Universal Adaptive Neural Network Predictive Algorithm for Remotely Piloted Unmanned Combat Aerial Vehicle in...

Area of Science:

• Robotics and Control Systems
• Aerospace Engineering
• Artificial Intelligence

Background:

• Remotely piloted unmanned combat aerial vehicles (UCAVs) offer future air combat advantages but face challenges.
• Network time delays degrade piloting performance in UCAV systems.
• A universal predictive method is needed for piloting various UCAVs with unknown control systems.

Purpose of the Study:

• To develop a universal modeling method for remotely piloted UCAVs.
• To address nonlinear uncertainties and time delays in UCAV control systems.
• To ensure stable and accurate remote piloting of diverse UCAVs.

Main Methods:

• A universal nonlinear uncertain model was established for UCAV dynamics.
• A neural network observer was employed for online identification of nonlinear dynamics.
• An adaptive law was designed for neural network weight stability.
• A time-varying delay state predictor was developed to compensate for network transmission delays.

Main Results:

• The proposed method accurately identifies nonlinear UCAV dynamics online.
• The observer-predictor system demonstrated uniform ultimate boundedness (UUB), ensuring stability.
• Simulations confirmed the method's effectiveness in compensating for time delays.
• The approach proved universal for piloting two different UCAV models.

Conclusions:

• The novel method effectively overcomes time delays and model uncertainties in UCAV remote piloting.
• The developed system offers a stable and accurate solution for future unmanned aerial combat.
• The universality of the method allows for flexible control of various UCAV platforms.