Adaptive learning control speeds up robot arm precision tracking
Differentiable Iterative Learning Control for Black-Box Position-Commanded Manipulators
Robotics
Summary
High-precision movements are hard for robot arms controlled by preset commands because their inner controls can’t be changed. The authors created a method that adjusts the robot’s control settings automatically between tries using a clever math technique. This method learns from each attempt without needing the robot to run more times. They tested it on a 6-joint robot arm and found it quickly cut errors much more than traditional fixed methods. This helps robots repeat tasks more accurately with less manual tuning.
What this means in practice
- •For industrial robot operators: Improve repetitive task accuracy in robotic arms without accessing low-level controls or lengthy manual tuning.
- •For robotics software developers: Implement adaptive learning modules that update control parameters offline to speed up robot precision improvements.
Authors
Mohammed Al-Ghaili, Yakoop Qasim, Hassan Almubarak
Abstract
High-accuracy repetitive tracking is challenging for position-commanded manipulators whose low-level controllers are inaccessible, while constrained model-based iterative learning control (ILC) can require substantial manual parameter tuning. We present Differentiable Iterative Learning Control (Diff-ILC), an online bilevel framework that adapts the parameters of a constrained ILC update between physical robot executions. A closed-loop differential model is identified once from paired smooth perturbations and subsequently kept fixed. At each trial, a constrained quadratic program computes a feasible command correction, while a real-anchored next-trial prediction defines a task-level objective. Hypergradients are obtained by implicitly differentiating the locally active KKT system, enabling multiple computational parameter updates without additional robot executions. Experiments on a 6-DoF DOBOT CR3 across unseen planar and three-dimensional trajectories show substantially faster physical-trial convergence than fixed-parameter model-based ILC. Across the four primary trajectories, after only two trial-to-trial updates, Diff-ILC achieves 78.8% and 76.4% lower mean joint-vector and Cartesian RMSE, respectively, than fixed-parameter model-based ILC.