CAVU Aerospace UK, 2024 – As the space industry rapidly evolves, the demand for robust and efficient space edge computing solutions has never been greater. The primary drivers behind this demand include the exponential growth in satellite launches, the increasing complexity of satellite missions, and the rising volume of data generated by these missions. Traditional methods of data management, which involve sending large amounts of raw data from satellites to ground stations for processing, are becoming less feasible due to bandwidth limitations, latency issues, and the sheer volume of data.

Key Factors Driving Demand:

• Proliferation of Satellite Constellations: The number of operational satellites, particularly in Low Earth Orbit (LEO), is projected to increase significantly. With over 5,401 Earth observation satellites expected to be operational by 2033, the data generated from these constellations will require advanced onboard processing to manage and analyze efficiently.

• Data Volume and Transfer Costs: The cost of transmitting data from space to Earth is considerable, especially for high-resolution imaging and SAR data. With costs estimated at around $1,048,576 per terabyte using traditional RF communication methods, reducing the amount of data transmitted by processing it in space can lead to substantial cost savings.

• Latency and Real-Time Processing: For applications such as disaster response, defense, and environmental monitoring, real-time data processing and decision-making are crucial. Space edge computing enables satellites to perform these tasks onboard, reducing latency and improving response times.

• Energy Efficiency and Power Constraints: Space missions, particularly those involving CubeSats and small satellites, are limited by the power available from solar panels. Therefore, efficient data processing solutions that minimize power consumption are essential.

Emerging Trends in Space Edge Computing

The future of space missions is heavily influenced by several emerging trends in space edge computing, which are aimed at enhancing the efficiency and capabilities of satellite operations.

• Laser Communication Systems: A significant trend is the adoption of laser communication systems, which offer high-speed data transfer capabilities. For example, the TeraByte InfraRed Delivery (TBIRD) system has demonstrated speeds of up to 100 Gbps, enabling the transmission of multiple terabytes of data during a single satellite pass over a ground station. This technology not only reduces the need for large and expensive RF ground stations but also minimizes latency and increases the volume of data that can be transmitted back to Earth. However, laser communication requires precise alignment and is susceptible to atmospheric conditions, which can pose challenges for consistent data transfer.

• Advanced AI and Machine Learning Applications: The use of AI and machine learning in space is becoming increasingly prevalent. Satellites equipped with AI capabilities can autonomously process data, identify relevant information, and make decisions without human intervention. This capability is particularly valuable for applications such as anomaly detection, predictive maintenance, and autonomous navigation.

• Onboard Data Compression and Processing: Another trend is the development of advanced algorithms for data compression and processing onboard satellites. By compressing data before transmission, satellites can reduce the required bandwidth and save on transmission costs. Onboard processing also enables satellites to perform preliminary analysis and send only critical data, further reducing the data load on ground stations.

• Integration of Edge Computing with Satellite Constellations: As satellite constellations become more complex, integrating edge computing capabilities across multiple satellites will be crucial. This integration allows for distributed processing, where data can be processed collaboratively by multiple satellites in the constellation, enhancing overall system efficiency and resilience.

Estimating Data Transfer and Processing Needs

Given the projected growth in satellite deployments and the increasing data demands, the total amount of data that needs to be transferred to Earth is expected to rise dramatically. For example, a constellation of Earth observation satellites could generate several petabytes of data annually. If a constellation like ICEYE’s SAR satellites were to generate approximately 4.3 terabytes per day, the annual data output could exceed 1.57 petabytes. When considering the combined data from multiple constellations, the total volume could reach tens of exabytes per year.

To handle this data deluge, space edge computing solutions can significantly reduce the amount of data needing transmission. By processing and filtering data onboard, these systems can cut down data transfer volumes by up to 90%, saving costs and reducing the strain on communication infrastructure.

Embedded AI Processing Platforms

To address the growing need for advanced data processing capabilities in space missions, various embedded AI processing platforms are available, each offering unique advantages in terms of performance, power consumption, and suitability for space applications. This table compares the specifications of leading AI processors, including those from NVIDIA, Google Coral, Teledyne, and others, focusing on their use in CubeSats and small satellites. These processors enable real-time data processing, AI inference, and efficient data management, essential for the operational success of modern satellite constellations.

