1. Background and Current Status

UAVs play an important role in outdoor rescue, forest monitoring, and disaster management. However, outdoor environments are complex, including hills, rivers, forests, geohazards, and dynamic obstacles (moving objects, animals, wildfire spread, etc.).

So far, our research has accomplished:

Environmental data processing: Extracted roads from UAV images and fused 2D images with 3D height information to build maps with road conditions and elevation data.
Path planning algorithm comparison: Implemented and compared static planning algorithms (e.g., A*, Dijkstra), providing a baseline for dynamic path planning.
Static planners work in static environments but cannot handle dynamic obstacles in complex outdoor scenarios. The next step is to develop dynamic path planning to improve UAV autonomy, safety, and adaptability in outdoor environments.

2. Research Objectives and Expected Outcomes

Dynamic environmental perception and modeling：Update 2D/3D maps with road and elevation data dynamically；Detect and model dynamic obstacles (moving objects, wildfire spread, falling rocks, etc.) in real time

Dynamic obstacle prediction model：Predict obstacle movement to provide spatiotemporal information for planning

Global + local dynamic path planning system：Global planning: Generate initial paths considering terrain, roads, and height information
Local planning: Real-time obstacle avoidance using spatiotemporal graphs or learning-based planners (e.g., RL or diffusion-based methods)

Dynamic path planning experiments：Compare static, incremental dynamic, and learning-based planners in dynamic outdoor environments
Verify feasibility and performance advantages

Performance evaluation metrics

Path safety: success rate of avoiding dynamic obstacles, minimum safe distance
Path optimality: path length or flight time relative to ideal path
Dynamic adaptability: path update efficiency and stability
Computational efficiency: real-time planning performance suitable for UAV flight