Build the autonomy that keeps deep-space missions alive

When a spacecraft is 20 light-minutes from Earth, nobody on the ground can fix a problem in time. Signals arrive buried in noise, carriers drift, and a single failed CO₂ scrubber can cascade into a habitat emergency before mission control even sees the alert.

At Matlab in Space Hackathon, you'll build the software that handles these problems on its own, using real NASA and ESA spacecraft data.

Tracks

Track 1: Deep-Space Communication & Signal Intelligence
Build agents that find, clean, and decode weak radio signals with no human tuning. Think adaptive filtering, re-acquiring a signal after dropouts and frequency hops, and turning raw bits into readable telemetry.

Track 2: Energy, Power Grid & Closed-Loop Life Support
Keep a lunar or Martian habitat alive. Detect anomalies in real spacecraft telemetry, predict battery failure before it happens, and design power systems that shed load intelligently when things break.

What makes it special

  • Real mission data: telemetry from NASA's SMAP satellite and Curiosity rover, NASA battery aging data, and ESA's satellite anomaly benchmark
  • Real problems: both tracks come from challenges space agencies are actively working on
  • Open to all skill levels: signal processing, ML, controls, or just curiosity about space

Requirements

What to build

Build a working software prototype for one of the two tracks. It should run with as little manual configuration as possible. Autonomy is the point.

  • Track 1: A pipeline that takes noisy radio data and automatically filters it, acquires or re-acquires the signal, and outputs decoded telemetry.
  • Track 2: A system that monitors habitat or spacecraft telemetry and does at least one of the following: detects anomalies, forecasts time-to-failure, or makes autonomous power allocation decisions under failures.

What to submit

  1. Project description: the problem you tackled, your approach, and which dataset(s) you used
  2. Code repository (required): a public GitHub link. Judges will primarily review your repo, so the README should explain the project in detail:
    • Problem and approach
    • Datasets used
    • How to run it
    • Results with plots or images
    • Limitations and next steps
  3. Results: your key outputs and metrics, included in the README or attached as images
    • Track 1: telemetry log plus before/after spectral plots
    • Track 2: detection accuracy, forecast error, or a failure-scenario simulation, depending on what you built
  4. Slide deck : a short deck with your key plots and images, uploaded to Devpost or linked in the repo
  5. Track selection: tell us which track you're entering

Hackathon Sponsors

Prizes

1 non-cash prize
Gift Card
1 winner

$100 for the team
Matlab Merch

Devpost Achievements

Submitting to this hackathon could earn you:

Judges

Jon Loftin

Jon Loftin

Neha Sardesai

Neha Sardesai

Neha Sardesai Gmail

Judging Criteria

  • Autonomy (30%)
    How well your system runs and adapts on its own. Top scores handle signal drift, dropouts, or component failures with no human input. Ties are broken by this score.
  • Technical Approach (25%)
    How well your methods fit the problem. We're looking for working solutions with clear reasoning behind your design choices.
  • Results & Evidence (25%)
    Proof that it works on real mission data. Show metrics, before/after plots, and how your system handles failure cases.
  • Communication (20%)
    How clearly you present your work. A strong 5-minute demo and a README that lets anyone reproduce your results will score highest.

Questions? Email the hackathon manager

Tell your friends

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.