than a couple of percent. Moderating the distance to the mat was important, but shielding outside light didn't matter much. Just wondering if others have experimented directly, and what you found.

                  Comment


                  • #24
                    Originally posted by korey99 View Post
                    Do you folks think that maybe shielding the light sensors might be unnecessary? My team spent a whole meeting last year shining 1000 watt halogen lights, LED flashlights, fluorescent tubes, etc. at our unshielded sensors, and never moved the reading more than a couple of percent. Moderating the distance to the mat was important, but shielding outside light didn't matter much. Just wondering if others have experimented directly, and what you found.
                    A TV remote causes more problems than a bright halogen bulb or LED flashlight. In reflected light mode the sensor is primarily interested in light in the red to infrared range.

                    The EV3 light sensor appears to compensate for ambient light, turning the red LED off for a moment and grabbing an ambient light reading. The reading is used to offset the raw sensor value. This works pretty well, but you will lose sensitivity on both ends of the range. If the light is bright enough, say outdoors on a clear day, the ambient light overwhelms the sensors ability to compensate (sensor reading with the lamp off = sensor reading with the lamp on, compensated reading is always 0). Color sensing either has no such compensation or it is much less effective. In experiments I found it easy to confuse the color sensor by flooding the sensing area with different colored ambient light.

                    Whether ambient light causes problems has less to do with shielding and how the sensor works than how you are using the sensor. If you are always looking for high contrast the sensor will be robust. If you are trying to differentiate between dark and really dark it will be less so. One of the kids I'm working with this year uses thresholds of 98 to find a white line or 1 to find a black line. His programs fail more often than they work. Changing the program to use 90 and 10 made a big difference. Now I just need to convince him the only valid threshold is 50.
                    Last edited by Dean Hystad; 11-16-2018, 01:46 PM.

                    Comment


                    • #25
                      One of the first things I teach my team is, "the world is not perfect."

                      Why doesn't the robot drive straight?
                      Why did it work the last time but not this time?
                      Why does the sensor change while stationary?

                      The first answer is always, the world is not perfect. Now, what can we do to mitigate imperfection? How can we be more robust?

                      I had a control systems prof who used to say, "all models are wrong, some are useful." That was always my favorite motto. The kids don't really think in terms of models and controllers but they can understand, "the world is not perfect."

                      Comment

                      Working...
                      X