Human-Assisted Reinforcement Learning Enhances Robotic Disassembly Efficiency

Strategic Insights -

Researchers at Mondragon University have successfully integrated human hints into reinforcement learning frameworks, significantly improving the adaptability and performance of disassembly tasks. The study reveals that manipulation generalization capabilities are enhanced when policies receive human hints regarding where to focus their actions. Employing force overlay types, such as Lissajous curves and spiral shapes, under different parameters reduces contact forces caused by friction during disassembly. This innovative approach leads to substantial improvements in efficiency and flexibility, making it highly applicable to industrial settings.

Key Takeaways:

  • The escalating rates of material consumption and energy usage pose significant environmental challenges, and circular economy principles, remanufacturing, and disassembly strategies emerge as potential solutions.
  • Disassembly processes face a critical hurdle due to the inherent variability of products, and human-assisted reinforcement learning approaches can effectively manage this variability and improve adaptability and performance.
  • The application of reinforcement learning with robot-human interactions addresses variability in disassembly, particularly in contact-rich tasks, such as occluded contact-rich manipulations.
  • Human-assisted reinforcement learning enhances manipulation generalization capabilities by providing human hints regarding where to focus actions, leading to up to 10% reduction in mean force magnitude and 55% decrease in contact forces exceeding the established threshold.
  • Employing force overlay types, such as Lissajous curves and spiral shapes, under different parameters significantly reduces contact forces caused by friction during disassembly.
  • This research demonstrates substantial improvements in efficiency and flexibility, making it highly applicable to industrial settings.
  • The paper's value lies in its innovative integration of human hints into reinforcement learning frameworks, addressing the limitations of previous methods and showing substantial improvements in efficiency and flexibility.

Statistics:

  • Up to 10% reduction in mean force magnitude
  • 55% decrease in contact forces exceeding the established threshold
  • 28% reduction in task completion time compared to non-overlay execution
  • Human-assisted reinforcement learning shows substantial improvements in efficiency and flexibility

Sources:

  • "Human-assisted Reinforcement Learning and Dynamic Force Patterns In Contact-rich Manipulation for Robotic Disassembly" (Industrial Robot: the international journal of robotics research and application, 2025)
  • Mondragon University, Dept. of Electronics and Computer Sciences, Arrasate Mondragon 20500, Spain
  • Nestor Arana-Arexolaleiba, Mondragon University, Dept. of Electronics and Computer Sciences, Arrasate Mondragon 20500, Spain
  • Antonio Serrano, Mondragon University, Dept. of Electronics and Computer Sciences, Arrasate Mondragon 20500, Spain
  • Dimitrios Chrysostomou, Mondragon University, Dept. of Electronics and Computer Sciences, Arrasate Mondragon 20500, Spain