A CCO-PPO Framework for Autonomous UAV Trajectory Tracking in Complex and Disturbed Environments.

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence and Machine Learning
• Aerospace Engineering

Background:

• Accurate trajectory tracking is crucial for autonomous unmanned aerial vehicle (UAV) operations.
• Proximal Policy Optimization (PPO) shows promise for UAV control but requires extensive manual hyperparameter tuning due to interparameter coupling.

Purpose of the Study:

• To develop an automated framework, CCO-PPO, for optimizing PPO hyperparameters for UAV trajectory tracking.
• To enhance the robustness and performance of UAVs in complex and high-disturbance environments.

Main Methods:

• Formulated UAV trajectory tracking as a Markov decision process with a 20-dimensional state space.
• Integrated the cuckoo catfish optimizer (CCO) for offline, automated hyperparameter search in a four-dimensional space.
• Evaluated CCO-PPO across seven diverse test environments, including varying trajectories, wind, sensor noise, and scale.

Main Results:

• CCO-PPO achieved the lowest tracking error across all tested environments.
• Performance gains over baseline PPO increased with task complexity, reaching 18.8% under combined wind and sensor noise.
• Demonstrated statistically significant advantages over PPO, SAC, and TD3 in 85.7% of comparisons.

Conclusions:

• Metaheuristic hyperparameter optimization using CCO substantially improves policy robustness for UAV trajectory tracking, especially in high-disturbance conditions.
• Joint optimization of all four hyperparameters is critical for performance under challenging environmental factors.
• CCO-PPO offers superior cross-seed stability compared to Bayesian optimization.