Designing modular robotic architecture for e-commerce bin picking task fulfillment

This paper presents a high level architecture for a robot, designed for item picking task for E-commerce fulfillment center. The architecture was designed to be modular and has features that allow the robot to operate in an E-commerce warehouse environment, where bin content changes over time. Our system was built upon the Robot Operating System (ROS), with the benefits of abstracting low level machine implementations such as joint control and communication, enabling the user to focus on higher level tasks. We built a software interface that includes a robot program manager node supported by core libraries. We tested the system on a mock copy of warehouse shelves and simulated camera data input. We used the Amazon Robotics Challenge (ARC) 2017 as a reference to design the over-all behavior of the system as well as to evaluate its performance. The robot's item picking performance was measured based on execution timing and the success of each pick.