Secure Containment Control for Multi-UAV Systems by Fixed-Time Convergent Reinforcement Learning.

Area of Science:

• Robotics and Control Systems
• Cyber-Physical Systems Security
• Artificial Intelligence in Autonomous Systems

Background:

• Autonomous aerial vehicles face containment control challenges due to cyber attacks manipulating control commands.
• Existing control methods may be vulnerable to sophisticated cyber threats, risking containment failure.
• Secure containment of multi-vehicle systems requires robust strategies against adversarial actions.

Purpose of the Study:

• To address the secure containment control problem for multiple autonomous aerial vehicles under cyber attacks.
• To develop a reinforcement learning-based framework for designing secure distributed control policies.
• To enhance the convergence speed and robustness of control systems using fixed-time convergence techniques.

Main Methods:

• Formulated the secure containment control as a zero-sum graphical game, a min-max optimization problem.
• Employed a critic-only neural network (NN) structure within a reinforcement learning (RL) framework to solve Hamilton-Jacobi-Isaacs (HJI) equations.
• Integrated fixed-time convergence and experience replay mechanisms to accelerate RL convergence and relax excitation conditions.

Main Results:

• Achieved optimal distributed secure control policies by solving coupled HJI equations using RL.
• Demonstrated NN convergence and closed-loop stability analysis for the proposed control scheme.
• Obtained optimal feedback control laws for the attitude loop by solving Hamilton-Jacobi-Bellman equations.

Conclusions:

• The proposed fixed-time convergent reinforcement learning method effectively ensures secure containment control for multiple autonomous aerial vehicles.
• Simulation and quadrotor experiments validate the practical effectiveness and robustness of the developed secure control strategy.
• The approach provides a significant advancement in securing autonomous aerial vehicle systems against cyber-physical threats.