The Discovery Campaign on OPS-SAT experiments

Below is the list of projects implemented through the Open Space Innovation Platform (OSIP) Campaign 'OPS-SAT experiments'.

Onboard Multi-Frame Super Resolution Image (OHB Hellas)

The small size of CubeSats often prevents them from obtaining high-quality images for use in Earth observation. This study will use OPS-SAT's powerful hardware and software to run demanding deep learning algorithms that will substantially increase the quality of images acquired with OPS-SAT's camera. By effectively fusing short sequences of low-quality images together, 'multi-frame image super resolution' can be achieved beyond the maximum resolution of the imaging hardware.

Angle-based Multiple Cross-Correlation (AbC) (Observatoire de Paris – PSL)

Whether pinpointing the location of a satellite in distress, or tracking asteroids in bright regions of the sky, nanosatellites will rely on optimised astrometry for their autonomous navigation. This study will develop an algorithm on board OPS-SAT to dramatically enhance the accuracy of optical measurements made with CubeSat hardware. OPS-SAT offers a wide range of different geometry, aberration and straylight conditions to test the algorithm.

AI at the edge: DeepCube service IOD/IOV (AGENIUM Space)

This study seeks to use OPS-SAT to demonstrate the DeepCube deep learning service in orbit. Use-cases include the detection of deforestation and ships in satellite images. Deforestation is a powerful demonstration of the service as change detection requires analysing individual pixels on board the satellite and storing and handling reference data. Solutions developed for the deforestation use case in a previous project will be validated here in a real environment in orbit.

Data Compression as a Service (VisionSpace Technologies GmbH)

Telemetry data is a valuable asset produced by all spacecraft. Compressing data generated in space can increase the mission science throughput of large ESA missions and increase the downlink rates of smaller missions. This study will develop the CompressionCache, a technology inspired by the computer cache that can compress any given data according to multiple criteria simultaneously and automate the management of which criteria are applied to each type of data and will be offered as a service to other OPS-SAT experimenters.

SaaSy ML: Onboard Machine Learning Software for Experimenters (VisionSpace Technologies GmbH)

SaaSy ML will provide other experimenters on OPS-SAT with onboard Machine Learning (ML) functionalities that applications can subscribe to via a Software as a Service (SaaS) app hosted on board. The ML features provided by the SaaS app will cover both training and prediction. SaaSy ML's service-oriented approach will save other experimenters the effort of implementing their own data provisioning and ML solutions and allow them to focus on the objectives of their experiment.

Deep Active Tracking: an AI-based system for an active tracking of Earth features from OPS-SAT (Adática engineering S.L.)

In this study, an artificial intelligence ‘agent’ will be taught how to control OPS-SAT and actively track features on Earth. The Deep Active Tracking (DAT) system will have two core AI-based algorithms, one that processes the information from images taken by OPS-SAT’s optical camera to detect selected features, and a second with control over the satellite’s reaction wheels that decides the optimal sequence of actions to place and keep the feature centred in the images.

Versatile data compression software for sustained high-throughput in-orbit data acquisition (DAPCOM Data Services S.L.)

In this study, new data compression algorithms will be developed, as will their software implementations for handling radiofrequency data and multi-band images. The team will identify the limitations of existing data compression frameworks in use, investigate existing solutions and design new algorithms where necessary. The end goal of the activity is to allow for real-time, in-orbit compression of demanding data sources, such as long radiofrequency acquisitions or video.

Smallsat DTN (Hellenic Aerospace Industry)

As space missions increasingly involve communications between multiple mission segments, space agencies and private actors, the need to manage and transfer large volumes of data is increasing. There are reliable and interoperable standards emerging for this purpose, and in this activity, the team will develop and adapt the necessary software for use on OPS-SAT's satellite and ground segment to demonstrate automation in spacecraft control and file transfer using these standards.

Hybrid Online Policy Adaptation Strategy (HOPAS) (Airbus Defence and Space Ltd.)

The systems that control how a spacecraft orientates itself in space are often built to be very robust, as knowledge of the exact stresses they will face in space is not well known during the design phase. But this can limit the satellite's pointing performance. The Hybrid Online Policy Adaptation Strategy (HOPAS) team from Airbus uses data and artificial intelligence to continuously improve pointing performance and surpass these limitations. It will be applied on OPS-SAT for testing on real flight data.

A Comparative Study of High-Level and Low-Level Implementations of Deep Learning Models for Spacecraft (Mission Control Space Services)

Not so long ago, updating the software on a spacecraft after launch was a nail-biting affair. Today, software updates are a key tool for recovering a faltering satellite. Reconfiguring hardware in flight is the new frontier and an exciting area of opportunity for innovation. OPS-SAT hosts powerful Field Programmable Gate Arrays (FPGA) and in this study, the team will use it to test new ways of deploying artificial intelligence never before flown on a spacecraft.

Satellite Identification and Localisation experiment (Libre Space Foundation)

Satellite Identification and Localization (SIDLOC) is a method of identifying and tracking satellites in low-Earth orbit (LEO). Even though SIDLOC operates using the downlink, a feature that is not available in OPS-SAT's software defined radio, this study till retrieve vital information by performing the situation in reverse, where OPS-SAT functions as the ground station and one or more ground stations transmit SIDLOC signals as if they were satellites.

Onboard anomaly detection from the OPS-SAT telemetry using deep learning (KP Labs)

Detecting anomalies in satellite telemetry is critical for safe operation. To detect types of anomalies beyond out-of-limit checks, this study will use data-driven deep learning-powered models, building upon current advances in the field and deploying them on OPS-SAT. The main objective is to demonstrate that deep learning can be effectively deployed on-board OPS-SAT for the automated detection of anomalous events using its operational telemetry data.