Waste-Container Lifting Using Residual Reinforcement Learning On Large-Scale Crane with Underactuated Tools

Qi Li, Karsten Berns
Proceedings of the The 39th Canadian Conference on Artificial Intelligence, PMLR 318:1100-1107, 2026.

Abstract

This paper studies the container lifting phase of urban waste-container recycling task with a hydraulic loader crane and an underactuated discharge unit. The task requires accurate hook–ring alignment under tight geometric tolerances while suppressing oscillations of the suspended unit. To address this, we propose a residual reinforcement learning framework that combines a nominal Cartesian controller for trajectory tracking and anti-sway control with a learned residual policy for compensating unmodeled dynamics. The residual policy is trained with PPO. Simulation results show improved tracking accuracy, reduced oscillations, and higher lifting success than the nominal controller alone.

Cite this Paper

BibTeX
@InProceedings{pmlr-v318-li26b, title = {Waste-Container Lifting Using Residual Reinforcement Learning On Large-Scale Crane with Underactuated Tools}, author = {Li, Qi and Berns, Karsten}, booktitle = {Proceedings of the The 39th Canadian Conference on Artificial Intelligence}, pages = {1100--1107}, year = {2026}, editor = {Bouzar-Benlabiod, Lydia and Leung, Carson}, volume = {318}, series = {Proceedings of Machine Learning Research}, month = {25--29 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v318/main/assets/li26b/li26b.pdf}, url = {https://proceedings.mlr.press/v318/li26b.html}, abstract = {This paper studies the container lifting phase of urban waste-container recycling task with a hydraulic loader crane and an underactuated discharge unit. The task requires accurate hook–ring alignment under tight geometric tolerances while suppressing oscillations of the suspended unit. To address this, we propose a residual reinforcement learning framework that combines a nominal Cartesian controller for trajectory tracking and anti-sway control with a learned residual policy for compensating unmodeled dynamics. The residual policy is trained with PPO. Simulation results show improved tracking accuracy, reduced oscillations, and higher lifting success than the nominal controller alone.} }
Endnote
%0 Conference Paper %T Waste-Container Lifting Using Residual Reinforcement Learning On Large-Scale Crane with Underactuated Tools %A Qi Li %A Karsten Berns %B Proceedings of the The 39th Canadian Conference on Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Lydia Bouzar-Benlabiod %E Carson Leung %F pmlr-v318-li26b %I PMLR %P 1100--1107 %U https://proceedings.mlr.press/v318/li26b.html %V 318 %X This paper studies the container lifting phase of urban waste-container recycling task with a hydraulic loader crane and an underactuated discharge unit. The task requires accurate hook–ring alignment under tight geometric tolerances while suppressing oscillations of the suspended unit. To address this, we propose a residual reinforcement learning framework that combines a nominal Cartesian controller for trajectory tracking and anti-sway control with a learned residual policy for compensating unmodeled dynamics. The residual policy is trained with PPO. Simulation results show improved tracking accuracy, reduced oscillations, and higher lifting success than the nominal controller alone.
APA
Li, Q. & Berns, K.. (2026). Waste-Container Lifting Using Residual Reinforcement Learning On Large-Scale Crane with Underactuated Tools. Proceedings of the The 39th Canadian Conference on Artificial Intelligence, in Proceedings of Machine Learning Research 318:1100-1107 Available from https://proceedings.mlr.press/v318/li26b.html.

Related Material