Our project hit all of our design criteria that we set out to accomplish. We successfully threw a ping pong ball into a moving cup, our AR tag tracking algorithm worked for many different patterns and our gantry reliably moved the cup in a periodic motion. The integrated throwing motion has 80+% success rate to be within the moving cup!
At the onset of our project, we encountered significant challenges in accurately determining the throwing motion and calculating the joint velocities. One of the main hurdles was formulating an initial throwing pose that ensured each joint attained the requisite velocity at the precise moment/position of release.
Additionally, we faced the complex task of time-normalizing cup pattern data, a necessity due to the irregular intervals of our AR tag history. This step was crucial to maintain the consistency and accuracy of our analyses.
Another technical difficulty we grappled with was AR tag detection. The AR tag exhibited erratic behavior, often abruptly shifting its orientation to an incorrect alignment. This led to significant distortions in our path data logging, resulting in inaccurate and undesirable outcomes.
Sawyer missing cup because of error of the AR tag detection.
Try task based control instead of commanding joints directly , this should reduce the amount of calibrations needed
Apply more oil on the gantry for smoother motion.
Try to use nonlinear optimizer to use all 7 joints for different throwing motion, e.g. side throws, which would allow for throwing in a cup in different orientation other than vertical.
Use other robotics platform,e,g,another Sawyer or turtle-bots, to perform the movement of the cup