NASA SpaceApps Challenge 2023
Bringing
AI Research
in Space
to everyone
Team Ker-boom
What is the project?
A research platform to develop AI solutions for the space problems
How
does it work?
3) AI learns from simulator experience
An AI algorithm uses the game data and
learns to solve the problem by trial and error!
1) Kerbal Space Program
Create scenarios of interest in this high-fidelity physics simulator.
2) Game Api
Use our version of the API to access the game data (like spacecraft orientation) and to send commands back to the simulator.
Proof-of-concept: Detumbling using AI
We can use the Q-learning algorithm to learn how to detumble a satellite (say, just after deployment or after a micrometeriote hit).
Input: the real-time orientation of the satellite (roll, pitch, yaw) Output: torque values for the reaction wheels.
The possibilities are endless!
next steps
Compare the AI solution with classic detumbling methods to assess its potential
(criteria: time taken to detumble, amount of energy/propellant used)
1
Learn AI solutions for other space problems
(orient a satellite’s solar panels to always face the sun, find interesting rocks to zap on Mars, and so on!)
2