Robust Formation Control for Cooperative Underactuated Quadrotors via Reinforcement Learning

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Cooperative control of multi-agent systems, particularly quadrotors, is complex due to underactuation and unknown dynamics.
• Achieving robust formation control under external disturbances and system uncertainties remains a significant challenge.

Purpose of the Study:

• To develop a model-free robust formation control strategy for cooperative underactuated quadrotors.
• To address unknown nonlinear dynamics and external disturbances in multi-quadrotor systems.

Main Methods:

• A hierarchical control scheme integrating reinforcement learning theory.
• A distributed observer for leader state estimation.
• Independent position and attitude controllers for formation and rotational motion.

Main Results:

• A novel robust controller was designed without prior knowledge of individual quadrotor dynamics.
• The proposed method effectively estimates leader states and achieves desired formations.
• Simulations demonstrated the effectiveness of the model-free approach in multi-quadrotor systems.

Conclusions:

• The model-free robust formation control method is effective for cooperative underactuated quadrotors.
• Reinforcement learning and hierarchical control offer a viable solution for complex robotic systems with unknown dynamics.
• The approach ensures stable formation and attitude control despite uncertainties and disturbances.