Announcement

Collapse
No announcement yet.

Off-season training

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

  • Off-season training

    The FLL team I'm mentoring has switched into once-per-week after-school-activity mode. They're all 6-to-8th grade students at a middle school, with a cabinet of robot parts in the computer lab, and a teacher as coach. They use one end of the lab for a two-hour robot club activity, one day per week.

    The team had a good experience at the state championship, not due to world-beating performance, but because of the exposure to other teams. The more mature team members spent some time roaming the pits during the open-house session, talking with other teams about their robots, and the members who were very into the project aspects dove into the display boards and presentations.

    The team has been around for a couple of years, and has accumulated enough parts to build several identical tri-bots. This makes it possible for all of the members to work in parallel on their own solutions to a challenge-of-the-day. These challenges are focused on programming and strategy. In two hours, there is just sufficient time to tackle a simple problem from scratch. Just completing the challenge has been motivation enough to keep the members busy.

    The coach is a PE teacher, and seems to have an unlimited supply of sports tape. This tape works really well against the tile floor in the lab for use with the light sensor.

    Up to this point, the team's mission planning style has been all about odometry. So the coach and I have been posing a series of challenges focused on getting experience with use of sensors for navigation.

    The first week, they were all doing basic straight line tracking, with varying degrees of success.

    The second week, an obstacle course was taped-out on the floor. The goal was to reliably complete the course. Line tracking wasn't necessary, just line detection was sufficient.

    Last week, a 4-ft square was laid-out, and a half-dozen paper cups weighed with beanbags were provided. The object was to clear the cups from within the square. Not everyone finished, so we'll set that up again this week, along with the next challenge for those who are ready.

    Other ideas in the planning stage include driving around a series of different sized cardboard boxes, and laying out a large 'X' on the floor with the goal of parking your robot as close as possible to the center.

    I'd also like to explore effects such as battery voltage level, motor speed, and turning methods. I haven't figured out yet how to build that into an exciting weekly challenge.

    Navigating to a light beacon would be fun also (simulated space docking mission, anyone?).

    I'm interested in hearing what training tools other coaches and mentors are using to build their team's skills.
    FIRST LEGO League Mentor and Referee/Head Referee since 2011.

  • #2
    Re: Off-season training

    My recommendation is to find a way to teach simple Proportional controllers. This is just the "P" in a PID controller.

    Both line following using a light sensor and wall following using an ultrasonic sensor are great ways to apply this technique.

    It sounds like you've done straight line following - try to graduate to an oval track, and see if the robot can follow all the way around.

    For using the ultrasonic sensor to wall-follow, our team used a big card board box set on the ground. Eventually, the kids could get the bot to follow the box all the way around.

    For both line following and wall following, the kids can experiment with either controlling the B and C motors directly using Motor blocks, or they can try passing the correction into the steering plug on a Move block.

    Comment


    • #3
      Re: Off-season training

      Originally posted by Tom Mosher View Post
      I'm interested in hearing what training tools other coaches and mentors are using to build their team's skills.
      During the summer, I tried to link math and geometry concepts to robot movements:
      1. How many motor degrees to an inch
      2. How many motor degrees to a turn (single motor) one degree angle turn
      3. How many motor degrees to a spin (wheels moving in opposite direction) one degree angle

      From that, they created myblocks that use inches instead of motor degrees. Also myblocks for turning using angle instead of motor degrees.

      To make it challenging, design a problem and split them into groups and see who can come up with the answer first (to get the snicker bars). Have each group explain their solution to the others.

      On the mechanical side, matching motors is a VERY important off-season activity. If you would run each motor for 5 seconds and record the number of degrees, you'll be surprised to find a huge disparity between motors. Pairing the ones that are close helps you build a more reliable robot.

      Another activity is to make the robot's arm with gears to create side-to-side as well as up-and-down motion.

      Another one is the build a fast robot (with gears).

      I agree with Tim, about the P controller. Our team did that two summers ago and were surprised at the complete elimination of the zig-zag movement. It was slower, but extremely precise. See it here. All but one of the missions used Proportional line follower; the final mission shows how precise it is.
      Tony A.
      Los Angeles FLL Region OP
      www.la-fll.org
      https://www.facebook.com/LosAngelesFLL/

      Comment


      • #4
        Re: Off-season training

        Originally posted by LOLComets View Post
        During the summer, I tried to link math and geometry concepts to robot movements:
        1. How many motor degrees to an inch
        2. How many motor degrees to a turn (single motor) one degree angle turn
        3. How many motor degrees to a spin (wheels moving in opposite direction) one degree angle

        From that, they created myblocks that use inches instead of motor degrees. Also myblocks for turning using angle instead of motor degrees.
        ...
        .
        Those are great suggested activities also. My only quibble would be that you should calculate how many motor degrees to a centimeter, instead of to an inch. Dean would suggest doing calculations and measurements in millimeters. Using millimeters actually works well for doing the math, since LEGO kindly provides the diameter of wheels in millimeters on the sides of tires.

