Sliding Flexible Prescribed Performance Boundary-Guided Reinforcement Learning Control for Input-Constrained...

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Nonlinear Dynamics

Background:

• Input-constrained nonlinear systems (ICNSs) present significant control challenges.
• Existing prescribed performance control (PPC) methods often require parameter tuning and may compromise initial performance.
• Balancing input constraints with desired performance is crucial for practical applications.

Purpose of the Study:

• To develop a novel sliding flexible prescribed performance boundary-guided reinforcement learning (SFPPB-RL) control approach for ICNSs.
• To overcome limitations of existing PPC methods regarding parameter tuning and initial error handling.
• To achieve a balance between input safety and control performance.

Main Methods:

• Design of a sliding flexible prescribed performance boundary (PPB) that adaptively adjusts to initial errors and dynamically modifies constraint relaxation.
• Integration of an auxiliary system to manage the coupling between input and performance constraints.
• Combination of an identifier-critic-actor structure-based reinforcement learning (RL) strategy with backstepping techniques.

Main Results:

• The proposed SFPPB-RL approach eliminates the need for repeated parameter debugging.
• It addresses the trade-off between initial transient performance and error requirements.
• The method effectively balances input safety and control performance, minimizing a cost function.

Conclusions:

• The developed SFPPB-RL control algorithm ensures input safety and prescribed performance indicators for ICNSs.
• Simulation results validate the effectiveness of the proposed approach.
• This method offers an adaptive and robust solution for complex control problems.