UAVRM-A*: A Complex Network and 3D Radio Map-Based Algorithm for Optimizing Cellular-Connected UAV Path Planning.

Area of Science:

• Robotics and Autonomous Systems
• Network Engineering
• Aerospace Engineering

Background:

• Path planning for cellular-connected Unmanned Aerial Vehicles (UAVs) often relies on computationally intensive methods like deep reinforcement learning (DRL).
• Conventional A* algorithms lack UAV-specific flight dynamics considerations and effective obstacle avoidance.
• Existing methods face challenges with long training times and poor generalization.

Purpose of the Study:

• To develop an improved path planning algorithm for cellular-connected UAVs that addresses limitations of traditional A* and DRL.
• To enhance situational awareness using a 3D radio map for more effective navigation.
• To create a computationally efficient and reliable path planning solution.

Main Methods:

• Proposed the UAVRM-A* algorithm, an enhancement of the A* algorithm built upon complex network navigation models.
• Integrated a 3D radio map to improve situational awareness and guide path planning.
• Evaluated the algorithm's performance against conventional methods and DRL-based approaches.

Main Results:

• The UAVRM-A* algorithm demonstrated effective obstacle avoidance and generated flight paths consistent with UAV dynamics.
• Achieved performance comparable to DRL methods with significantly reduced radio outage duration.
• Showcased a substantial reduction in computational time for model training compared to DRL.

Conclusions:

• UAVRM-A* offers a more efficient, reliable, and practical path planning solution for cellular-connected UAVs.
• The algorithm provides enhanced situational awareness and improved flight path generation.
• This research paves the way for advanced UAV applications in delivery, surveillance, and emergency response.