DDPG-Based Adaptive Robust Tracking Control for Aerial Manipulators With Decoupling Approach.

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Aerial manipulators offer high agility but face challenges due to complex dynamics and uncertain parameters.
• Implementing practical control for these systems requires addressing coupled dynamic models and system uncertainties.

Purpose of the Study:

• To develop a robust and adaptive control strategy for aerial manipulators.
• To ensure precise trajectory tracking for quadrotors with robotic arms, minimizing interference with drone dynamics.

Main Methods:

• A decoupling approach combining adaptive/robust control techniques with reinforcement learning (Deep Deterministic Policy Gradient - DDPG).
• Utilizing nominal inputs and adaptive algorithms to manage uncertainties in quadrotor, robotic arm, and payload dynamics.
• Employing robust controllers to compensate for residual interactive forces/torques.

Main Results:

• Demonstrated effective control of the robotic arm with minimal impact on quadrotor dynamics.
• Successfully guaranteed system stability and tracking performance under uncertain conditions.
• Validated the control structure and algorithms through numerical simulations and experimental results.

Conclusions:

• The proposed control structure effectively addresses the complexities of aerial manipulator control.
• The integration of adaptive/robust techniques and reinforcement learning provides a robust solution for trajectory tracking.
• The study confirms the efficacy of the developed algorithms for practical implementation.