USS-Net: A neural network-based model for assisting flight route scheduling.

Area of Science:

• Aviation Safety and Air Traffic Management
• Artificial Intelligence in Aerospace
• Computer Vision for Navigation

Background:

• Current aviation navigation systems (GPS, INS) lack environmental awareness and early warning capabilities.
• Air traffic congestion poses risks, necessitating advanced solutions for coordinated flight and safety.
• Wake turbulence detection is crucial for preventing accidents during multi-aircraft operations.

Purpose of the Study:

• To propose a multi-aircraft parallel approach for coordinated flight using advanced AI.
• To develop a neural network-based semantic segmentation model for enhanced situational awareness and route scheduling.
• To improve flight safety by enabling early detection of wake turbulence.

Main Methods:

• Development of a U-shaped State Space Block UNet (USS-Net) integrating Mamba for temporal dynamics.
• Utilizing StateConvBlock and ResConvBlock for feature extraction and multi-scale fusion.
• Employing a neural network-based semantic segmentation model for aircraft monitoring and environmental awareness.

Main Results:

• USS-Net achieved high precision pixel-level segmentation on an aircraft simulation dataset.
• The model attained a mean Intersection over Union (mIoU) of 95.70% and pixel accuracy (PA) of 97.80%.
• Demonstrated effectiveness in assisting multi-aircraft parallel route scheduling tasks.

Conclusions:

• USS-Net significantly enhances situational awareness and safety in aviation.
• The model's efficiency and accuracy support advanced multi-aircraft coordination.
• This approach offers a promising solution for mitigating risks associated with air traffic congestion.