Overview

Software flow

Our software had 5 main steps, listed below

1. Picking up the ball

2. Moving to a viewing angle

3. Extracting cup positions (x, y, z) from image and selecting cup (see subpage Computer Vision)

4. Calculating throw pose via optimization (see subpage Initial Throw Position)

5. Throwing the ball (see subpage Throwing)

Steps 3, 4, 5 are linked in separate sub-pages.

*** Please note that in-line code may have adjustments from the code in the repository due to readability and further development of our codebase.

Step 1: Picking up the ball

To pick up the ball, we decided that we would hold the ball in front of the grippers and manually execute a close; this was executed by this code flow;

gripper = robot_gripper.Gripper('right_gripper')

# ...

c = input('to close')

gripper.open()

Step 2: Moving to a viewing angle

We similarly waited for a manual input to confirm the ball was in the grippers before executing a shell command to move the robot to a good viewing position. We chose to use an intera_examples script called go_to_joint_angles.py to do this quickly.

import subprocess

# ...

def go_to_joint_angles(theta0, theta1, theta3, theta5):

subprocess.run(f"rosrun intera_examples go_to_joint_angles.py -q {theta0} {theta1} 0.0 {theta3} 0.0 {theta5} 1.57".split())

rospy.sleep(0.5) # for safety

# ...

_ = input('find cup')

# Vision code covered in Step 3

go_to_joint_angles(-0.5, -0.785, 0, 0)

Steps 3, 4, 5: See subpages

3: Computer Vision

4: Throw Initial Position

5: Throwing