SLIDE 1 / 9

Navigate: [←] / [→] keys

[ 4-PIECE TACTICAL SHATTER ARMED ]

CAPSTONE DEFENSE // SLIDE 01

PROJECT AMRAI

Autonomous Military Reconnaissance & Action Intelligence

A Sovereign C6ISR, Multi-Layer UAV Surveillance, 100km Radar Grid, and Autonomous Sensor-to-Shooter Defense Architecture of Bangladesh.

CORE ARCHITECTURAL PILLARS :

01 // 3-TIER UAV RECONNAISSANCE

Strategic HALE/MALE (45k ft), Tactical Mid-Tier Swarms (15k ft), and Low-Altitude Laser Designators (Nap-of-the-Earth).

02 // 100KM RADIUS AESA RADAR GRID

Six interlocking nationwide 360° GaN phased-array domes providing overlapping multi-static stealth and low-RCS tracking.

03 // COGNITIVE EW & HYBRID TELECOM

Multi-frequency FHSS across dedicated Airtel 5G slices & military bands with NIST FIPS 203 Post-Quantum Cryptography.

04 // SOVEREIGN AI DUAL-GOVERNANCE

Seamless switching between Human-in-the-Loop (HITL) authorization and sub-10ms Sovereign Autonomous (HOTL) saturation defense.

05 // INTEGRATED MULTI-DOMAIN KILL-WEB

Unified Air-to-Air micro-interceptors, Ground-to-Air SAM/CIWS, High-Power Microwave & Laser DEW, and guided 155mm artillery coordination.

SPEAKER NOTES & ORAL DEFENSE SCRIPT VERBATIM PRESENTATION TEXT

"Honorable faculty members and evaluation board, welcome to the formal capstone defense of Project AMRAI—Autonomous Military Reconnaissance and Action Intelligence. Modern high-intensity conflicts have demonstrated that legacy, human-paced command structures cannot survive saturation drone swarms, electronic warfare blackouts, and hypersonic or cruise missile threats. Project AMRAI proposes a comprehensive, sovereign C6ISR defense architecture tailored for the geographical and electromagnetic environment of Bangladesh. Our architecture unifies a three-layer autonomous drone reconnaissance fleet, an interlocking nationwide network of 100-kilometer radius phased-array radars, and a frequency-hopping, quantum-encrypted telecom backbone into a single Sovereign AI brain. Most importantly, AMRAI features a dual-governance architecture—allowing seamless switching between human Command Controller approval and sub-10-millisecond autonomous defensive execution."

CAPSTONE DEFENSE // SLIDE 02

The Operational Problem Statement & The C6ISR Gap

Four critical asymmetric vulnerabilities in conventional defense grids.

1. SATURATION ASYMMETRY COST DEFICIT

Low-cost loitering munitions ($500–$2,000 FPV swarms) used to intentionally exhaust high-cost Surface-to-Air Missile (SAM) inventories ($150,000–$500,000 per interceptor). Traditional missile reserves bankrupt within 48 hours.

2. RADAR HORIZON & TERRAIN MASKING BLIND SPOTS

Low-Radar-Cross-Section ($\sigma_{ ext{RCS}} < 0.01\, ext{m}^2$) drones exploiting riverine delta clutter, mangrove canopies (Sundarbans), urban canyons, and hilly border topography to bypass ground line-of-sight.

3. ELECTROMAGNETIC CONTEST DENIAL OF GNSS

Wideband high-power barrage jamming and GNSS/GPS spoofing completely severing standard drone radio command links and causing autonomous drift failure in commercial autopilot navigation.

4. SENSOR-TO-SHOOTER LATENCY HUMAN BOTTLENECK

Legacy voice and manual fire-direction centers require 3 to 8 minutes to coordinate artillery, air defense, and ground armor—whereas modern hypersonic and saturation swarm threats require sub-second closure.

