← Projects

CAESAR

Combat aircraft engagement and strategic AI response

Period

Mar 2024 – Mar 2026

Affiliation

Defence R&D Organisation (DRDO) — collaboration

• Electronic warfare
• Sensor fusion
• Real-time ML

A collaboration with the Defence R&D Organisation on the machine-learning side of electronic support measures. What follows is the requirement set the work was scoped against; the implementation is not described here.

Operational requirements

Two broad requirements drive everything else:

1. Detect enemy radars as early as possible.
2. Deny them a detection — jam them, or otherwise avoid being seen.

Derived requirements

The broad pair decomposes into a chain, each stage feeding the next:

1. Detection of radar pulses out of a noisy mixed-signal environment.
2. Clustering of those pulses, separating interleaved sources.
3. Track generation, deriving further parameters from each cluster.
4. Classification of the radar behind each track.
5. Localisation of that radar, accurately enough to act on.
6. Jamming, by noise or deception, once a threat is identified.

Engineering constraints

• Real time. The pipeline runs on an embedded system under an RTOS — a correct answer that arrives late is a wrong answer.
• SWaP. Size, weight and power are fixed by the airframe, which bounds how much computation the models may ask for.
• Certification. Airborne software follows DO-178C, so the design has to stay auditable rather than merely accurate.

Workstreams

• Signal clustering over emitters
• Emitter classification
• Multi-sensor data fusion (MSDF)
• Target detection and tracking
• Voice analysis
• A pilot-facing dashboard presenting the fused picture