Actor-critic learning based coordinated control for a dual-arm robot with prescribed performance and unknown...

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Dual-arm robots (DARs) face complex control challenges, including unknown input backlash-like hysteresis and system uncertainties.
• Ensuring precise tracking performance is critical for advanced robotic applications.

Purpose of the Study:

• To develop an adaptive coordinated control strategy for DARs that addresses unknown input backlash-like hysteresis.
• To impose prescribed performance constraints on the DAR system for enhanced tracking accuracy.
• To improve the self-learning capabilities of the control system.

Main Methods:

• An adaptive coordinated control approach utilizing an actor-critic (AC) design is proposed.
• An actor-critic learning (ACL) algorithm is introduced to handle system uncertainties and hysteresis.
• Neural networks (NNs) are employed within the actor and critic networks to approximate unknown system dynamics.
• Lyapunov direct method is used to prove system stability.

Main Results:

• The proposed AC-based coordinated control effectively manages unknown input backlash-like hysteresis in DARs.
• Prescribed performance is successfully imposed, guaranteeing desired tracking performance.
• The ACL algorithm demonstrates improved self-learning ability in handling system uncertainties.
• Numerical simulations validate the effectiveness and stability of the proposed control strategy.

Conclusions:

• The developed adaptive coordinated control with AC design offers a robust solution for dual-arm robot tracking problems with hysteresis.
• The integration of AC learning and prescribed performance significantly enhances control accuracy and adaptability.
• The findings provide a foundation for more sophisticated control of robotic systems in complex environments.