        For our teams, I've had the kids do the math to calculate expected values for the motor degrees to centimeter ratio, and then run some experiments to see how well things work in reality. Doing the experiments first, and determining the ratio via measurement is certainly a valid approach also.

        If you want examples of the math and measurements involved, look at this excerpt of "Classroom Activities for the Busy Teacher: NXT".

        For translating the math to NXT-G, take a look at the blog entry we did last year, or at Dave Parker's wonderful nxtprograms.com site.

        Comment


        • #5
          Re: Off-season training

          Thanks, those are great suggestions.

          School is closed today due to snow, so no robots for me. Sadness...

          During the line tracking exercise, one of the stronger programmers added a second light sensor, and worked on controlling the steering input by stradding the line and calculating the difference between the sensor readings (the P part of the PID controller). We made one observation I thought was pretty interesting, which relates to another other thread about light sensor calibration.

          With two sensors, when driving across the white tile floor after losing track of the line, the robot entered a persistent gradual right turn. He was puzzled, so started looking into the readings from each light sensor. Because the two sensors share one set of calibration data, the slight difference in light sensor readings was causing a small steering input.

          That's another deposit in that student's bank of experience.
          FIRST LEGO League Mentor and Referee/Head Referee since 2011.

          Comment


          • #6
            Re: Off-season training

            Our team worked on two things between the seasons.
            first task was to develop a strategy how to obtain different movements from one motor (up and down, tip over and turning at differnt places of the robot.
            second task was the aforementioned PID controller, although the kids got the math for it right after the season. But the description of the PID quoted in this thread was very helpful for the kids to experiment with the PID
            The result can be watched http://www.youtube.com/watch?v=OR_Io...ature=youtu.be
            The kids won the champion at the regional (with first place in design).

            Comment


            • #7
              Re: Off-season training

              Originally posted by Tom Mosher View Post
              Thanks, those are great suggestions.

              School is closed today due to snow, so no robots for me. Sadness...

              During the line tracking exercise, one of the stronger programmers added a second light sensor, and worked on controlling the steering input by stradding the line and calculating the difference between the sensor readings (the P part of the PID controller). We made one observation I thought was pretty interesting, which relates to another other thread about light sensor calibration.

              With two sensors, when driving across the white tile floor after losing track of the line, the robot entered a persistent gradual right turn. He was puzzled, so started looking into the readings from each light sensor. Because the two sensors share one set of calibration data, the slight difference in light sensor readings was causing a small steering input.

              That's another deposit in that student's bank of experience.
              Surprising to many, a two sensor proportional tracker may perform much worse than a single sensor tracker. The deadband introduced when the sensors straddle the line (or are both on the line) results in larger (though) less frequent) oscillation than you get with a single sensor that always stays on the edge and is always is under control. This isn't much of a problem if you only use proportional gain. Things get ugly if you start using integral gain which hates deadband. Closed loop feedback control really needs each correction to cause a predictable change in feedback. Seeing the effects of deadband could be an interesting exercise. Build a robot with two sensors that are easy to reposition. Adjust the spacing of the sensors and record the robot behavior (video or maybe a pen mark on paper).

              Comment


              • #8
                Re: Off-season training

                Originally posted by timdavid View Post
                Those are great suggested activities also. My only quibble would be that you should calculate how many motor degrees to a centimeter, instead of to an inch. Dean would suggest doing calculations and measurements in millimeters. Using millimeters actually works well for doing the math, since LEGO kindly provides the diameter of wheels in millimeters on the sides of tires.
                Millimeters is a good choice because the tire diameters are printed in millimeters on the side wall, but I started with millimeters for the girls because prior to version 2 of the software you couldn't pass real numbers across a wire. Entering tire diameter to the nearest centimeter or inch doesn't give good results. We could create an artificial unit, such as tenths or hundredths of an inch, but millimeters are widely used and really fit the bill.

                I suggest finding the relationship between distance and duration through experimentation. When the kids divide distance by tire diameter and come up with the same value regardless of tire size it is a real "Aha!" moment. A good way to find out who has mechanical aptitude is look for the first kid who can predict how far the next set of wheels will make the robot travel.

                My boys originally thought that tire width was going to affect the speed the robot travels. Funny that the girls never had a tire width bias. Less emotional attachment to racing cars perhaps? Anyway, having two robots running tests at the same time is informative. Which robot will go faster, the one with the short wide wheels (that looks like a drag racer), or the one with the tall narrow wheels?