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"To justify our engineering design, we first analyzed four critical vulnerabilities in conventional defense grids: First is cost-exchange asymmetry: an adversary can launch hundreds of low-cost drones to bankrupt a nation's missile reserves. Second is terrain and clutter masking across riverine deltas, mangrove forests, and hilly border sectors where ground radars alone suffer from horizon blind spots. Third is severe electronic warfare, where GPS and standard radio links are denied. Finally, traditional sensor-to-shooter loops take several minutes to relay target coordinates from a scout to an artillery battery. Project AMRAI solves all four challenges simultaneously through automated multi-domain fusion."

CAPSTONE DEFENSE // SLIDE 03

Three-Layer Drone ISR & 100km AESA Radar Architecture

Eliminating terrain and altitude blind spots through elevation-stacked multi-static sensing.

NATIONWIDE 100KM GA-N AESA OVERLAPPING RADAR GRID 6 MULTI-STATIC DOMES // 360° COVERAGE

TIER / LAYER	ALTITUDE	ENDURANCE	SENSOR PAYLOADS	TACTICAL KILL-CHAIN ROLE
Layer 1: Strategic HALE/MALE	40,000–45,000 ft	36–42 Hours	X-Band SAR/GMTI, Wideband ELINT/COMINT, Laser Link	Persistent wide-area maritime & border surveillance; 250km slant horizon.
Layer 2: Tactical Mid-Tier Swarm	8,000–15,000 ft	8–14 Hours	EO/MWIR Thermal Gimbal, Edge NPU, 4x A2A Micro-Interceptors	Optical target classification, airborne comms relay, Air-to-Air drone defeat.
Layer 3: Micro-UAV & Cavalry	150–1,500 ft (NOE)	90–180 Mins	3D Solid-State LIDAR, Stereo VIO, Semi-Active Laser (SAL)	Foliage/urban penetration; real-time laser painting for guided artillery.

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"Slide 3 details our physical sensing architecture. We eliminate altitude and horizon blind spots by stacking three specialized drone layers over a nationwide 100-kilometer radar grid. At 45,000 feet, Layer 1 HALE UAVs use Synthetic Aperture Radar to see through monsoon cloud cover across 250 kilometers. At 15,000 feet, Layer 2 Tactical Swarms act as the workhorse—verifying targets with thermal optics, relaying encrypted communications, and launching air-to-air micro-interceptors. Below 1,500 feet, Layer 3 Micro-UAVs operate organically with our ground troops and armored cavalry, flying under trees and into urban zones to paint targets with coded lasers. Anchoring these airborne layers are six interlocking 100-kilometer radius Gallium Nitride AESA radar domes positioned across Dhaka, Chattogram, Sylhet, Rajshahi, Khulna, and the Bay of Bengal littoral zone. Because their coverage rings overlap, AMRAI performs multi-static radar fusion—allowing one dome to transmit while adjacent domes listen passively to triangulate stealth targets."

CAPSTONE DEFENSE // SLIDE 04

Hybrid Telecom Backbone, Cognitive FHSS & Post-Quantum Cryptography

Airtel multi-band slice integration, 25,000 hops/sec spectral water-filling & FIPS 203 PQC.

DYNAMIC SPECTRUM WATER-FILLING FORMULATION

$$\max_{\{P(f)\}} \int_{B_{ ext{avail}}} \log_2 \left( 1 + rac{P(f)|H(f)|^2}{N_0(f) + J(f)} ight) df \quad ext{s.t.} \quad P(f) \le \Theta_{ ext{LPD}}(f)$$

Under enemy barrage jamming $J(f)$, our Cognitive Radio AI dynamically measures channel frequency response $|H(f)|^2$, redistributing RF energy into unjammed spectral pockets at 25,000 hops per second synchronized via Chip-Scale Atomic Clocks (CSAC).

POST-QUANTUM CRYPTOGRAPHY (PQC) SECURITY STACK

Key Encapsulation: NIST FIPS 203 ML-KEM-1024

Digital Signatures: NIST FIPS 204 ML-DSA-87

Bulk Telemetry: Hardware AES-256-GCM

LPI/LPD Steganography: CDSSS @ SNR < -15 dB

Embeds encrypted C2 telemetry inside unused guard bands of civilian cellular broadcasts, achieving full electromagnetic stealth.