Processor/Platform	Processing Power	Power Consumption	Suitability
NVIDIA Jetson AGX Orin	Up to 275 TOPS	15W – 60W	High-performance tasks, co-processor capable
NVIDIA Jetson Orin NX	Up to 100 TOPS	10W – 25W
NVIDIA Jetson Xavier NX	Up to 21 TOPS	7.5W – 15W	Moderate AI tasks, co-processor
NVIDIA Jetson Orin Nano	Up to 40 TOPS	5W – 15W	Entry-level AI applications, co-processing
Google Coral Edge TPU	4 TOPS (per TPU)	2W per TPU (0.5W/TOPS)	Low-power AI inferencing, co-processor
Intel Movidius Myriad X	Up to 1 TOPS	<2W	Computer vision, low-power AI tasks
Mythic M1076	Up to 25 TOPS	3W – 4W	High efficiency AI inference, low power
Teledyne Qormino	Quad ARM Cortex A72, 1.8GHz	Low, exact value not specified	Robust processing in compact spaces, radiation tolerant

It is important to note that most embedded AI processing platforms, except for specific solutions like Teledyne’s Qormino, are not inherently designed for space environments due to vulnerabilities to Single Event Upsets (SEU), Total Ionizing Dose (TID), and latch-up issues. To address the challenges posed by space environments, CAVU Aerospace UK employs a specialized design incorporating Microchip’s flash-based FPGA SoCs, such as the PolarFire series. These SoCs are equipped with robust features including Error Correction Code (ECC), Error Detection and Correction (EDAC), and Triple Modular Redundancy (TMR), which significantly enhance reliability by mitigating effects like Single Event Upsets (SEU). Additionally, Magnetoresistive memories (MRAM) and radiation-tolerant or radiation-hardened versions of the PolarFire SoC are used to safeguard against radiation-induced failures. These measures, combined with distributed processing among co-processors, ensure robust performance and data integrity in the challenging conditions of space.

CAVU Aerospace UK’s Space Edge Computing Solutions

CAVU Aerospace UK offers a suite of cutting-edge space edge computing solutions, each designed to meet the specific needs of CubeSats and small satellite missions. These solutions integrate the versatile Microchip Flash Based PolarFire SoC with a variety of powerful co-processors, providing a range of performance options to suit different mission requirements:

• Tartan Typhoon: Combining the PolarFire SoC with the Jetson AGX Orin, this solution offers high computational power, ideal for intensive tasks such as detailed environmental monitoring and complex data analytics.
• Tartan Might: Featuring the PolarFire SoC paired with the Jetson Orin NX, this configuration is designed for demanding AI applications, providing powerful processing capabilities with efficient power consumption.
• Tartan Edge: Utilizing the PolarFire SoC alongside the Jetson Xavier NX, this balanced solution excels in moderate AI and data processing tasks, offering a harmonious blend of performance and energy efficiency.
• Tartan Light: This solution includes the PolarFire SoC with the Jetson Nano, providing a cost-effective option for basic AI applications. It ensures low power consumption, making it suitable for energy-constrained missions.
• Unicorn Lite: Incorporating the PolarFire SoC with Teledyne Qormino co-processors, Unicorn Lite delivers reliable performance in harsh space environments. It is particularly well-suited for real-time data processing and communication systems.
• Unicorn Might: Paired with four Mythic M1076 M.2 cards, this solution leverages the PolarFire SoC to maximize data processing capabilities, offering efficient AI processing for data-intensive applications.

These solutions are specifically tailored to address the complex challenges of space missions, including real-time data processing, cost efficiency, and operational reliability. By leveraging advanced edge computing technologies, CAVU Aerospace UK enables satellite missions to operate more effectively, reduce data transfer costs, and deliver real-time insights from space. As the space industry continues to expand, these solutions are set to play a crucial role in advancing satellite technology and exploration.