Unlocking New Frontiers in Space Applications with Off-the-Shelf Edge AI Processors

Article originally posted on All About Circuits.

Learn how government space experts are re-evaluating radiation environments and identifying commercial electronics, including new edge AI co-processors, that can meet the mission requirements of future missions.

When we think about space exploration and satellite operations, a common misconception persists: that only highly customized, radiation-hardened microelectronics can survive the harsh environment of outer space. Historically, this view has been valid, contributing to the noticeable lag between the computational capabilities of terrestrial systems and those deployed on space systems.

For example, the Space Shuttle relied on the AP-101 General Purpose Computer (GPC), a radiation-hardened system based on 1960s technology. Even with mid-life upgrades in the 1990s, by the end of the Shuttle program, its processing speed was fully 1,000 times slower than consumer-grade desktop computers while being over 2,000 times more expensive in constant dollars.

The Changing Landscape of Space Electronics

Despite their vast benefits in power and affordability, consumer technology was often seen as too unreliable for space missions. However, the landscape of space electronics is evolving. The industry is undergoing pivotal changes that open the potential for a greater use of commercial off-the-shelf (COTS) microelectronics in space applications, particularly in the domain of edge AI processing.

Figure 1. For space qualification, products must undergo radiation testing. Here, the EdgeCortix SAKURA-I AI Accelerator undergoes heavy ion testing as part of the NASA Electronics Parts and Packaging (NEPP) program. Image used courtesy of EdgeCortix

This shift is not merely theoretical. There are several trends that promise to transform the way we approach space systems. And, as the global space economy totaled $570 billion in 2023, we are at a critical point in time to bring many of these transformative efforts to life.

The Need for Greater Autonomy

As space missions grow in complexity and ambition, the demand for greater autonomy has intensified. Both commercial and government customers are seeking systems that can operate with minimal reliance on ground control. This is critical for missions that aim to deploy complex systems at dramatically lower costs, especially when the orbital environment is becoming increasingly crowded with active objects and space debris. In such scenarios, real-time decisions—free of communication delays with Earth—become essential.

The U.S. Space Force (USSF) and allied national security operators are particularly interested in what has been termed "dynamic space operations." These operations would require systems that can autonomously analyze their surroundings and react accordingly, paving the way for a new generation of space missions that prioritize efficiency and responsiveness.

Advancements in AI Algorithms and Edge Processing Capabilities

The advancements in AI algorithms have significantly bolstered the capabilities of edge processing systems. Technologies such as generative AI and large language models (LLMs) are now being adapted for deployment on edge devices. A notable development in this space is the advent of custom-designed edge AI co-processors (like the one shown in Figure 2), which offer remarkable efficiency compared to traditional CPUs and GPUs.

Figure 2. EdgeCortix SAKURA-II AI accelerator in a small form factor M.2 module. Image used courtesy of EdgeCortix

For example, processors that deliver high Tera Operations Per Second (TOPS) of real-time processing while consuming minimal wattage are critical. Solutions that support fully generative AI capabilities and boast high DRAM bandwidth can manage complex multi-billion parameter models. This level of processing prowess is crucial for the autonomous systems that future space missions will rely on, allowing them to perform complex analysis and decision-making processes onboard.

The Suitability of Commercial Microelectronics

Commercial microelectronics are rapidly becoming more suited for space applications. Modern transistor technology, characterized by smaller feature sizes, reduces the volume of silicon that a single radiation particle can impact, thereby lowering the risk of operational disruptions. For instance, Fin Field Effect Transistor (FinFET) technology provides better control over the charge within the channel, making systems less susceptible to destructive effects like Single Event Latch-ups (SELs) and Total Ionizing Dose (TID) radiation.

Additionally, mission designers now have access to advanced mitigation techniques, including improved Error Correction Codes and other established algorithms, which help manage transient radiation effects such as Single Event Upsets (SEUs) and Single Event Functional Interrupts (SEFIs). These tools allow for greater reliability without compromising mission performance.

A compelling example of this new mindset is the Ingenuity Mars Helicopter (Figure 3), which utilizes a Qualcomm Snapdragon 801 processor—an instance of commercial technology successfully operating in a space environment and performing functions that no traditional radiation-hardened processor could manage in that mission.

Figure 3. NASA technicians inspect the Ingenuity helicopter before launch to Mars. Image used courtesy of NASA

Government space experts are evaluating radiation environments and currently assessing commercial electronics to identify components that can meet mission requirements. For example Sandia National Laboratories has published findings suggesting that the threshold for TID tolerance for many missions could be lowered to 50 Krad, expanding the range of commercially available parts suitable for space applications.

Furthermore, the NASA Electronics Parts and Packaging program (NEPP) has conducted radiation testing on several commercial components, including Edge AI co-processors as shown in Figures 1 and 4, affirming their potential viability for future missions.

Figure 4. EdgeCortix SAKURA-I AI Accelerator undergoing proton testing for NEPP. Image used courtesy of EdgeCortix

Transforming the Space Economy with Edge AI

The convergence of these trends signifies a transformative era for the space economy. The integration of edge AI into space systems promises to enhance autonomy and operational efficiency, pushing the boundaries of what is possible in space exploration.

These capabilities not only enhance mission efficiency but also lower costs, making space exploration more accessible. As commercial entities increasingly participate in space endeavors, collaboration between public and private sectors will yield innovative solutions that propel the space economy forward.

The Future of Space Exploration

The landscape of space applications is rapidly changing, driven by advancements in COTS edge AI processors and a growing acceptance of commercial microelectronics in space. The need for greater autonomy, coupled with the rise of sophisticated AI algorithms, sets the stage for a new era of space exploration. By leveraging these technologies, we can realize ambitious visions of autonomous space systems that not only operate efficiently but also are agile to the challenges of the space environment.

The future of space exploration is not only about pushing the limits of technology but also about reimagining the possibilities of what humans can achieve in the vast expanse of space. Powered by COTS edge AI processors, the next generation of space missions is set to transform our understanding of the universe and our role in it.

Feature image used courtesy of EdgeCortix (foreground) and Canva (background).