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"A surveillance and weapons grid is only as strong as its communication nervous system. On Slide 4, we present our hybrid telecommunications and cryptographic architecture. Rather than relying solely on vulnerable point-to-point military radios, Project AMRAI integrates dedicated multi-frequency slices across nationwide commercial telecom infrastructure—modeled on Airtel's multi-band cellular and fiber grid—fused with military L-band and S-band links. When an adversary attempts electronic jamming, our Cognitive Radio AI executes spectral water-filling at 25,000 hops per second, shifting data into clean frequencies or hiding packets inside the ambient noise floor of civilian cellular broadcasts at minus 15 decibels SNR. Furthermore, every single telemetry frame and fire-control command is locked behind a Post-Quantum Cryptographic stack combining CRYSTALS-Kyber, Dilithium signatures, and AES-256-GCM."

CAPSTONE DEFENSE // SLIDE 05

"AMRAI" Sovereign AI Core & Dual-Decision Governance

Spatio-Temporal Graph Neural Networks fusing multi-sensor streams with state-machine governance.

MODE A LATENCY: ~1.2 SEC

Command Controller Mode (HITL)

Peacetime border surveillance and low-intensity skirmishes. AMRAI ingests radar, EO/IR, and SIGINT streams, ranks threats, and computes optimal weapon assignments.

Human Commander Authorization Required [1-CLICK AUTH]

MODE B LATENCY: < 10 MS

AMRAI Sovereign Autonomous Mode (HOTL)

Armed for hyper-velocity saturation attacks (hypersonic missiles or 500-drone swarms) where human reaction time (>3 sec) is lethal. Executes automated intercepts within pre-cleared ROE geofences.

Human-on-the-Loop Instant Emergency Veto [INSTANT OVERRIDE]

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"Slide 5 addresses the core requirement of our capstone: the AMRAI Sovereign AI Decision Engine and its dual-control governance model. In peacetime border patrol or low-intensity skirmishes, the system operates in Mode A: Command Controller HITL Mode. AMRAI fuses data from the 100km radars and 3-layer drones, ranks the threats, selects the best artillery piece or interceptor, and presents a one-click authorization prompt to the human commander. However, if the sector faces a sudden multi-vector saturation attack—such as simultaneous cruise missiles and loitering munition swarms—the commander can flip the master governance switch to Mode B: Sovereign Autonomous Mode. In this state, AMRAI closes the sensor-to-shooter loop in under 10 milliseconds, automatically coordinating jammers, lasers, SAMs, and counter-battery artillery while human officers monitor on-the-loop with instant override authority."

CAPSTONE DEFENSE // SLIDE 06

Integrated Kinetic & Directed Energy Kill-Web

Dynamic Stochastic Weapon-Target Assignment (DS-WTA) minimizing Cost-Per-Kill.

DYNAMIC STOCHASTIC WEAPON-TARGET ASSIGNMENT (DS-WTA) OBJECTIVE FUNCTION

$$\min_{\mathbf{X}} \sum_{j=1}^M V_j \prod_{i=1}^W (1 - p_{ij})^{x_{ij}} + \eta \sum_{i=1}^W \sum_{j=1}^M \left( rac{C_{ ext{shot},i}}{V_{ ext{threat},j}} ight) x_{ij} + \zeta \sum_{i \in W_{ ext{kinetic}}} rac{x_{ij}}{I_{ ext{rem},i}}$$

The second term minimizes the Cost-Exchange Ratio ($rac{C_{ ext{shot}}}{V_{ ext{threat}}}$), pairing cheap loitering munitions with $2/shot lasers or HPM pulses, while the third term penalizes consumption of limited kinetic interceptor magazines ($I_{ ext{rem},i}$).

AIR-TO-AIR (A2A)

Layer-2 carrier UAVs launch hit-to-kill micro-interceptors and airborne RF jammers against hostile drones.

GROUND-TO-AIR (G2A)

300 kW High-Energy Lasers, High-Power Microwave (HPM) anti-swarm pulses, paired with 35mm automated CIWS.

AIR-TO-GROUND (A2G)

Armed tactical UCAVs and loitering munitions cued directly by Layer-1 SAR radar and Layer-3 scouts.

