Required Data & Inputs
- Sensor Telemetry: Raw and processed return signals from the 80 GHz Space Debris Radar and monochrome/RGB camera frames.
- Hardware Profiles: Technical specifications, memory limits, and power constraints of the TychoBoB edge-computing and AI/ML processing unit.
- Orbital Environment Data: Simulated and empirical orbital ephemeris data, background noise models, and synthetic trajectories for mm-sized debris classes.
Objectives
- Lightweight ML Architecture Design: Develop computer vision models tailored to execute efficiently within the strict power, thermal, and compute constraints of the TychoBoB edge processor.
- Real-Time Signal & Image Processing: Implement automated filtering and feature extraction pipelines to isolate mm-sized debris signatures.
- Autonomous Event Triggering: Build an on-board event-driven warning system that logs debris encounter metrics and immediately triggers high-priority downlinks or safe-mode maneuvers via deployable antenna systems.
- Hardware-in-the-Loop (HIL) Validation: Test the inference pipeline on physical engineering models of the TychoBoB processor using simulated space debris scenarios to evaluate processing latency, accuracy, and power draw.
This thesis project is supervised by the CENSSS partner Jotne Connect.
Contact persons are Kjell.Bengtsson@jotne.com and aditya.gupta@jotne.com