No announcement yet.

Precision robot placement

  • Filter
  • Time
  • Show
Clear All
new posts

  • Precision robot placement

    Well, the season is over for us, but I am still curious. How did you folks resolve the issue of initial robot placement? All of the autonomous programs that I saw, including ours, had an arm that lowered with color sensor(s) on the end to read the color of the jewels. Then, different methods were used to knock off the correct ball. Some twisted, some drove forward or backward, and some had rotating arms.

    The biggest problem for us was not the programming. It was placement. Our arm was on the left side of the robot. If the robot was placed too far to the left, the arm hit the wall on the way down. If it was placed too far to the right, the color sensor wasn't close enough to read the color. If it was placed too far forward or backward, it knocked off a ball when the arm lowered. The margin for error for the placement was fairly small. In addition, the balancing board has inherent variance in it due to the way it is constructed.

    Using the rivets as a guide gets it in the general area. We tried several different things. At first, we lowered the arm to check to make sure it was properly located. At an early competition, we learned that was not allowed. We considered range sensors looking at the walls. The budget was tight and these seemed cost prohibitive. Also, there is variance on the walls.

    We ended up using Vuforia to determine the precise location of the robot in relation to the pictograph. The game manual specifically states that it is to be placed 3 inches to the left of center of the jewels. We calibrated to this and it worked wonderfully in practice. At the beginning of the autonomous period, the robot would judge the location of the robot in relation to the pictograph and either move forward or backward to get into the proper spot. The color sensor worked almost every time. It did knock off the ball once out of dozens of attempts and we reduced the increment of the initial moves. After that, it never knocked off the ball when lowering the arm. We used two color sensors and it worked properly every time.

    However, in the two competitions we went to after getting this worked out, the pictographs are not precisely located 3 inches offset. Both times we appealed and pointed out where they were to be located. Both times, we were told there is a 1 inch allowance. At this last one, we were allowed to measure them. They ranged from 2 1/2 inches to 3 1/2 inches. That is much more than the room for error in the location of the robot. With the pictographs placed incorrectly, we could not find another way for precision location of the robot. We did try reading the jewels in Vuforia, but there are not enough identifying features on the balls.

    If the room for error in placing the robot is under 1/2 inch from the perfect spot in order to work and the allowance given in rule G25 is 1 inch, how in the world did anybody achieve precision placement in order to consistently knock off the proper jewel?

  • #2
    We eyeball it. We have a jewel arm that slides out from underneath the robot linearly. The front of it sticks a little out from the robot so when placing we eye it to see if the arm will come out to where the white tape is on the jewel platform. For forward/backward placement, we align in the middle of the balancing stone. For determining the jewel colors, we use a Pixy Camera sensor which has proved extremely reliable. The jewel arm extends to between the jewels and then the robot drives left or right to knock the correct jewel off. So the placement has been fairly easy. The accuracy of it all has been very good this season outside of some growing pains our first competition.


    • #3
      That sounds like a good design! A linear extension would negate the issue of hitting the wall on the way down. We eyeball it as well in the manual placement. We use two color sensors, one facing forward and one facing backward to reduce the risk of the sensor being placed by one of the holes. The downside is that the room for error is reduced. That is where the precision movement came into play. Often, the robot would not need to adjust. It just gave that extra cushion in case the board is shifted a bit forward or backward.


      • #4
        We use one color sensor, so it isn't as bulky, and we have that arm like a lot of teams do. We only have one arm though, which makes some of our starting positions have the robot be backwards, not ideal. Some teams have 2 arms with what looks like one color sensor. One of the higher performing teams also uses several sensors including the NXT ultrasonic sensor for key placement to square into the cryptobox more accurately. My team's robot has not been totally consistent in scoring autonomous, but generally has gotten some points, we aim for the jewel, the key, and parking in auto. (and early on in practice, we learned it's best to be too far from the wall since hitting the wall broke the arm motor. Every week something different breaks after heavy use, so we are very happy when things aren't breaking down in an official match, and we ended up in the Winning Alliance at the last competition. The highest performing teams that we've seen practice driving a lot and that seems to help, too. They've got the ciphers memorized and are fast.)
        Last edited by RoboticsFun; 02-15-2018, 10:46 AM.