AI-Powered Platform Takes Autonomous Flight to New Heights

Statistics indicate that small planes are more prone to accidents compared to larger commercial aircraft, primarily due to errors from less experienced pilots and the absence of advanced computer-aided safety systems. Technologies such as traffic advisory, predictive, and avoidance systems are often missing in smaller aircraft. However, advancements in Edge Computing and Artificial Intelligence (AI) are addressing these challenges, enabling the development of autonomous systems for airborne vehicles powered by compact edge AI servers.

A leader within the eVTOL (Electric Vertical Takeoff and Landing) market  wished to utilise a AI inference platform to develop its machine learning and autonomous flight solutions. The clients innovations are transforming the aviation industry, with them introducing a new line of autonomous aircraft with one to four passenger seats, commonly referred to as “air taxis.” These self-piloted, battery-powered aircraft are designed for vertical takeoffs and landings, heralding the dawn of “urban air mobility” where flying cars become a commuting option. Although civilian autonomous flight regulations are still evolving, a Morgan Stanley Research BluePaper estimates the electric vertical takeoff and landing (eVTOL) market will reach $1.5 trillion by 2040. The autonomous system developed by this Silicon Valley-based innovation centre has also been tested on large commercial aircraft (weighing up to 315 tons) for automated takeoffs and landings to mitigate human error caused by fatigue and stress.

Several critical challenges needed addressing in the development of the AI-based autonomous flight system:

• Safety Concerns and Certifications:
  While autonomous vehicles have long been used in the defence sector, their expansion into civilian aviation requires public trust. Demonstrating that machines can make better, safer decisions than human pilots is essential. The project’s primary goal was to develop scalable, certifiable systems to establish industry standards and ensure public confidence.
• Data Capacities and Real-Time Reactions:
  AI-driven autonomous flight demands extensive data acquisition to build imagery across varying conditions, including day, night, and poor visibility. Real-time processing capabilities and immediate reactions based on dynamic sensor inputs are critical for safe and efficient operation.
• Super Ruggedness to Withstand Shock and Vibration:
  Small aircraft are more vulnerable to turbulence and adverse weather, as they cannot fly at higher altitudes to avoid such conditions. Therefore, onboard computers must be exceptionally rugged to endure continuous shock and vibration.
• High Performance vs. Compactness and Ruggedness Requirements:
  Balancing high performance with ruggedness and space constraints is challenging, particularly in small aircraft. The solution had to deliver maximum performance and reliability while fitting within limited space and maintaining durability.

The Neousys AI inference platform, Nuvo-7166GC, powered by an NVIDIA® Tesla T4 GPU and an Intel® 9th/8th-Gen Core™ processor, was selected for this autonomous flight project. Each autonomous system on the aircraft leverages the Nuvo-7166GC for real-time data collection, AI inference, and precise aircraft control.

Advantages of Neousys Nuvo-7166GC

• Compact Design:
  The Nuvo-7166GC is one of the smallest NVIDIA-Tesla qualified servers designed for T4 support. Its compact size allows it to fit seamlessly into space-constrained systems while delivering unparalleled processing power and AI acceleration.
• Super Rugged Construction:
  Built for mission-critical operations, the Nuvo-7166GC utilises Neousys’ patented Cassette module and an optimised cooling solution. It can operate at temperatures up to 60°C under 100% CPU and GPU load and withstand continuous shock and vibration, meeting military standards.
• Versatile I/O Deployment:
  The system offers rich I/O options to support data acquisition and sensory inputs necessary for autonomous flight. It includes an M.2 NVMe interface for ultra-fast disk access, USB 3.1 Gen2, GbE, and PoE ports for data input, and a Gen3 x8 PCIe slot for installing high-performance PCIe cards or various sensor/image acquisition cards.

Key Benefits

• Real-Time AI Processing:
  The Nuvo-7166GC’s powerful GPU and CPU combination ensures real-time AI processing for autonomous flight control, minimising delays and enhancing safety.
• Scalability and Certification:
  The system’s scalable architecture and rugged design support certification processes, contributing to the development of industry standards for civilian autonomous flight.
• Enhanced Safety and Reliability:
  The rugged construction and advanced AI capabilities reduce human error and enhance the overall safety and reliability of autonomous flights.
• Energy Efficiency:
  Despite its high-performance capabilities, the Nuvo-7166GC is designed for energy efficiency, supporting sustainable and eco-friendly aviation solutions.

By integrating Neousys’ Nuvo-7166GC AI inference platform, our client successfully developed an autonomous flight system capable of real-time data processing, precise aircraft control, and robust performance under challenging conditions. This solution not only advances the development of air taxis and urban air mobility but also sets a new standard for safety and reliability in the aviation industry. With the eVTOL market poised for significant growth, Neousys’ AI platform is at the forefront of transforming modern aviation through cutting-edge technology and innovation.