A Reinforcement-Learning-Driven Multi-Strategy Spherical-Vector Grey Wolf Optimizer for UAV 3D Path Planning

Area of Science:

• Robotics and Automation
• Artificial Intelligence
• Computational Optimization

Background:

• Unmanned Aerial Vehicles (UAVs) are crucial for tasks like surveying, inspection, and environmental monitoring.
• Effective path planning for UAVs in complex 3D terrain presents significant challenges, including terrain undulations, no-fly zones, safety clearance, and trajectory smoothness.
• Conventional optimization algorithms often exhibit instability and premature convergence, hindering optimal path generation.

Purpose of the Study:

• To develop an advanced path planning algorithm for UAVs operating in complex 3D environments.
• To address limitations of existing optimization techniques by improving search stability and convergence.
• To enhance the safety, feasibility, and efficiency of UAV trajectories.

Main Methods:

• Proposes a reinforcement-learning-driven multi-strategy spherical-vector grey wolf optimizer (TLQ-SGWO).
• Employs Tent-Logistic hybrid initialization for enhanced population diversity and Q-learning for adaptive search strategy scheduling.
• Utilizes spherical-vector increments for trajectory encoding and a comprehensive cost function for path optimization.

Main Results:

• TLQ-SGWO demonstrated superior performance on the CEC2017 benchmark functions, achieving the best average rankings in mean error and standard deviation.
• In complex mountainous terrain scenarios, TLQ-SGWO achieved the lowest mean path cost in most cases.
• The algorithm successfully generated stable and feasible 3D trajectories even with increasing no-fly-zone complexity.

Conclusions:

• The proposed TLQ-SGWO algorithm offers a robust and effective solution for UAV path planning in challenging 3D terrains.
• The integration of reinforcement learning and hybrid optimization strategies significantly improves path optimization accuracy and stability.
• This advancement contributes to safer and more efficient UAV operations in diverse real-world applications.