6G | Space - header illustration

6G | Space

6G | Space: Edge computing

Share this page

woman wearing a hardhat walking across a factory floor with an overlay of cloud, server and edge-computing icons

With a focus on edge computing, we bring computational power closer to the source, reducing latency and enhancing real-time processing capabilities.

space connections 6G - AI training in space diagram

Artificial intelligence training in space

Large-scale deployments of low Earth orbit satellites collect massive amount of Earth imageries and sensor data, but it is increasingly infeasible to download all the high-resolution images and train the corresponding AI models on the ground. In this project, we focus on novel distributed and federated learning frameworks (opens in new tab) that enables ground stations and satellites collaboratively train AI models without downloading all the data to the ground.

Related vertical(s): Space connectivity

Image of Sentinel-2 satellite scanning a section of Earth; Photo: ESA/Astrium

Context-aware compression of satellite imagery in space

Transmitting raw space imagery data to the ground for processing presents difficulties due to limited network bandwidth, resulting in data being captured in restricted modes and taking hours to days to downlink. To address this, compression right in space is a more promising approach to reduce the amount of data transmitted. However, current compression techniques treat all pixels as having equal weight, despite not all parts of an image being equally important. Our proposed solution, Earth+, is a smart filtering and compression method that is implemented directly on the satellite. Earth+ leverages the rich historical dataset on earth to intelligently select a reference image that best represents the near future and uploads it to the satellite. Utilizing lightweight context-aware cloud detection and diff-based comparison, Earth+ identifies only the changed areas and transmits them back to earth, thereby significantly reducing data transmission volume.

Related vertical(s): Space connectivity

space connections 6G - Kodan; illustration of the Earth surrounded by six computers in space

Kodan: Edgifying satellite applications for on-board computation in space

The decreasing costs of deploying space vehicles to low-Earth orbit have led to the emergence of large constellations of satellites. However, the high speeds of the satellites, the large sizes of image data, and the short ground station contacts have created a challenge for data downlink. Orbital edge computing (OEC) can filter data at the space edge and address the downlink bottleneck, but it shifts the challenge to the limited computation capacity onboard satellites. We present Kodan, an OEC system designed to maximize the utility of saturated satellite downlinks while mitigating the constraints of the computational bottleneck. Kodan has two phases: a one-time transformation step that uses a reference implementation of a satellite data analysis app, along with a representative dataset, to produce a set of specialized ML models targeted for deployment to the space edge. After deployment to a target satellite, a runtime system dynamically selects the best specialized model for each data sample to maximize valuable data downlinked within the constraints of the computational bottleneck.

Related vertical(s): Space connectivity