OCTUNE: Optimal Control Tuning Using Real-Time Data with Algorithm and Experimental Results.

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence
• Optimization Algorithms

Background:

• Autonomous robots require precise control tuning for optimal performance, such as trajectory tracking.
• Traditional tuning methods (manual or offline data-driven) are cumbersome and require repetition for system changes or new conditions.
• Existing methods often necessitate prior knowledge of the robot's plant dynamics.

Purpose of the Study:

• To propose a novel algorithm, online optimal control tuning (OCTUNE), for real-time tuning of discrete linear time-invariant (LTI) controllers.
• To enable optimal control tuning without prior knowledge of the plant dynamics.
• To ensure stable and guaranteed convergence during the iterative tuning process.

Main Methods:

• The OCTUNE algorithm utilizes backpropagation optimization to adjust controller parameters.
• Lyapunov stability theory is employed to guarantee convergence and ensure stable iterative tuning using real-time data.
• Validation is performed using realistic simulations of a quadcopter model and on a physical quadcopter platform.

Main Results:

• OCTUNE effectively tunes Proportional-Integral-Derivative (PID) controllers for Unmanned Aerial Vehicles (UAVs) in real-time.
• Guaranteed convergence was demonstrated through simulations and experiments.
• The algorithm proved capable of automatic, real-time PID controller tuning for UAVs.

Conclusions:

• The proposed OCTUNE algorithm offers an effective solution for real-time optimal control tuning in autonomous systems.
• It provides guaranteed convergence and adaptability to changing conditions without requiring plant dynamics knowledge.
• An open-source implementation is available for broader application and adaptation.