Plan V1-1
Translated version Here
Plan V1-1
Imported from the implementation repo:
~/dev/python/HDDS2/Plans/Plan V1-1.md
Summary
This plan extends the project from radar-only to a unified multimodal drone detection system.
Core Decision
Do not build three independent subsystems.
Build one synchronized system with:
- per-modality inference,
- temporal alignment,
- fusion,
- final tracking and alert logic.
Best Practical Model Choices
Radar
- OFDM range-Doppler pipeline,
- CA-CFAR,
- M/N confirmation,
- Kalman or alpha-beta tracking.
Vision
- YOLO11 + tracking as the practical route,
- RT-DETR as the more research-heavy route.
Audio
- MFCC + LightGBM baseline,
- PANNs or AST embeddings as the stronger upgrade path.
Required Proof
The thesis must show:
multimodal fusion reduces false alarms relative to radar-only detection
