These videos represent the body of research that I contributed to at AUTOlab in University of California Berkeley. They utilize robotic learning and human demonstrations to exhibit robust behavior in initial state variance
Description of training the robot to grasp in a cluttered environment using camera feed
The robot learns a policy to navigate through the objects to reach the yellow circle
The robot learns to separate one object from the rest
The robot learns a policy to perform overhead grasps for objects in arbitrary positons
The robot learns a policy to separate the objects with a paddle, identify, grasp, and place the object in the kit