Composite Neural Learning Fault-Tolerant Control for Underactuated Vehicles With Event-Triggered Input

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence
• Fault-Tolerant Control

Background:

• Underactuated vehicles present significant control challenges due to limited actuators.
• Traditional control methods struggle with system uncertainties and actuator faults.
• Event-triggered control offers reduced communication burden but requires robust fault tolerance.

Purpose of the Study:

• To develop a novel composite neural learning fault-tolerant algorithm for underactuated vehicle path-following.
• To address system uncertainties and unknown actuator faults effectively.
• To reduce communication load via an event-triggered input mechanism.

Main Methods:

• Integration of neural networks (NNs) with dynamic surface control (DSC).
• Utilization of a serial-parallel estimation model (SPEM) for error dynamics.
• Design of four adaptive parameters to handle gain uncertainties and actuator faults.
• Application of the direct Lyapunov theorem for stability analysis.

Main Results:

• The proposed algorithm ensures semiglobal uniformly ultimately bounded (SGUUB) stability.
• The event-triggered mechanism significantly reduces communication overhead.
• Neural networks effectively compensate for system uncertainties and actuator faults.
• Experimental validation confirms the algorithm's superiority.

Conclusions:

• The composite neural learning fault-tolerant algorithm is effective for underactuated vehicle path-following.
• The event-triggered approach enhances control system efficiency.
• The method provides robust stability and fault compensation.