Reinforcement learning-based trajectory tracking control of live-line working robot coupled with flexible transmission line and wind load vibrations
To address end-effector trajectory tracking degradation caused by floating-base motion under coupled flexible-transmission-line vibration and wind loads, a Soft Actor-Critic (SAC)-based residual trajectory compensation method is proposed. The effects of transmission-line vibration, base motion, and wind loads are represented as a joint-space composite disturbance. Within a nominal inverse-kinematics and PD control framework, the SAC agent uses joint states, trajectory errors, and floating-base states to generate bounded joint-position residuals for online correction of the desired trajectory. The reward function incorporates end-effector error, action magnitude and smoothness, and joint-limit penalties to improve compensation accuracy and safety. Under random fluctuating wind, PD-SAC reduces the mean, root-mean-square, and maximum errors by 88.17%, 87.87%, and 80.65%, respectively, compared with IK-PD. At fixed wind speeds of 5.7, 8.9, and 11.3 m/s, the mean and maximum errors remain below 1.28 and 2.91 mm, respectively. In the field experiment under approximately 12 m/s fluctuating wind, the corresponding reductions are 69.32%, 71.12%, and 72.00%. These results demonstrate that the proposed method effectively compensates for trajectory deviations induced by cable–wind coupled disturbances and improves tracking accuracy and robustness.
Authors
- Wei Jiang (ORCID: https://orcid.org/0000-0002-4870-5005)
- Wei Chen (ORCID: https://orcid.org/0000-0003-1188-4510)
- Yaowei Gan (ORCID: https://orcid.org/0009-0005-4801-6874)
- Lanlan Liu
- Qinghua Liu
Institutions
- Wuhan Textile University (CN)
Publication Details
- Journal
- Journal of Vibration and Control
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1177/10775463261506904
- Primary Topic
- Reinforcement Learning in Robotics
- Type
- article
- Field-Weighted Citation Impact
- 0.00