GROUND-TO-GROUND (G2G)

Guided 155mm howitzers and autonomous cavalry receiving live laser painting from Layer-3 micro-drones.

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"How does AMRAI decide which weapon fires at which target? Slide 6 presents our integrated multi-domain Kill-Web and the Dynamic Stochastic Weapon-Target Assignment equation. Notice the middle term of our optimization formula: the Cost-Exchange Ratio (C shot divided by V threat). If an adversary launches a $1,000 quadcopter, AMRAI mathematically refuses to waste a $150,000 Surface-to-Air Missile. Instead, it pairs the drone swarm with a $2-per-shot High-Energy Laser, a High-Power Microwave pulse, or an Air-to-Air reusable interceptor drone. Simultaneously, when hostile ground armor or artillery is detected, Layer 3 micro-drones paint the target with a semi-active laser, and AMRAI directly slaves our automated 155mm guided artillery and autonomous armored cavalry to neutralized the threat in under 1.2 seconds."

CAPSTONE DEFENSE // SLIDE 07

Decentralized Mosaic Warfare (GPS & Comms Blackout)

Autonomous survival when central headquarters is decapitated and GNSS is jammed.

MULTI-AGENT VISUAL-INERTIAL-RANGING ODOMETRY (MA-VIRO)

When GPS is completely jammed, drones and autonomous cavalry do not crash or drift. MA-VIRO fuses stereo optical/thermal cameras, high-rate IMUs, Digital Terrain Elevation Data (DTED) contour matching, and peer-to-peer Ultra-Wideband (UWB) / mmWave ranging.

POSITIONAL DRIFT BOUNDED TO < 0.1% OF DISTANCE TRAVELED

BYZANTINE FAULT-TOLERANT (BFT) SWARM CONSENSUS

Local clusters of $N$ drones elect temporary leaders and auction targets using the Consensus-Based Bundle Algorithm (CBBA). The swarm is mathematically immune to cyber-hijacked or jammed nodes up to:

$$f \le \left\lfloor rac{N-1}{3} ight floor$$

In a 36-drone swarm, up to 11 nodes can be completely compromised or jammed without impacting mission consensus.

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"A frequent question from defense reviewers is: What happens if the enemy destroys the central AI server or jams all GPS satellites? Slide 7 explains AMRAI's decentralized Mosaic Warfare Mode. Our drones and autonomous cavalry do not rely exclusively on GPS. By combining onboard stereo vision with peer-to-peer millimeter-wave distance ranging—an algorithm we term MA-VIRO—the swarm maintains positional accuracy within 0.1 percent drift in total satellite blackout. Furthermore, using Byzantine Fault-Tolerant consensus, local clusters of drones, artillery, and ground vehicles form an ad-hoc peer-to-peer mesh network. Even if up to one-third of the drones are jammed or cyber-compromised, the remaining units mathematically reach consensus and complete the defensive mission autonomously."

CAPSTONE DEFENSE // SLIDE 08

Continuous R&D & Digital Twin OTA Evolution Lab

The Continuous Combat Learning Loop (CCLL) delivering fleet-wide model updates in < 180 seconds.

1. ANOMALY CAPTURE

Tactical units flag novel DRFM jamming or evasive maneuvers ($E_{ ext{anom}} > au$), burst-streaming raw I/Q samples.

2. SYNTHETIC TWIN

Reconstructs exact encounter in a 3D synthetic physics sandbox and spawns 10,000 parallel combat episodes.

3. MAPPO TRAINING

Centralized Critic / Decentralized Actor reinforcement learning discovers optimal evasive and counter-jamming policies.

4. LORA COMPRESSION

Exports quantized INT8 low-rank weight matrices ($\Delta W = B \cdot A$) under 1.8 MB (98.4% bandwidth reduction).

5. OTA HOT-SWAP

Signed with CRYSTALS-Dilithium and latched into edge NPU dual-bank memory in < 180 sec without reboot.

