- HPE’s Spaceborne Computer-2.5, upgraded from its predecessor, Spaceborne Computer-2, is pioneering advanced AI and machine learning tasks in space.
- Equipped with modern hardware and extensive storage capacity, Spaceborne-2.5 enables experiments and technology showcases for researchers.
- The computer’s robust capabilities, including extensive Kioxia flash memory, facilitate accelerated DNA monitoring and AI testing for space applications.
- HPE is forging collaborations between space-based and terrestrial systems, demonstrating the potential for federated learning in space missions.
- With plans for additional federation members on the ISS and aspirations to commercialize space-based computing, HPE is paving the way for expanded commercial utilization of space data centers.
- Future missions, including Spaceborne Computer-3’s journey to the moon, signal continued innovation and adaptation to the challenges of space computing.
Main AI News:
In a recent upgrade, HPE’s Spaceborne Computer-2 aboard the International Space Station is embarking on increasingly intricate artificial intelligence and machine learning tasks.
“It’s a remarkably capable system currently orbiting in space,” remarked Mark Fernandez, HPE Space Technologies and Solutions chief scientist and principal investigator for Spaceborne Computer-2, to SpaceNews. “I haven’t turned down any experiment requests due to hardware constraints.”
The initial Spaceborne Computer, deployed on the ISS in 2017, served as a trial to assess the functionality of a commercial off-the-shelf HPE processor in space. Its successor, Spaceborne Computer-2, derived from HPE’s more robust Edgeline computers, has executed numerous experiments since its arrival at the ISS in 2021.
Half of Spaceborne Computer-2 returned to Earth in 2023 for upgrades. The latest iteration, officially designated Spaceborne Computer-2 but colloquially referred to as Spaceborne-2.5, journeyed to the space station on Jan. 30 via a Northrop Grumman Commercial Resupply mission.
Spaceborne-2.5, equipped with 130 terabytes of Kioxia storage, a modern operating system, new software, and enhanced security features, is showcasing potential applications of artificial intelligence in future lunar and planetary endeavors.
Crafting the Future with SPACEBORNE-2.5
HPE extends an invitation to researchers to propose scientific experiments and technology showcases for Spaceborne-2.5.
“I have top-notch CPUs. I have cutting-edge GPUs,” Fernandez declared. “Alongside a contemporary operating system supporting Docker, we offer a suite of artificial intelligence frameworks and machine learning capabilities.“
Clients incur no charges for Spaceborne Computer services.
“We are prohibited from levying fees for our services,” Fernandez explained. “Conversely, we are unable to deploy any production applications on behalf of clients. Our focus lies in conducting numerous proof of concepts.“
Spaceborne-2.5’s extensive Kioxia flash memory resolves a fundamental challenge of space-based edge processing.
“I possess an immense volume of edge data. However, the dilemma lies in storage,” Fernandez pondered. “Now, I can archive copious DNA sequences, Earth observations, and video footage. This prospect excites me greatly.“
The computer’s update has facilitated expedited DNA analysis for HPE and NASA. Previously, NASA researchers compared each astronaut’s DNA with the standard human genome, leading to daily identification of any anomalies.
However, HPE has now uploaded the standard human genome to Spaceborne-2.5.
“Today, I compare Astronaut A’s genome with yesterday’s iteration,” Fernandez elaborated. “It’s far more reassuring to inform the astronaut, ‘No changes since yesterday,’ rather than repeating prior anomalies.”
Advancing AI Paradigms
Spaceborne-2.5 is pioneering AI applications for space missions. As highlighted in a recent SpaceNews webinar, AI models, necessitating extensive computational resources and data storage, undergo training on vast datasets in terrestrial clouds.
Cloud processing yields an inference engine, the cornerstone software facilitating predictions and conclusions by AI systems. These inference engines can be shared with space-based and terrestrial edge processors within a federated learning framework. As federation members acquire new data, they collaborate to refine the underlying AI model.
“Completely replacing your inference engine could consume substantial bandwidth. However, leveraging Docker layers can expedite the process,” Fernandez noted.
In a recent experiment, HPE interconnected an AI model running on Spaceborne-2.5 with the terrestrial AWS Cloud.
“The inference engine on ISS suggests parameters for the model, transmitted to the mothership,” Fernandez described.
The AWS cloud-based mothership scrutinizes these parameters against extensive public cloud datasets and proposes refinements. After hours of deliberation, the two inference engines converge on an agreement. Fernandez has devised programming directives enabling seamless resumption of discussions upon restoration of satellite communications.
Pioneering New Frontiers
As spacecraft venture farther from Earth, synergy between edge computers and terrestrial clouds or potent space-based motherships assumes greater significance.
For instance, entities seeking water on the moon are likely to deploy multispectral sensors. Rather than dispatching observations to Earth for analysis, data may be transmitted to nearby edge devices equipped with inference engines trained on Earth to detect water indicators.
“In seconds, I can identify dry areas, allowing rovers or instruments to relocate swiftly,” Fernandez affirmed.
As inference engines evolve, lunar datasets could undergo reassessment.
“Given ample storage capacity, I can reprocess raw data from previous sites,” Fernandez elaborated. “All raw data remains at the edge, with insights relayed back to Earth.”
Expanding Collaborative Endeavors
HPE is primed to install additional federation members on the space station.
“These units will engage in dialogue with the mothership,” Fernandez stated. “By amalgamating insights from various sources, the mothership, armed with extensive computational power and public cloud storage, can devise optimal parameters for distribution.”
“This initiative is currently underway,” Fernandez confirmed.
Pioneering Commercial Ventures
Simultaneously, HPE explores avenues for commercializing space-based computing. Demonstrating SAP Hana database software in space holds prominence.
“Deploying a commercially supported database will foster greater commercial utilization of space-based data centers,” Fernandez envisaged.
HPE also aims to validate the resilience of space-based applications with HPE Serviceguard, designed to ensure continuous application functionality even in the event of hardware failures.
In the ensuing decade, Spaceborne Computer-3 will journey to the moon.
“After experiencing the cocooned environment of the ISS, venturing to the moon presents an entirely different realm,” Fernandez remarked.
Heat management poses minimal concerns for Spaceborne Computer-3, as the model slated for deployment boasts a certified operating threshold of 55 degrees Celsius, well within the lunar environment’s temperatures, which peak at 50 degrees.
However, the lunar night presents a challenge.
“Unlike the necessity to cool computers on Earth, our priority shifts to maintaining warmth,” Fernandez reflected. “Our engineers are accustomed to grappling with cooling concerns, but the prospect of preserving warmth on the airless lunar surface is a novel endeavor. Overcoming this hurdle will unlock myriad opportunities.”
Conclusion:
HPE’s advancements in space computing, as exemplified by the Spaceborne Computer-2.5, signify a significant leap forward in AI, machine learning, and data processing capabilities in space. This not only expands the horizons of scientific research and exploration but also opens up new opportunities for commercial ventures in space-based computing and data analytics. As technology continues to evolve, the market can anticipate a surge in demand for innovative solutions to support space missions and exploration endeavors.
