An End-to-End Path Planning Network for UAV in Noisy Environments

Unmanned Aerial Vehicles (UAVs) have become crucial tools for post-disaster rescue and equipment inspection, offering significant social value. However, this demands exceptional navigation capabilities from UAVs in unknown and noisy environments. This paper proposes a novel robust path planning strategy to address the challenges faced by UAVs navigating in such conditions. The strategy integrates an image denoising module, ALSP-ID, which calculates similarity weights using Euclidean distance and vectorized representation. This effectively removes noise from depth maps while preserving edge information. By constructing an end-to-end network architecture, the perception and planning functions are trained simultaneously, reducing computational resource requirements. Additionally, the introduction of a channel attention mechanism allows the network to better focus on critical feature channels, enhancing the model's expressive ability and generalization performance. Experiments firmly validate the effectiveness of the proposed method in improving path planning robustness, path smoothness, and obstacle avoidance capabilities. Notably, it excels in handling noise and adapting to unknown environments, providing strong technical support for the autonomous navigation of UAVs in complex scenarios.