Transforming Aviation: How Project Neural Wing Aims for Fully Autonomous Flight

The future of aviation is moving fast, and Project Neural Wing is at the forefront of this change. This initiative is transforming the Sling HW aircraft from a plane controlled by a pilot to a fully autonomous system guided by artificial intelligence. Using a Vision Transformer (ViT) architecture, the aircraft will handle complex tasks such as take-offs, pattern flying, and emergency landings by interpreting visual data directly into control commands. This post explores how this project works, its technical foundation, regulatory environment, and safety measures.

How Project Neural Wing Works

Project Neural Wing uses a Vision Transformer, a type of AI model that processes images to make decisions. Instead of relying on traditional sensors alone, the system analyses raw pixel data from cameras to control the plane. This approach allows the aircraft to perform manoeuvres that usually require a skilled pilot, including:

• Take-offs: Smooth and precise lift-offs using real-time visual input.

• Pattern Work: Flying standard traffic patterns around airports autonomously.

• Emergency Landings: Detecting safe landing spots and executing landings without human input.

  
  

The AI interprets the environment through multiple cameras and sensors, converting visual information into control signals instantly.

Technical Foundation of the Project

The Sling HW is an ideal platform for this AI integration because of its advanced hardware and design. Key technical elements include:

• Powerful Compute Hardware: The NVIDIA Orin AGX with 64GB memory delivers trillions of operations per second (TOPS), enabling the AI to process complex data in less than 50 milliseconds.

• Comprehensive Perception System: A 4-camera GMSL2 global shutter array offers 360-degree visual coverage. This is paired with solid-state LiDAR sensors that provide sub-centimetre accuracy, especially useful during landing flares.

• Safety Architecture: A Simplex Architecture runs on an independent STM32 chip, acting as a watchdog. This system ensures the AI does not make unsafe decisions by providing a safety envelope that prevents "hallucinations" or errors in AI perception.

  
  

These components work together to create a reliable and responsive autonomous flight system.

Navigating the Regulatory Landscape

The FAA’s 2026 MOSAIC rule has opened new doors for projects like Neural Wing. This regulation simplifies certification and testing for automated aircraft by:

• Introducing Full Automation Designation: Aircraft with complete automation can now follow new certification paths that reduce pilot training requirements.

• Allowing Flexible Weight Limits: The removal of strict weight caps for Light Sport Aircraft (LSA) means the extra hardware needed for AI and sensors can be added without losing legal status.

• Enabling Experimental and Restricted Operations: Initial testing can happen under "Experimental" status, with a clear progression to "Restricted" commercial uses such as cargo delivery or aerial surveying as flight hours increase.

  
  

This regulatory environment supports innovation while maintaining safety standards.

Managing Risks with Safety Protocols

Safety is a top priority for Project Neural Wing. The system uses a Redline Safety Protocol to handle unexpected situations, especially those outside the AI’s training data. For example:

• Engine Failure on Take-off (EFATO): If the engine fails during take-off, the AI performs a rapid 100-millisecond energy-state calculation. It then selects the safest option, such as returning to the runway or finding an emergency landing spot.

• Out-of-Distribution Events: When the AI encounters scenarios it has not seen before, the Simplex Architecture steps in to maintain control and prevent unsafe actions.

  
  

These measures ensure the aircraft can handle emergencies with minimal risk.

What This Means for Aviation

Project Neural Wing shows how AI can take on complex flight tasks safely and efficiently. By combining powerful computing, advanced sensors, and strict safety protocols, the project moves closer to fully autonomous flight. This technology could reduce pilot workload, improve safety, and open new possibilities for commercial and recreational aviation.

The FAA’s updated rules make it easier to test and certify these systems, accelerating their adoption. As autonomous flight becomes more common, it will change how we think about flying and aircraft operation.

The next step is watching how these systems perform in real-world conditions and how they evolve with more flight hours and data. For anyone interested in aviation or AI, Project Neural Wing offers a clear example of how technology can reshape the skies.