Adaptive Neural Network Fixed-Time Leader-Follower Consensus for Multiagent Systems With Constraints and Disturbances

Area of Science:

• Control Theory
• Robotics
• Artificial Intelligence

Background:

• Multiagent systems (MASs) face challenges like output constraints, unknown dynamics, and control input issues.
• Achieving consensus in MASs is crucial for coordinated behavior and task execution.
• Existing methods often struggle with fixed-time convergence under these complex conditions.

Purpose of the Study:

• To develop a fixed-time leader-follower consensus control scheme for MASs.
• To address challenges including output constraints, unknown control direction, unknown system dynamics, unknown external disturbance, and dead-zone control input.
• To ensure the consensus is achieved within a predetermined finite time.

Main Methods:

• A fixed-time distributed observer is designed for followers to estimate leader states.
• A modified nonlinear mapping transforms output-constrained systems into unconstrained ones.
• Power integrator technique, radial basis function neural network (RBFNN) approximation, and adaptive methods are employed.
• A virtual control protocol is derived, followed by an actual control protocol using dead-zone parameter bounds.

Main Results:

• The proposed control scheme guarantees fixed-time leader-follower consensus for the studied MAS.
• The control strategy effectively handles unknown control direction, system dynamics, and external disturbances.
• The consensus is achieved within a specified finite time, outperforming traditional methods.

Conclusions:

• The presented control scheme provides an effective solution for fixed-time leader-follower consensus in MASs with significant uncertainties and constraints.
• The method's applicability is demonstrated through its successful implementation on inverted pendulum systems.
• Simulation results confirm the robustness and effectiveness of the proposed control strategy.