Research Byte - Brain networks inspire the next generation of space communication

7 August 2023

JWST enjoys a roughly 1MB/s internet link with Earth when leveraging the massive radio dishes out at Tidbinbilla. The Nancy Grace Roman Telescope will be taking images almost 1GB in size. There’s an obvious problem looming for science, and for anyone who relies on space communication. Laser communication has been growing over roughly the last two decades as a viable alternative to radio communication for alleviating issues of spectral bandwidth and data rates. However, there are a number of reasons why laser communication hasn’t taken off like radio communication and why JWST isn’t enjoying a gigabit laser link back to Earth. While many of these obstacles are slowly being overcome by new technological advances led by newfound momentum in the communications field, one that is here to stay, and that everyone except for radio astronomers can relate to, is clouds.

Laser links to and from space are blocked by all but the thin and icy clouds found in the upper atmosphere, which makes selecting the location for an optical ground station very important. Luckily for Australia, we have some pretty great locations to get away from clouds but even the middle of the desert can’t provide the reliability offered by a radio antenna. This issue is typically addressed by constructing so-called “diverse networks”, i.e. networks that operate many ground stations across different weather and climactic environments to ensure at least one is available. Therefore, when selecting sites for a network, one must not only be careful to select sites that aren’t too cloudy, but also select sites that are climatically independent from one another, such that being cloudy at one site won’t mean that the whole network goes down. 

This is exactly the problem we aimed to address in our latest work, “Availability, outage, and capacity of spatially correlated, Australasian free-space optical networks” led by Marcus Birch & James Beattie. First, we developed an algorithm for selecting sites in the Australasian region that minimised the average cloud cover at any given site, whilst maximising the statistical independence between them. Once we had our network, the next problem we faced was building a statistical model for the network, such that we could answer questions like “what’s the probability of at least one site being unavailable at any given time?” To answer this kind of question, we borrowed some tools from computational neuroscience, treating our network of spatially correlated ground stations like neurons in the brain. If one neuron fires, then it is more likely that other neurons will fire in the same spatial vicinity, which maps perfectly to our problem that if one ground station is cloudy, then other ground stations will be cloudy close-by. Taking this analogy to fruition, we built a model that had tune-able point-wise correlations between ground stations, allowing us to flexibly capture weather events that are correlated across space.

By combining our firing neuron network model with decades of remote sensing data from many different satellites, alongside global orbital simulations (shown in the Figure), we explored how different network configurations and correlations between ground sites influenced the availability statistics of the network. We showed that a realistic network ground station configuration, spread across the Australasian region, may be capable of providing tens of terabits per day to a low Earth orbit satellite, with up to 99.97% availability to geostationary satellites.

With the Mt Stromlo optical ground station nearing completion, our laser communications group is moving onto the next stage of actively demonstrating a network with partners like University of Auckland and University of Western Australia. Our work also did a great job at highlighting desert locations for optimal placement, raising some interesting questions about the atmospheric turbulence at some of these locations. So, in addition to trying to demonstrate a network, we are working to deploy a novel type of turbulence monitor for testing these sites.

Beyond the important scientific outcomes of the study, this project really highlights the potential synergies between industry-focused and theory-focused researchers here at Stromlo, with the solutions we came up with only possible through collaborations between the AITC and the RSAA. We hope this inspires some more cross-pollination between the two institutes.

Marcus Birch & James Beattie

Figure Caption: Low Earth orbital simulations (blue) shown orbiting around our realistic ground station network (green). We combine these orbits, different ground station network configurations, decades of remote sensing data, and our spiking neuron network model to construct detailed availability statistics to inform us about optimal network placement and performance in the Australasian region.