Dynamic UAV Deployment Scheme Based on Edge Computing for Forest Fire Scenarios.

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Engineering

Background:

• Forest fires present dynamic challenges requiring adaptable resource allocation.
• Unmanned Aerial Vehicles (UAVs) offer flexible deployment but require efficient management.
• Edge computing enhances UAV capabilities for real-time data processing.

Purpose of the Study:

• To develop a two-timescale dynamic deployment scheme for UAVs in forest fire scenarios.
• To address the varying resource demands posed by the evolving nature of forest fires.
• To optimize UAV deployment for efficient task execution and reduced latency.

Main Methods:

• A two-timescale model considering dynamic changes in UAV number and position.
• Utilizing a Gate Recurrent Unit (GRU) for predicting user numbers and determining UAV quantity.
• Implementing a deep reinforcement learning algorithm for real-time UAV position adjustment.

Main Results:

• The proposed scheme demonstrates improved prediction accuracy for resource demands.
• Effective adaptation of UAV numbers and positions to changing requirements.
• Significant reduction in task execution delays compared to existing methods.

Conclusions:

• The dynamic UAV deployment scheme effectively manages resources in forest fire scenarios.
• The integration of GRU and deep reinforcement learning optimizes UAV operations.
• This approach enhances the efficiency and responsiveness of UAVs in critical situations.