TelePIX has announced the completion of the in-space demonstration of TetraPLEX, an artificial intelligence (AI) processor for satellites launched in August. This marks the world’s first successful demonstration of real-time, high-speed parallel processing of satellite big data AI model in the extreme conditions of space.
TetraPLEX, developed by TelePIX, is an on-board processor that enables AI processing and edge computing directly on the satellite without the need to transmit information collected from space to a ground station. TetraPLEX has the advantages of reducing cost and time, reducing data loss.
TetraPLEX, launched using SpaceX Falcon 9 rocket in August, reached an altitude of approximately 597 kilometers. It conducted a series of in-orbit performance tests for its satellite image preprocessing and attitude control estimation algorithms, both of which performed successfully and stably in the space environment. The satellite image preprocessing was completed in just 11 seconds, approximately 35 times faster than it would have under a conventional FPGA (Field Programmable Gate Array) board.
The demonstration confirmed that all three data processing boards of TetraPLEX were functioning normally. The three boards are APB (Application Processor Board) containing the satellite image preprocessing and satellite attitude control estimation algorithms, the DIB (Data Interface Board) verifying the sensor data interface, and the IPB (Intelligent Processing Board) responsible for AI processing based on a commercial GPU (graphics processing unit) platform.
The primary focus of the demonstration is the operation of the satellite image preprocessing algorithm. TetraPLEX development team, led by Dongshik Won, the director of Future Innovation Research Team, confirmed that the images were processed at the same level of quality as on the ground, despite the harsh space environment of radiation, thermal shock and vacuum. The computational results of the logic on board the TetraPLEX matched the results on the ground with 100% accuracy. This breakthrough is expected to influence satellite image preprocessing without the need for extensive ground station infrastructure, allowing for various satellite image preprocessing algorithms to be used in space.
The attitude control estimation algorithm was also performed normally. In order to operate satellites stably, the complex sensor data processing operations required for attitude control and determination must be performed quickly and accurately. TelePIX achieved these results by equipping TetraPLEX with a Kalman filter algorithm (a statistical technique that filters out noise from reality data to calculate accurate numbers) for satellite attitude estimation.
TelePIX will conduct a final confirmation of the stable operation of TetraPLEX in space over the next four months. During this time, the TetraPLEX will be evaluated for the possibility of unexpected failure due to radiation buildup in the space environment. The data gathered from this demonstration will be used to further advance the reliability and sustainability of the next-generation model.
“Enhancing the efficiency of space missions through high-performance on-board processors is just the beginning,” said Seonghui Kim, the CTO of TelePIX. “Building on the success of this demonstration, we plan to implement more complex and advanced algorithms on satellites. This technological advancement will maximise the success rate of space missions and lead the future of satellite data processing.”
TelePIX is a Korean satellite specialised company with technologies for the entire satellite industry cycle, from satellite payloads to satellite data processing and value-added solutions. The company has earned global acclaim, including winning the CES 2024 Innovation Award and being selected as a World Economic Forum (WEF) Technology Pioneer 2024. TelePIX recently showcased its innovations, including TetraPLEX, at International Astronautical Congress (IAC) 2024, the world’s largest space congress, where TelePIX had an exclusive pavilion as an official sponsor.
Comment on this article via X: @IoTNow_ and visit our homepage IoT Now
