Spring Validation Demo (May 2019)
We randomly place 5 toys around starting location and turn on Cubi. Cubi searches for a toy autonomously and moves towards it while keeping track of path it moves. It slowly approaches the toy and then lowers the gripper when it is near the toy. It pushes the toy a bit further to increase the chances of grasping. Then it uses the caging strategy and picks the toy. It starts to follow back its path back to the start position with the help of wheel odometry. At the destination, it re-orients itself with the designated drop-off location (box) and drops the toy. This procedure is repeated until all toys are picked up.
At the end of the demo we tried to put more random objects like phone, mouse, markers and CuBi was able to generalize well to unseen examples.
Exploration Demo
Fall Validation Demo (December 2019) (5x Speed)