EA-AHS: A Perception-Driven Adaptive Heuristic Framework for Real-Time UAV Path Planning in Complex Urban...

Area of Science:

• Robotics
• Artificial Intelligence
• Autonomous Systems

Background:

• Urban low-altitude airspace opening necessitates advanced drone navigation.
• Traditional path planning algorithms face challenges with real-time sensor data processing in obstacle-dense environments.
• Computational bottlenecks and unsafe trajectories hinder drone operations.

Purpose of the Study:

• To develop a novel framework for efficient and safe drone navigation in complex urban environments.
• To bridge the gap between raw environmental perception and agile decision-making for drones.
• To enhance the practicality of drone path planning for resource-constrained onboard systems.

Main Methods:

• Proposed the Environment-Aware Adaptive Heuristic Search (EA-AHS) framework.
• Implemented adaptive heuristic weights using a sliding window based on local obstacle density.
• Incorporated a historical feedback loop for macro-level parameter adaptation.

Main Results:

• EA-AHS reduced planning time by up to 87.9% compared to standard A*.
• In 3D urban scenarios, EA-AHS achieved significantly lower cumulative risk (48% of standard A*) and increased minimum obstacle distance to 6.04 m.
• Demonstrated substantial safety gains with only a modest computational overhead.

Conclusions:

• EA-AHS offers a practical, lightweight solution for real-time drone navigation.
• The framework requires no pre-training or neural inference, making it suitable for onboard systems.
• EA-AHS enhances both the efficiency and safety of drone path planning in complex environments.