An adaptive direct data-driven controller for nonlinear system reference tracking

Area of Science:

• System control theory
• Data-driven control methods
• Robotics and automation

Background:

• Reference tracking is crucial in system control.
• Data-driven approaches are gaining prominence for control problems.
• Existing methods may face challenges in complex or unexplored system dynamics.

Purpose of the Study:

• To propose a novel data-enabled predictive controller for enhanced reference tracking.
• To develop a controller that ensures safety, constraint satisfaction, and robustness.
• To evaluate the controller's performance on a simulated unmanned surface vehicle (USV).

Main Methods:

• Development of an Adaptive Trajectory Data-Enabled Predictive Controller (ATDeePC) algorithm.
• Local linear model construction and prediction storage.
• Selection of appropriate linearized model trajectories for Hankel matrix construction based on online and reference trajectories.

Main Results:

• The ATDeePC algorithm demonstrates safe closed-loop operation.
• The controller successfully satisfies operational constraints.
• The controller exhibits robustness and adaptability to new system states.

Conclusions:

• The ATDeePC algorithm offers a robust and adaptive solution for data-driven reference tracking.
• The proposed controller outperforms common methods in simulations for USV applications.
• This approach advances the field of predictive control through data enablement.