APPA-3D: an autonomous 3D path planning algorithm for UAVs in unknown complex environments.

Area of Science:

• Robotics
• Artificial Intelligence
• Aerospace Engineering

Background:

• Unmanned Aerial Vehicles (UAVs) offer flexibility and cost-effectiveness for tasks in complex environments.
• Current UAV path planning often assumes known environments, limiting safe navigation in unknown areas.
• Autonomous collision-free path planning is crucial for reliable UAV operations.

Purpose of the Study:

• To propose an autonomous collision-free path planning algorithm for UAVs in unknown complex 3D environments (APPA-3D).
• To enhance UAVs' ability to navigate safely and reliably in unpredictable conditions.

Main Methods:

• Designed an anti-collision control strategy using a UAV collision safety envelope and environmental awareness.
• Developed a dynamic reward function for reinforcement learning (RL) tailored to flight environments.
• Proposed an optimized RL action exploration strategy based on action selection probability.
• Utilized an improved RL algorithm for training and simulating UAV flight in unknown environments.

Main Results:

• The APPA-3D algorithm was trained through interaction with the simulated unknown environment.
• Demonstrated effective autonomous collision-free path planning for UAVs.
• Comparative experiments confirmed APPA-3D's ability to guide UAVs from start to target safely in unknown complex 3D environments.

Conclusions:

• APPA-3D enables autonomous, collision-free path planning for UAVs in unknown complex 3D environments.
• The algorithm enhances UAVs' capability to perform missions reliably in unpredictable settings.
• This research contributes to safer and more robust UAV navigation systems.