Robot system learns efficient dual-arm brick assembly tasks
BrickCraft-Duo: Efficient Dual-Arm Skill Learning and Refinement for Compositional Long-Horizon Assembly
Robotics
Summary
Assembling interlocking bricks with robots is tricky because it involves precise fitting, multiple steps, and teamwork between two arms. The authors created BrickCraft-Duo, a robot system that learns reusable skills for both single-arm and two-arm actions from examples. This system smartly shares skills between symmetric arms and plans the order of assembly steps by checking stability. It can also improve itself by asking humans for corrections during use. The system successfully completes long assembly tasks with high accuracy on real bricks.
What this means in practice
- •For manufacturing engineers: Automate complex assembly processes involving precise multi-step part insertion with dual-arm robots using learned reusable skills.
- •For robotics software developers: Develop modular robotic systems that combine single-arm and dual-arm skills to plan and execute long, stability-sensitive assembly sequences.
Authors
Jichuan Yu, Zhenyu Xiao, Ze Wang, Ruixuan Liu, Changliu Liu, Chuxiong Hu
Abstract
Interlocking brick assembly provides a representative testbed for evaluating real-world robotic manipulation capabilities, where diverse structural designs, complex inter-step dependencies, intricate mechanical interactions and tight insertion tolerances pose substantial challenges. We present BrickCraft-Duo, a modular framework for long-horizon dual-arm collaborative assembly of interlocking bricks through data-efficient skill learning and composition. BrickCraft-Duo learns reusable single- and dual-arm assembly skills from diverse demonstrations, with bilateral symmetry alignment facilitating skill sharing across symmetric arms and assembly--support role assignments. Guided by stability-aware assembly reasoning, BrickCraft-Duo composes heterogeneous skills to achieve autonomous long-horizon execution, and further integrates human-in-the-loop correction for targeted skill refinement. The resulting system achieves long-horizon success rates of at least 60% and step-level completion rates of at least 95% across five real-world assembly tasks involving partially supported configurations, with horizons of up to nine steps. Project website: https://jichuan-yu.github.io/BrickCraft-Duo.