Future Improvements

These are some ideas we had for how to improve the robot to get more continuous bouncing and hopefully get it working more consecutively on a ball.

Sensing

• Increase the reaction time of the robot by improving perception update rate
• Try utilizing only the depth/infrared camera so we would not have to worry about the 30 Hz RGB limit on Realsense.
• Fuse HSV and depth clustering for more reliable centroid measurements.
• Track the ball across frames and smooth out noise to prevent perceived jittery motion.

Planning

• Implement trajectory prediction: right now we're naively shadowing the ball which made it really hard react accordingly to the ball drops
• Time the latency delay between sensing and the forward velocity controller and adjust paddle pop accordingly.
• Currently track velocity realtime -> keep a running state of ball velocity to ensure smoother timing.
• Dynamically update strike height based on current paddle position and the ball velocity.
• Paddle currently remains parallel to the ground -> predict / tune end effector angle to be more adjustable for random behavior (especially seen in the balloon case).