Deep Dive: Satellite Swarm Technology
When a natural disaster breaks out, often the first few minutes after the catastrophic event determine the nearby future. Sometimes emergency teams react well to the case and are on the charge immediately, other times, maybe not. Satellites already try to contribute in a positive way, but it still is not always enough. They sometimes act as glorified digital cameras, that take the image of the disaster, but hours after the impact is severely felt, they send down the data in raw files. An example is a wildfire. By the time the fire has moved miles from its origin, first responders only then receive the satellite data. The solution to this problem is not getting bigger cameras, but putting artificial intelligence (AI) into orbit and letting satellite constellations communicate like a synchronised swarm.
Firstly, what is swarm AI or swarm intelligence? It is where many simple, leaderless individuals interact using basic rules to create a complex, intelligent group behaviour. This was all inspired by nature in the form of bees, ants and fish. Swarm intelligence works when an individual in a group immediately identifies its surroundings and communicates insights with its closest neighbours, adjusting its actions accordingly. Some key benefits of this is resilience. Say one drone breaks down or loses connection; the rest of the swarm continues on and keeps the system running. Another benefit is scalability. The more units you add to swarm intelligence, the more capable the swarm becomes, making it easier to achieve more difficult tasks. The last benefit is the ability to dynamically problem solve. By working together as one unit, swarm intelligence becomes excellent for tasks such as multi-step reasoning breaking down larger problems between the swarm. Edge AI, by definition, is the practice or running AI algorithms and machine learning on physical hardware such as smartphones, cameras or autonomous vehicles. It works when a device collects data by its own sensors and also immediately processes the data by its own microchip. Some benefits to this is that crucially, there is ultra - low latency, which is ideal in this case for satellites regarding natural disasters. It also works well in autonomous cars in terms of crash avoidance. Another benefit is that these devices can work when they are offline as well. Finally, there is also enhanced privacy. As the data collected on the Edge AI device is not shared with anything, it stays private and sensitive and personal data can be secure.
Coming back to the primary problem, equipping LEO (Low Earth Orbit) satellite constellations with Neural Processing units, specialised microchips that are designed to enhance artificial intelligence and machine learning tasks, allows running the computer models directly in orbit. The satellite would then be able to filter out clouds, isolate targets and beam immediate alerts to ground teams via inter-satellite laser networks. This would speed up the process and take out the latency factor as a whole and could potentially save lives from these natural disasters.
To help understand the maths and physics of swarm networks, we need to look at a few formulas, the first of which is the latency equation. The latency equations tells the total latency (Ttotal) by combining the transmission delay across distance (d) and processing time (Tprocessing). The distance has to be divided by the speed of light in a vacuum (c) to find the total travel time as the laser beams are at the speed of light in the vacuum of space. The formula looks like so:

The next piece of maths or physics is optical mesh network bandwidth. Inter - satellite links operate with infrared lasers with 1550 nanometres wavelength in the vacuum of space, relaying data between satellites at the speed of light (c). Why 1550? This is because of the wavelength preventing scattering of air molecules on Earth. In addition, 1550nm can easily be absorbed by water, and thus, the eye having a layer of water, is still protected. Below is the equation:

The final and arguably most complex physics formula is the Doppler Shift formula. In satellite networks, the Doppler Shift formula applies wherever there is relative motion between a transmitter and a receiver. It calculates the change in frequency of a wave caused by relative motion between the source and the observer. The formula is below:

The formula may seem very complicated at the start, but by breaking it down, it is easier to understand. First, we need to identify the base signal, which in this case is the original frequency sent out by the transmitting satellite. Think of f0 as a constant rhythm. The satellite fires an infrared laser beam at a constant rate across space. Looking at the next part of the equation: (Vrealtive/c) is the relative speed between two satellites divided by the speed of light. The ratio measures how fast the spacecrafts are moving compared to how fast light travels. Because LEO satellites travel at speeds of 7.8km/s, this fraction reveals what percentage of the Lightwave gets physically compressed or stretched per second. The final step is combining all of this together to find the shift ( Δf ). This is the actual Doppler Shift and is found by multiplying the base signal by the speed ratio to give you the exact number of Hertz the receiving satellite must adjust to in its tuning dial to stay connected and help communication in the swarm. An example of the Doppler shift formula on the ground is by imagining a machine throwing tennis balls at you once every second. If you stand still, you can catch one ball per second, however, if you start running toward the machine, that catch every second becomes a catch every half a second and decreases until you reach the machine itself. This means Δf goes up, opposite if you run away, the gap between the catches grow longer Δf goes down. In orbit, satellites are like those runners, just at 7.8km/s, compressing or stretching the laser light so quickly the receiver must retune their systems. This all has to be done in real time to prevent dropped connections between the swarm.
Therefore, a way of explaining the difference of capabilities between a swarm of satellite and a traditional one can be by saying a single satellite is like a stationary lighthouse, where it is functional, but unable to adapt unless if a human physically has to come and change the beams. However, a swarm of satellites are like a murmuration of starlings, when one bird detects a hawk or obstacle, the information ripples through the pack causing the entire group to shift shape and react instantly without waiting for instructions from a so called ‘leader’.
Some real swarm technologies already in use are: Radiation Hardened Neural Processing Units, which are onboard AI accelerators running neural networks to classify land features, fires or ships in real time, something that will be desperately needed in the future especially with El Nino coming around the corner. Another swarm technology is inter-satellite lasers (ISLs). These are low power laser transmitters and receivers creating self-healing mesh networks across hundreds of CubeSats in orbit. Finally, there is Autonomous Orbit Control which is algorithms maintaining precise spacing using different drag systems or micro thrusters, all without any intervention at all from the ground.
This sets the future for satellites and swarm networks in space, with autonomous disaster detection an idea for the future along with the NPUs in the satellites to potentially send messages to people residing in areas where natural disasters may just have broken out or about to break out.


Comments