                For turning we used a one wheeled robot that turned about a central pivot. I printed out a clock face and we did turns over that. I designed the robot so it was easy to adjust the distance between the pivot and the wheel to simulate wide or narrow robots. At the end of the season the boys still used clock references to specify angles "Then we have to do a two o'clock turn to the left, or maybe two thirty."

                Comment


                • #9
                  Re: Off-season training

                  Originally posted by Dean Hystad View Post
                  ...
                  For turning we used a one wheeled robot that turned about a central pivot. I printed out a clock face and we did turns over that. I designed the robot so it was easy to adjust the distance between the pivot and the wheel to simulate wide or narrow robots.
                  Dean, I can't visualize what you mean by a "one-wheeled robot". Would you describe it further, or post a picture? Thanks.

                  Comment


                  • #10
                    Re: Off-season training

                    Originally posted by timdavid View Post
                    For translating the math to NXT-G, take a look at the blog entry we did last year....
                    Thanks, that's a great simple presentation of both math and MyBlocks. I've forwarded it to the coach for their next session.

                    Personally, I keep forgetting that one can provide input to a MyBlock by entering a constant, rather than always driving it with a "variable" block.
                    FIRST LEGO League Mentor and Referee/Head Referee since 2011.

                    Comment


                    • #11
                      Re: Off-season training

                      Originally posted by timdavid View Post
                      Dean, I can't visualize what you mean by a "one-wheeled robot". Would you describe it further, or post a picture? Thanks.
                      I am picturing a 'clock' hand. Fixed pivot point in the middle, with a motor and wheel mounted on the 'hand' (or a car axle, but only half). If NXT also mounted on, it could do multiple 360's without tangled wires.


                      Our team has gone anti-bot for some change of pace, choosing to work on a ferris wheel of sorts that has a 'loader' to put in more supplies. More assembly-line-machine contraption than roaming robot.

                      There are many more great 'goals' to use Mindstorms for than what next September may bring in the form of deliver/fetch missions. And building any of those examples can only make for better engineers and problem solvers come FLL time again. And in my opinion can guard against burnout.

                      Comment


                      • #12
                        Re: Off-season training

                        Originally posted by dna1990 View Post
                        I am picturing a 'clock' hand. Fixed pivot point in the middle, with a motor and wheel mounted on the 'hand' (or a car axle, but only half). If NXT also mounted on, it could do multiple 360's without tangled wires.
                        Pretty good description. I wanted something quick to build, easy to adjust and didn't use many parts.

                        Comment


                        • #13
                          Re: Off-season training

                          Originally posted by dna1990 View Post
                          There are many more great 'goals' to use Mindstorms for than what next September may bring in the form of deliver/fetch missions. And building any of those examples can only make for better engineers and problem solvers come FLL time again. And in my opinion can guard against burnout.
                          Good point, thanks for the reminder about the bigger picture.
                          FIRST LEGO League Mentor and Referee/Head Referee since 2011.

                          Comment


                          • #14
                            Re: Off-season training

                            The team I coached for the Food Factor season (8 6th grade boys) was comprised of kids from 2 different teams from the body forward season. Last summer, we sat down and discussed issues that both teams had with their robots. The number one item was accurately finding a line followed by inaccuracies after several moves and turns.

                            The team wrote a program using NXT education version 2.0 that used the data collection block to collect light sensor data as the robot traversed a ripply mat. They then uploaded the data to a computer and used Microsoft Excel to view the data. They could then see what the robot "sees" when it encounters a ripple vs a true black line (notice how the black lines are always sandwiched between white areas). They also tested with different light sensor to mat heights and different lighting conditions. They then came up with an "optimal" height to mount the sensors, light shield, and a my-block which results in repeatably finding a line under varying light and mat ripple conditions. (No sensor calibration required)

                            Next, they wrote a my block to align the robot perpendicular to a line using two light sensors using the previously mentioned line finding my block to first stop on black.

                            They used these my blocks this season to have the robot either align itself to a line on the table or to a wall after making 4 or 5 moves. This also allowed them to have no accurate alignment of the robot in base. The run that was the least reliable was the first one where they did actually align the robot in base instead of using features on the mat or table to align the robot outside base.

                            Comment


                            • #15
                              Oh happy day!

                              My shipment of new parts has arrived! Now I have two light sensors, and a battery pack. There's no line I cannot orient myself to orthogonally now!
                              Last edited by Tom Mosher; 03-09-2012, 03:19 AM. Reason: (post-midnight typing anomaly)
                              FIRST LEGO League Mentor and Referee/Head Referee since 2011.

                              Comment

                              Working...
                              X