Most Goals Scored in Autonomous?

[apalrd]

Originally Posted by dag0620

...I'm sorry but without the Autonomous In my oppion (I know this is not Dictionary Definition) it is just once big RC car...

We generally write the autonomous code after the robot is built, which leads it to be the last thing, however a teleoperated FIRST robot can have as much automation in Teleop as Autonomous mode. For the Killer Bees, we generally have closed-loop feedback on many things, especially arms, kickers, crab pods, etc. with possibly more automation that just that.

For example, this year's robot had a three-speed DeWalt transmission on the arm, and a servo to shift it from first to second gear to deploy faster. At the operator's request to deploy, it would move over until the shift point, shift into first, and use a separate set of gains with the different gear ratio. As the operator grabbed the pole, they would press a button to lift the robot, and it would lift. If the latch was not latched, they could deploy again, and it would stay in first so as not to risk missing a shift under load. All of this was automated, as the operator has a multi-position switch with each mode (start, vertical, deploy, and flip).

Another, possibly better example: 33 in 2007. 3-axis arm, with two joints on the same plane. The operator had a box for each position as well as a joystick. They would press the, for example, Floor button and it would recall the position and go to it. This is simple. However, as the operator pressed the, for example, Score Mid button, the robot would then:
1. recall the position and set that as setpoint of each joint
2. Calculate the error on the first joint and use the second joint to correct for that error, so the end-effector is at the same height
3. Allow the operator to use the Y axis of their joystick to slide the end-effector in and out at the same height, while changing the shoulder and elbow positions automatically. This allowed some wiggle room for the driver to be closer to or farther from the rack then expected, or there was a robot in their way (just score over them).
4. Calculating the position of the mass (claw) and adjusted the D term of the shoulder based on the horizontal distance from the joint, to prevent it from speeding up and overshooting as the distance (and torque on the joint) decreased.
5. The claw was very wide, to control the tube, so when the claw was pulled into the robot, it would automatically slow down or wait for the claw to rotate to a position that would not hit the ifself before bringing it into the machine.
6. The claw was also asymetrical to allow easier collection from the floor, and the top jaw would close 1/10th of a second before the bottom jaw.

Some more robots. 33 in 2008. This robot had a two-speed AndyMark transmission, geared for 14fps high. It generally looped the track at high speed with the trackball, then slowed down a little to lift it over the overpass, etc. This robot had a semi-automatic transmission. If the driver was in low, flooring it, and the robot was moving at its top speed in low gear, for long enough time, it would assume that the driver is racing and shift up. In a real scenario, the driver could have shifted to low to push an obstacle (defensive robot?) out of the way, then broken through and continued on, and the robot would return to high gear for racing. Several robots have had this, this is the one I know the most about.

Another robot. 33 in 2001. Back in the day of the PBASIC processor with 63 bytes of variable space. The robot had an inclinometer, a weight attached to a potentiometer, and with the push of a button it would very quickly level itself on the ramp. Much faster than an operator could've done it. (the points in that game were based partially on time it took for you to complete the match; you could end the match early to gain bonus points)

Many FIRST robots have advanced past the remote-control car stage. They now use automation to help lower the load on the drivers and operators, so they can focus on playing the game instead of translating commands into what the robot wants. A great thing that many teams tried to do in this game was use the target to align to the goal. Autonomy in teleop, although many of them found that vision processing ate up so much CPU it caused problems with PID controls.



Back to this year's autonomous modes, 33 could kick five balls. We focused on tuning the first three kicks, and by Nationals, at least two of three went in (providing the drivers lined the robot up, something they weren't very good at). The last two were more tricky, because the robot's alignment would change when it went over the bump, but it usually made one of those as well. We had closed-loop control of speed per side and triggers on distance, but did not use the camera to line up any shots.

@GDC: A longer autonomous period could create a greater incentive to write one. 20 or 30 seconds of full autonomy would make me very, very happy.