DRISHTI—AKSHI
An indigenous, AI-fused, cost-matched counter-UAS platform for India's $1,000 drone problem.
DETECT · IDENTIFY · TRACK · NEUTRALIZE
An Indian answer to the $1,000 drone problem.
THE OPPORTUNITY · Operation Sindoor (May 2025) exposed a 200× cost asymmetry — a $1,000 drone defeating $200,000-class interceptors. The IAF SWAC RFI explicitly demands a launch-and-forget answer. India has roughly 24 months to deploy it.
THE SOLUTION · DRISHTI-AKSHI fuses four sensing modalities through an indigenous transformer-based AI pipeline running on edge compute, with a six-tier effector ladder that selects the cheapest sufficient response.
THE ASK · ₹3–4 Cr non-dilutive grant for Phase 1. Full programme ₹210–335 Cr through induction. 10-year revenue potential >₹2,500 Cr including export.
Self-reliant C-UAS. Built to win the economics.
Multi-sensor fusion
Radar, RF, EO/IR and acoustic fused at feature and decision level. No single blind spot.
Edge-autonomous AI
Each node operates locally when comms degrade. Mesh-cooperative when uplinks are healthy.
Cost-matched kill chain
Six-tier effector ladder. Median engagement ₹500 to ₹2 lakh — not ₹4.2 Cr.
Friend / Foe / Neutral
DigitalSky + RF fingerprint + behavioural model. Friendly UAVs stay alive.
Indian-context AI
Models trained on subcontinental birds, DGCA platforms, local terrain. The data is the moat.
Four modalities.
One decision in under 3 seconds.
Sensor diversity beats sensor sophistication. Every node fuses X-band radar, passive RF, EO/IR and acoustic at feature and decision level — eliminating the single-sensor blind spots adversaries exploit.
Robust AI by design,
not by retrofit.
Game-theoretical defences today. Quantum-enhanced equilibria for the Phase 5+ adversary. Aligned with India's National Quantum Mission (₹6,003 Cr, 2023–2031).
Adversarial oracle
Worst-case attack simulation against every sensor pipeline — co-evolved with the defender model in a min-max loop.
Hybrid quantum solver
QAOA / VQE on NISQ devices to compute Nash-equilibrium defender policies as quantum hardware matures.
Certified robustness
≥70% AI capability under 100 W/m² broadband jamming with provable ε-adversarial-radius bounds.
From PGD-hardened CNNs to post-quantum mesh comms.
PGD, AutoAttack and FGSM augmentation across radar, RF and EO/IR feature spaces. Hardens models before any quantum layer engages.
Stackelberg defender–attacker formulation solved with mirror-descent. Yields min-max policies that survive the worst rational adversary.
NISQ-era QAOA / VQE on 20–127 qubit backends to approximate Nash equilibria in defender policy space. Quantum hardware abstracted behind a classical fallback.
Post-quantum KEM (ML-KEM / Kyber) on mesh uplinks. Aligned with National Quantum Mission Thrust Area on quantum communication.
Six classes. One strategic threat.
The economic gap is the threat. A 16-drone saturation wave costs an adversary ~$16,000 — and bankrupts any defender forced to answer it with legacy SAMs. DRISHTI-AKSHI closes the ratio with a cost-matched effector ladder.
Detect to engage in six fused steps.
Sensors → Fusion → AI → Effector mesh.
Heterogeneous sensors fielded at every node. Time-synchronised via PTP, geo-referenced via INS.
- ◆Node-level autonomy when uplink degrades
- ◆Mesh-cooperative fusion when uplinks healthy
- ◆Akashteer / IACCS C2 northbound integration
- ◆Air-gapped install option for classified sites
Operator view, live.
| ID | CLASS | TYPE | BRG | RNG (m) | SPD | THREAT | EFFECTOR | STATUS |
|---|---|---|---|---|---|---|---|---|
| TRK-1000 | DJI Mavic 3 | FOE | 012° | 1800 | 18 | 88 | T1 · RF spoof | CLASSIFY |
| TRK-1001 | Autel Evo II | FOE | 047° | 2220 | 24 | 88 | T2 · GNSS deny | CLASSIFY |
| TRK-1002 | Custom FPV | FOE | 088° | 2640 | 30 | 88 | T3 · RF burst | CLASSIFY |
| TRK-1003 | Heron-MK2 | FRIEND | 134° | 3060 | 36 | 15 | — | NEUTRAL |
| TRK-1004 | Bird flock | NEUTRAL | 176° | 3480 | 42 | 15 | — | TRACK |
| TRK-1005 | Unknown wing | UNKNOWN | 210° | 3900 | 48 | 55 | — | TRACK |
Subsystem specs. Issued on identification.
Datasheets are issued to verified defence, government and integrator contacts only. Submissions are logged and reviewed by the ParaMedha programme office.
Concept to induction in 30 months.
Every milestone is TRL-anchored and audited.
Indigenous IP. Indian consortium.
ParaMedha leads as prime integrator and AI/fusion IP owner. Consortium partners include BEL, SAMEER, Tonbo Imaging, IIT Madras CAI, Bharat Dynamics and DRDO labs.
PhD, UTS Sydney · Assistant Professor, BITS Pilani. Three Springer monographs, ten peer-reviewed papers on adversarial machine learning.
Detect. Identify.
Track. Neutralize.
Indigenous. Aatmanirbhar Bharat. Built for the next Operation Sindoor.