Improving Robotic Lego Assembly with Vibro-Tactile Feedback

Abstract

Robotic manipulation is an important area of research to improve the level of efficiency and autonomy in manufacturing processes.
Due to the high precision and repeatability of industrial robot arms, robotic manufacturing tasks
are dominated by simple pick, place, and peg insertion actions performed in a highly structured environment.
Lego blocks are an excellent analog for studying robotic assembly as they are well structured in a grid
along with only needing simple rotations and peg insertions to manipulate. They are also cheap, easily segmented with cameras, and robust to wear and tear.

In this paper, we explore both efficiency and autonomy improvements for Lego building tasks.
We introduce "BrickPick," an actuated end effector capable of carrying multiple Lego blocks at once. The BrickPick end effector improves assembly
efficiency by increasing the throughput of manipulated blocks per motion. We also introduce the "Vibro-Tactile Toolbox," a ROS software package used to integrate
vibro-tactile feedback into manipulation skills. The toolbox provides a framework for utilizing sensor data from contact microphones,
force-torque sensors, and cameras to inform the robot when actions should be terminated and what their outcomes have been.
The Vibro-Tactile Toolbox improves assembly autonomy by providing an online self-supervisor to handle failed actions and facilitate the collection of
labelled trial data.