Prescribed Performance Path-Following Control for Rotor-Assisted Vehicles via an Improved Reinforcement Learning...

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence
• Automotive Engineering

Background:

• Rotor-assisted vehicles require advanced control for energy-efficient navigation.
• Existing path-following controllers face challenges with model uncertainties and initial conditions.

Purpose of the Study:

• To develop an adaptive prescribed performance path-following control algorithm for rotor-assisted vehicles.
• To enhance energy-saving cruising missions through reinforcement learning integration.
• To address model uncertainties and external disturbances effectively.

Main Methods:

• Designed a concise prescribed performance control (PPC) algorithm with a shifting function.
• Integrated Backstepping method and optimal control with actor-critic neural networks (AC-NNs) for improved reinforcement learning (RL).
• Utilized Lyapunov theory to guarantee semi-global uniform ultimate bounded (SGUUB) stability.

Main Results:

• The proposed adaptive RL algorithm effectively constrains output errors within defined boundaries.
• Actor-NN generates control policies, while Critic-NN optimizes the cost-to-go function.
• Numerical experiments validated the algorithm's superiority and feasibility.

Conclusions:

• The developed adaptive prescribed performance control algorithm enhances path-following capabilities in rotor-assisted vehicles.
• The integration of RL with AC-NNs provides a robust solution for uncertainties and disturbances.
• The algorithm demonstrates significant potential for energy-saving autonomous missions.