Globally Guided Deep V-Network-Based Motion Planning Algorithm for Fixed-Wing Unmanned Aerial Vehicles.

Area of Science:

• Robotics
• Artificial Intelligence
• Aerospace Engineering

Background:

• Fixed-wing UAVs offer significant potential for diverse applications.
• Safe navigation in complex environments with static and dynamic obstacles remains a challenge.

Purpose of the Study:

• To develop a hierarchical two-layer motion planning algorithm for fixed-wing UAVs.
• To achieve safe, collision-free, and kinematically feasible flight paths.

Main Methods:

• A global planner generates reference guidance considering kinematic constraints and static obstacles.
• A local reinforcement learning (RL) planner handles dynamic obstacles within sensing range.
• A multi-stage, multi-scenario training strategy was employed for simulation.

Main Results:

• The proposed algorithm demonstrated superior performance compared to baseline methods in simulations.
• Effective collision avoidance with both static and dynamic obstacles was achieved.
• Kinodynamic feasibility was maintained throughout the planning process.

Conclusions:

• The hierarchical two-layer algorithm provides an effective solution for fixed-wing UAV motion planning.
• RL-based local planning enhances safety and adaptability in dynamic environments.
• The proposed method significantly improves flight safety and efficiency for UAVs.