Video
We have developed a new paddle for the juggler called the Cloverleaf. It enables the Blind Juggler to simultaneously juggle multiple balls. We can initialize (so far manually) the balls to bounce in different modes to create new interesting juggling patterns.
Higher Quality Video Available Here
Our prototype starts to juggle a ball and demonstrates feed-forward apex height transitions. Since the robot cannot sense the ball at any given time, the apex height is changed quasi-statically, i.e. by just adjusting the height a little bit each hit. In this video, the height increment between hits when transitioning is 0.08m, which is quite substantial. The large motion of the paddle (amplitude ca. 0.16m) allows for larger uncertainty in the impact time.
Higher Quality Video Available Here
Philipp moves the robot while it juggles a ball at 0.8m. The system demonstrates robustness in the horizontal directions.
Higher Quality Video Available Here
The robot can juggle up to heights of 2.1m. This is achieved with the new paddle with a smaller curvature. The analysis shows that the smaller the curvature, the higher the robot can juggle in a stable way. However, we are reaching the physical limits in apex heights as the stroke of the robot is limited.
Higher Quality Video Available Here
Our prototype juggles various balls to illustrate the effect of ball roundness on the juggling performance. Rounder balls introduce less noise in the horizontal speeds. It is interesting to see that even high-quality table-tennis balls introduce quite some noise. Initially, we optimized the design for a Super Ball, since they have excellent vertical coefficients of restitution. It turned out that Super Balls are too heavy (~44 grams) for the robot, i.e. the plate and linear guide are not stiff enough.
Experimenting with different balls, the precision ball made for valves introduces the least noise at impact. The ball's roundness is very precise (up to a few micrometers). This specific ball is made of PA-6.6, has a diameter of 0.012m and costs about 1 dollar. The vertical coefficient of restitution is about 0.81, but decreases with larger apex heights. We also thought about using balls from ball bearings, but they would damage the robot's aluminium paddle. Note that the motion is the same for all the balls in the video; this
Higher Quality Video Available Here
The new version of the juggler performing it's "first bounces". The pendulum trajectory control is just a very basic PD controller. With this simple controller and the present motors, we cannot reach the 30° amplitude the system was designed for. The motors are not powerful enough to force the pendulum to swing in the proper motion. The motors have to brake the pendulum as the system would - very similar to a child on a swing - quickly swing up to the physical limits of the pendulum. Read more about this new prototype in the V2.0 section.
Higher Quality Video Available Here
