Year: 2025
Type: article
Abstract: The study proposes a hybrid control architecture for industrial robots that incorporates deep neural networks for adaptive tuning in all operating domains through online dynamic compensation. The system synchronizes six-dimensional force/torque sensor data, combined motor current, and terminal acceleration via an Ether-CAT bus to construct a compensation basis that includes 23 dimensions of real-time state observation. Dynamic residual learning is performed using a two-stream network design, with the primary network processing the historical state window and the secondary network parsing joint-space parameters. The network is trained using a move-learning strategy and pre-trained in a high-fidelity simulation environment of the Franka Emika robot.
Author: Purui Liu
Language: English
Subfield: Law
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W4410474707