IBM may no longer have the relevance it used to have in personal and corporate computing. SHowever, it continues to be a leader in research on how to apply information technology to scientific uses.
In this post we are going to talk about two open source projects; Space Situational Awareness and Kubesat. These projects seek to improve communication between satellites and help predict the trajectory of space debris. The latter is to prevent them from colliding with the satellites and leaving them idle.
By making these two projects open source, the company seeks to give more people access to space technology and democratize access to space for all.
Open source and space. The two IBM projects
Space Situational Awareness (SSA)
Tens of thousands of human-made space objects (ASOs) orbit the Earth. To this must be added those that will be generated as a result of the entry of private companies and other countries such as Israel or India to the space race.
ASOs range in size from specks of paint a few centimeters in diameter to the International Space Station. A small piece of paint may not be dangerous on our planet, but when it travels at speeds of about 8.000 meters per second, a collision with a satellite or other manned vehicle can cause disaster.
The project, developed by the IBM Space Technology Team, allows you to determine where the ASOs are (orbit establishment) and where they will be in the future (orbit prediction). Those responsible had the advice of Dr. Moriba Jah of the University of Texas at Austin.
The most advanced methods for orbit prediction are based on physical models. To be successful, these models require extremely accurate data about the ASO and its environment.
The problem to be solved is that the location data obtained on the location of the ASOs by the ground sensors is infrequent and with interference. It is also not fully known how phenomena such as space weather and atmospheric density affect ASOs.
The project seeks use machine learning methods, not to predict orbits but to create models that learn when physical models incorrectly predict the future location of an ASO. If you tried to build a model that predicted all the orbital dynamics of an ASO, you would need a lot of data to learn orbital mechanics. Physical models know a lot about orbital dynamics, so it is only necessary to learn when and why they are wrong.
KubeSat
As an alternative to expensive traditional satellites, many small businesses and research groups plan to use swarms of satellites (groups of satellites that work together to perform specific tasks). In order for the swarm to be created and controlled, you need a software tool, as well as additional support software for your specific objectives.
KubeSat es an autonomous cognitive framework designed for swarms of cube satellites that allows the simulation and optimization of multisatellite communications.
The project allows to simulate the precise orbital mechanics for each object through orekit. It then uses these calculations to put restrictions on communications between the satellites, the ground station, and the ground sensors.
On the other hand, it incorporates the NATS messaging services; optimizing your communications through reinforcement learning.
This framework follows the principles of swarm intelligence such as awareness, autonomy, solidarity, expandability and resistance.
IBM hopes that by making the project open source it will democratize the satellite swarm industry and allow smaller operators to take advantage of space technology.
Using it, reinforcement learning models can be built for specific use cases. Collaboration and communications between those swarms of satellites can be autonomous, allowing the swarms to integrate or separate as needed.
The code for the Orekit application, the base library for orbital mechanics, also opens. These extensions are used to simulate communications between satellites, between satellites and ground stations, and between satellite sensors and the ground, through a NATS message platform.