Grasping Strategies for Picking Items in an Online Shopping Warehouse.

The purpose of this study is to investigate the most effective methodologies for the grasping of items in an environment where success, robustness and time of the algorithmic computation and its implementation are key constraints. The study originates from the Amazon Robotics Challenge 2017 (ARC’17) which aims to automate the picking process in online shopping warehouses where the robot has to deal with real world problems of restricted visibility and accessibility. A two-finger and a vacuum grippers were chosen for their practicality and ubiquity in industry. The proposed solution to grasping was retrieval of a final position and orientation of the end effector using an Xbox 360 Kinect sensor information of the object. Antipodal Grasp Identification and Learning (AGILE) and Height Accumulated Features (HAF) feature based methods were chosen for implementation on the two finger gripper due to their ease of applicability, same type of input, and reportedly high success rate. A comparison of these methods was done.