Helicopter trimming and tracking control using direct neural dynamic programming

Area of Science:

• Aerospace Engineering
• Control Theory
• Artificial Intelligence

Background:

• Complex control problems, such as helicopter flight control, require advanced methodologies.
• Existing control systems may struggle with nonlinear dynamics and uncertainties inherent in such systems.

Purpose of the Study:

• To develop and evaluate a neural-network-based approximate dynamic programming control mechanism for complex systems.
• To tailor the direct neural dynamic programming (DNDP) framework for helicopter flight control design and maneuver learning.
• To assess the robustness and applicability of the DNDP approach on a realistic helicopter model.

Main Methods:

• Implementation of a direct neural dynamic programming (DNDP) control framework.
• Integration of a trim network within the neural dynamic programming (NDP) controller for nonlinear system control.
• Extensive simulation studies using the FLYRT nonlinear validated model of an Apache helicopter.
• Evaluation of control system performance under various disturbance conditions to ensure design robustness.

Main Results:

• Successful application of the DNDP methodology to a complex, continuous state, multiple-input multiple-output nonlinear system (Apache helicopter).
• Demonstration of the control system's ability to learn helicopter maneuvers and perform tracking control.
• Validation of design robustness through simulations under diverse disturbance scenarios.
• The DNDP control system framework proved effective for the Apache helicopter model.

Conclusions:

• The DNDP-based control system offers a viable solution for complex nonlinear control problems, exemplified by helicopter flight control.
• This systematic application of approximate dynamic programming to a complex, uncertain system represents a significant advancement.
• The developed DNDP framework is adaptable and shows potential for general-purpose tracking control applications beyond helicopters.