As a visiting student researcher at Autonomous Systems lab, I was assigned the task of developing a stacking demo with ABB’s Yumi robot. The robot should be able to recognize a stacking pattern formed using blocks and then stack objects to build a similar structure.

I was awarded the ThinkSwiss Research Scholarship by the Swiss Embassy to conduct this research at ETH Zurich with Dr. Roland Siegwart.

YuMi pick-place simulation in Gazebo and Rviz; Object segmentation output.

The goal of this project is to scan a prebuild object stack and then rebuild it with the YuMi robot. The overall design is divided into two parts. First, a pick-place pipeline to move the robot arms. This was done using MoveIt! library for planning, manipulation and grasping. The second part includes an object detection pipeline to identify location of blocks required to build the stack. The PointCloud library was used to interpret data from ASUS PrimeSense sensor.

Here is a video showing a single pick-place routine in Rviz and Gazebo.