Onboard Data Processing
Technology Development
Onboard data processing allows satellites to compress and analyse increasing volumes of data in orbit, and is used to reduce the volume that must be downlinked to Earth. It can enable faster delivery of insights, improve spacecraft operations and communications efficiency, and facilitate timely responses for applications such as disaster monitoring and defence.
Advancements in Earth observation (EO) have increased the complexity of satellite systems, data products, and user needs. Developments in the broader fields of data science, Artificial Intelligence (AI), and autonomous tasking and spacecraft control are further increasing user expectations, and driving the need for improved in-orbit data processing and decision-making performance. These developments mean satellites require higher data throughput, faster downlink speeds, and increased computing power to address growing complexity and demands, and enable a number of specific benefits. 2)
Reduced data volumes transmitted. The volume of data generated each day by satellites is increasing as the performance of instruments spatially, spectrally, and temporally advances. Onboard data processing capacity is used to compress or pre-process data, which reduces the volume to be transmitted, helping to mitigate bandwidth constraints and lessens the load on downlink services and infrastructure.
Delivery of actionable insights. By making use of onboard processing and algorithms, data can be analysed in situ which can improve operational efficiency and accelerate the delivery of actionable insights, especially for time critical applications. 2) 3) 4)
Real-time AI data analysis. Onboard data processing includes the use of AI inference which can enable the autonomous parsing and selection of data, and prioritise data downlinking. This can provide accelerated response times, and pre-processed analysis ready data products from space. 3)
Onboard Processing Techniques
Compression
Algorithms, such as Block Adaptive Quantisation (BAQ) and Performance-Optimisation BAQ (using AI), can selectively compress data up to ~100x without losing relevant information. This allows operators to downlink compressed and refined data, reducing bandwidth and latency.
Image Selection and Tasking
Onboard image selection facilitates the removal of unusable data, such as images outside the area of interest or those taken in low visibility conditions. Dense environments such as cities can be imaged more intensively relative to less populated areas. This can increase the information richness of data acquired and downlinked. 2) 3) 4)
Preliminary Analysis
Processing techniques such as filtering and feature extraction allow for preliminary analysis to occur onboard the spacecraft. Feature extraction identifies and isolates significant features in the data such as edges, textures, or specific patterns for applications such as object recognition, land cover classification, and change detection. Preliminary analysis can include basic image formation, geolocation tagging, and area of interest (AOI) identification. AOIs are selected based on mission specific parameters or real time data analysis, allowing the satellite to focus on imaging relevant regions, easing storage demands and enhancing transmission efficiency. This AOI identification is increasingly performed using real time decision making algorithms powered by AI and machine learning (ML) techniques. 4)
Edge Computing
Edge computing enhances traditional cloud computing by bringing processing and storage closer to observation. This enables faster processing and reduced latency through localised ‘edge’ cloud servers. In Space-Air-Ground Integrated Networks, combining satellite and ground-based edge processing with AI/ML enables efficient, real-time data handling. This is valuable for latency-sensitive applications like disaster response, as well as tasks such as automatic target detection used in navigation and space situational awareness. Ground based edge servers near ground stations further refine satellite preprocessed data before cloud transmission, speeding up analysis and decision making. 2) 4)
Onboard Processing Applications
Through enhanced real time onboard processing capabilities, satellite systems can deliver critical data for applications in environmental monitoring, disaster management, and global security with improved speed and efficiency. 2) 4)
Disaster Monitoring
The reduced latency enabled by onboard processing and compression, enhanced by AI, can be critical for timely disaster response. The Ciseres project, of ESA’s Civil Security from Space programme, aims to enhance natural disaster detection and response by using onboard satellite image processing and predictive AI. The International Charter for Space and Major Disasters allows predefined disaster risk management authorities to mobilise satellites in the wake of disasters to efficiently transmit compressed data. 4) 6)
The Italian-Argentine Satellite System for Emergency Management includes the Argentine National Commission for Space Activities’ (CONAE) SAOCOM satellites and the Italian COSMO-SkyMed constellation. The satellites utilise onboard processing to provide near real time data for emergency response teams. 4)
Cloud Detection and Removal
Onboard processing can identify and remove cloud contaminated pixels before data is downlinked to ground stations. For optical satellite instruments, cloud detection and atmospheric correction are valuable preprocessing steps for applications such as change detection, and vegetation and water monitoring. For example, the Data Transformations and Auto-Calibration Sentinel-2 Network (DTACSNet) is a specialised atmospheric correction processor that was trained on Sentinel-2 imagery to provide cloud mask and estimated surface reflectance for the production of analysis ready data. 7)
Ship or Land Based Vehicle Detection
Situational awareness of ships and land vehicles, including their detection, localisation, and tracking, is crucial for safe navigation and the prevention of accidents. Optical and radar remote sensing systems, often used in combination with Automatic Identification Systems (AIS) and in-situ radar networks are useful for this application, but generate large volumes of data. Onboard processing can reduce the amount of data transmission by outputting parametric information on vehicles (e.g. location, heading, etc.). 8) 9)
Space Situational Awareness
Onboard processing for space situational awareness focuses on in-space object tracking and surveillance. This awareness is essential in mitigating space-to-space collision risk, and is expected to grow more important with the projected increase in the number of objects in orbit. Without improved automation, manual tracking and hazard resolution methods risk becoming saturated and ineffective. 5)
Related Missions
PhiSat-1 & -2 Nanosatellite Mission
ESA’s Federated Satellite System (FSSCat) mission, consisting of PhiSat-1, launched in September 2020, and PhiSat-2, launched in August 2024, was the first EO mission to use AI onboard. It demonstrated improvements in transmission efficiency provided by the application of AI processing for cloud screening to imagery from its HyperScout-1 payload. This allows the downlinking of only select data based on cloud coverage.
