Composite-Learning-Based Adaptive Neural Control for Dual-Arm Robots With Relative Motion.

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Dual-arm robot systems require precise control for complex bimanual tasks.
• Modeling uncertainties and relative motions pose significant challenges in traditional robotic control.
• Existing methods often struggle with dynamic uncertainties and convergence issues.

Purpose of the Study:

• To develop an adaptive control method for dual-arm robot systems performing bimanual tasks.
• To address challenges posed by modeling uncertainties and relative motions between robotic arms and a grasped object.
• To enhance trajectory tracking and contact force control capabilities.

Main Methods:

• A command filtered control technique is applied for trajectory and force control.
• A radial basis function neural network (RBFNN) is utilized to handle dynamic uncertainties.
• A novel composite learning law integrates historic data for improved neural network weight updates.
• Partial persistent excitation ensures estimation convergence, with stability analyzed via Lyapunov theorem.

Main Results:

• The proposed adaptive control method effectively manages modeling uncertainties in dual-arm robot systems.
• The command filtered control technique achieves accurate trajectory tracking and contact force control.
• The novel composite learning law with RBFNN demonstrates improved convergence and estimation accuracy.
• Numerical simulations validate the robustness and effectiveness of the control and learning algorithms.

Conclusions:

• The developed adaptive control strategy offers a robust solution for bimanual tasks in uncertain environments.
• The integration of RBFNN and composite learning significantly enhances the performance of dual-arm robots.
• The method provides a stable and reliable approach for complex robotic manipulations.