Path Planning Research of a UAV Base Station Searching for Disaster Victims' Location Information Based on Deep...

Area of Science:

• Robotics and Automation
• Artificial Intelligence
• Wireless Communications

Background:

• Unmanned Aerial Vehicles (UAVs) are increasingly used for search tasks, requiring efficient path planning.
• Traditional path planning algorithms face challenges with complex environments and real-time decision-making.
• Deep Reinforcement Learning (DRL) offers a promising approach for optimizing UAV operations.

Purpose of the Study:

• To develop an advanced path planning algorithm for UAVs performing search missions.
• To enhance the efficiency, stability, and convergence speed of UAV path planning.
• To reduce the decision-making complexity for UAVs in dynamic environments.

Main Methods:

• Proposed a Double DQN-state splitting Q network (DDQN-SSQN) algorithm integrating state splitting and optimal state concepts.
• Utilized Deep Reinforcement Learning (DRL) based on the Double Deep Q-Network (DDQN) algorithm.
• Incorporated Received Signal Strength Indicator (RSSI) into the agent's reward mechanism.
• Simulated UAV mission scenarios using the Open AI Gym platform.

Main Results:

• The DDQN-SSQN algorithm demonstrated faster optimal path planning compared to traditional methods.
• The proposed scheme exhibited superior stability and convergence speed in simulations.
• State splitting and RSSI integration effectively reduced UAV decision-making difficulty.

Conclusions:

• The DDQN-SSQN algorithm provides an effective solution for UAV path planning in search tasks.
• The integration of DRL, state splitting, and RSSI enhances planning performance.
• The proposed method offers significant advantages in speed, stability, and convergence for UAV operations.