The PROBA (Project for On-Board Autonomy) Missions
ESA’s Project for On-board Autonomy (PROBA) missions were controlled by an automated Mission Control System and a dedicated tracking terminal located at ESA’s European Space Operations Centre (ESOC) in Germany. GNSS position determination and onboard orbit propagation allowed for autonomous navigation.
The Advanced Data and Power Management Systems (ADPMS) onboard the Proba-3 satellites, launched in December 2024, has an onboard control algorithm that provides navigational support to maintain the high-precision, millimetre-scale flying formation of the dual satellites. PROBA-V (Project for On-Board Autonomy - Vegetation), launched in May 2013 and decommissioned in October 2021, carried an advanced ADPMS onboard computer that used ESA's LEON2-FT microprocessor to manage its autonomous operations and data storage, processing, and downlinking.
PROBA-3 (Project for On-Board Autonomy-3)
PROBA-V (Project for On-Board Autonomy - Vegetation)
OPS-SAT - Operations nanoSatellite
ESA’s OPS-SAT, launched in December 2019, was the first satellite mission dedicated to testing and validating new techniques in mission control and onboard satellite systems. OPS-SAT carried the GomSpace NanoMind A712D, a powerful onboard computer with a ten-fold increase in power over typical ESA spacecraft. The NanoMind gathered telemetry from the various onboard systems enabling in orbit testing of new software and control techniques for future missions.
Planet Pelican
Planet Labs Pelican is a commercial constellation of high-resolution imaging satellites that leverages AI for on-orbit data processing and analysis. The satellites are equipped with an NVIDIA AI chip allowing use of the NVIDIA Jetson edge AI platform. The AI platform provides edge computing for in orbit data processing which allows analysis-ready data to be delivered, enabling near real-time insights for applications in defence, government, and commercial sectors.
EO-1 (Earth Observing-1)
Launched in November 2000, the EO-1 (Earth Observing-1) spacecraft was a technology demonstration mission of NASA’s Goddard Space Flight Centre (GSFC). The ASE (Autonomous Sciencecraft Experiment) software package was installed on EO-1, utilising onboard processing and decision making to capture follow-up images of dynamic events. ASE includes three autonomy software components: CASPER (Continuous Activity Scheduling Planning Execution and Replanning) for replanning activities, SCL (Spacecraft Command Language) package to enable event-driven processing, and onboard science processing algorithms that detect trigger conditions including cloud detection for onboard image masking.
ICEYE
The commercial EO company ICEYE Ltd. of Espoo, Finland, is actively developing and implementing AI-based techniques for onboard processing of SAR data. This is crucial for handling the vast amounts of data generated by the ICEYE X-band SAR microsatellites, the first of which, ICEYE-X1, was launched in January 2018. This reduces the unnecessary transmission of data to ground stations while ensuring that the quality of the X-band SAR data is maintained.
ICEYE Microsatellites Constellation
ICEYE-X1 (SAR Microsatellite-X1)
D-Orbit
D-Orbit, an Italian space logistics and transportation company, provides various Space Cloud Services for onboard processing. The OBC aboard the ION Satellite Carrier, D-Orbit’s proprietary orbital transfer vehicle, integrates edge computing and AI acceleration. Onboard processing of EO imagery and signal intelligence data reduces latency and enables faster validation for time-sensitive use cases. Intelligence Logistics is an orbital infrastructure designed to support processing. It allows data from multiple satellites to be received, processed directly onboard, and routed to the appropriate destination according to mission requirements.
References
1) ESA, “Onboard Computers,” URL: https://blog.satsearch.co/2022-03-23-on-board-data-processing-with-unibap
2) H. Curtis, “Spotlight: on-board data processing for next-generation satellites – with Unibap,” 23 March 2022, URL: https://blog.satsearch.co/2022-03-23-on-board-data-processing-with-unibap
3) KP LABS, “What is on-board data processing?,” January 28, 2025, URL: https://www.kplabs.space/news/what-is-on-board-data-processing
4) Garcia et al., “Advancements in Onboard Processing of Synthetic Aperture Radar (SAR) Data: Enhancing Efficiency and Real-Time Capabilities,” 3 June 2024, URL: https://ieeexplore.ieee.org/document/10547107
5) Lim et al., “Onboard Artificial Intelligence for Space Situational Awareness with Low-Power GPUs,” 2020, URL: https://amostech.com/TechnicalPapers/2020/Poster/Lim.pdf
6) CSC, “ESA to leverage AI-enabled satellite for disaster response,” 15 October 2024, URL: https://connectivity.esa.int/news/esa-leverage-aienabled-satellite-disaster-response
7) Aybar et al., “Onboard cloud detection and atmospheric correction with efficient deep learning models,” 15 October 2024, URL: https://ieeexplore.ieee.org/document/10716772
8) Lee et al., “Detection and tracking for the awareness of surroundings of a ship based on deep learning,” 23 September 2021, URL: https://academic.oup.com/jcde/article/8/5/1407/6374687
9) Ghosh et al., “On-Board Ship Detection for Medium Resolution Optical Sensors,” 28 April 2021, URL: https://www.mdpi.com/1424-8220/21/9/3062
10) Coursera, “What Are the Differences Between Machine Learning and AI?” 30 June 2025, URL: https://www.coursera.org/articles/machine-learning-vs-ai