Learning Deformable Object Manipulation Using Task-Level Iterative Learning Control

2026-02-24

AI Summary

Manipulating deformable objects such as ropes presents a significant challenge for robotics due to their infinite degrees of freedom and underactuated dynamics. This work proposes a task-level iterative learning control method that enables a robot to directly learn non-planar “flying knot” tasks on real hardware using only a single human demonstration and a simplified rope model, without requiring large datasets or extensive simulation. The approach formulates a task-space error-driven quadratic program to construct a local inverse model of the robot–rope system and iteratively refines control actions. Experiments demonstrate 100% success within ten trials across seven rope variants differing in material, thickness, and density, with cross-type transfer achieved in just two to five trials, substantially reducing data dependency and simulation costs.

Technology Category

• Intelligent Robots: Manipulation
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Dynamic manipulation of deformable objects is challenging for humans and robots because they have infinite degrees of freedom and exhibit underactuated dynamics. We introduce a Task-Level Iterative Learning Control method for dynamic manipulation of deformable objects. We demonstrate this method on a non-planar rope manipulation task called the flying knot. Using a single human demonstration and a simplified rope model, the method learns directly on hardware without reliance on large amounts of demonstration data or massive amounts of simulation. At each iteration, the algorithm constructs a local inverse model of the robot and rope by solving a quadratic program to propagate task-space errors into action updates. We evaluate performance across 7 different kinds of ropes, including chain, latex surgical tubing, and braided and twisted ropes, ranging in thicknesses of 7--25mm and densities of 0.013--0.5 kg/m. Learning achieves a 100\% success rate within 10 trials on all ropes. Furthermore, the method can successfully transfer between most rope types in approximately 2--5 trials. https://flying-knots.github.io

Problem

• deformable object manipulation
• dynamic manipulation
• infinite degrees of freedom
• underactuated dynamics
• rope manipulation

Innovation

• Iterative Learning Control
• Deformable Object Manipulation
• Task-Level Control
• Dynamic Manipulation
• Hardware Learning