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"To dominate future warfare, a defense system cannot remain static. Slide 8 illustrates our Continuous R&D and Digital Twin Evolution Pipeline. Whenever a frontline drone encounters a zero-day threat—such as a new evasive swarm formation or an unknown frequency jammer—it captures the raw electromagnetic and kinematic signature and sends it to our Digital Twin server. The server spins up 10,000 parallel simulated battles using Multi-Agent Reinforcement Learning to discover an optimal counter-tactic. Instead of transmitting a massive 500-megabyte neural network over contested radio links, we use Low-Rank Adaptation (LoRA) to compress the newly learned counter-strategy into a 1.4-megabyte weight delta. Within three minutes of first contact, every drone, radar, and autonomous ground gun across the country receives an Over-The-Air neural upgrade mid-combat."

CAPSTONE DEFENSE // SLIDE 09

System Verification Metrics & Operational Conclusion

Empirical engineering leap over legacy human-paced defense architectures.

OPERATIONAL METRIC	LEGACY BASELINE	PROJECT AMRAI SPECIFICATION	ENGINEERING ENABLER
Sensor-to-Shooter Closure	3–8 Minutes	< 1.2s (HITL) / < 10ms (Auto)	Direct EKF sensor fusion + DS-WTA fire control
Nationwide Low-RCS Radar	Fragmented / LOS Gaps	100km Interlocking GaN AESA Rings	Multi-static GaN phased arrays + 3-layer UAVs
Anti-Jamming Resilience	Fixed/Slow FHSS	25,000 Hops/sec + SNR < -15 dB	Cognitive spectral water-filling + Airtel 5G slice
GPS Blackout Navigation	Mission Failure / Drift	< 0.1% Drift + BFT Mesh	Collaborative MA-VIRO + UWB ranging + PBFT
Zero-Day Threat Adaptation	Weeks / Months (Patches)	< 180 Seconds Fleet-Wide	Digital Twin MAPPO self-play + LoRA INT8 push

Launch Interactive Battlefield Simulator → Evaluation Committee Ready for Oral Examination

SPEAKER NOTES & ORAL DEFENSE SCRIPT

"In conclusion, Slide 9 summarizes the quantitative leap provided by Project AMRAI across sensor-to-shooter latency, radar coverage, electronic warfare resilience, and real-time AI adaptation. We have accompanied this theoretical specification with a fully interactive C6ISR Command & Control Web Simulator. I now invite the evaluation committee to direct our live demonstration—where we can toggle between Human Command Controller mode and AMRAI Sovereign Autonomous mode, trigger a multi-vector missile and drone swarm attack over the Bangladesh 100km radar grid, sever central communications to observe Mosaic mesh fallback, and inject a zero-day threat into the R&D Digital Twin lab. Thank you, and I welcome your questions."

APPENDIX // VIVA VOCE DEFENSE PREPARATION GUIDE

Evaluation Board Questions & Anticipated Defense Responses

3 CRITICAL DEFENSE PROOFS

Q1: How do you prevent an adversary from hacking the commercial Airtel telecom towers to inject false commands into AMRAI?

Defense Response: Commercial telecom towers are treated strictly as an untrusted physical transport layer under our Zero-Trust Architecture (ZTA). Every packet is encapsulated in end-to-end Post-Quantum Cryptography (ML-KEM-1024 + AES-256-GCM) and signed by hardware security modules (HSMs) using CRYSTALS-Dilithium. Even if an adversary compromises a cellular base station, they cannot forge a valid signature or decrypt a single byte of telemetry, and the FHSS engine immediately routes around compromised nodes via Layer-2 airborne UAV relays.

Q2: How does the system avoid friendly fire or civilian collateral damage when operating in Full Autonomous (Mode B) governance?

Defense Response: AMRAI does not allow unconstrained neural networks to pull triggers directly. The MARL neural network outputs candidate actions into a Deterministic Formal Safety Filter (Control Barrier Functions). This hardware-enforced gate checks 3D civilian airspace corridors, friendly IFF (Identification Friend or Foe) cryptographic transponders, and Rules of Engagement geofences before any kinetic ordinance or laser energy is released.

Q3: Why use a 3-layer drone architecture instead of just one versatile drone type?

Defense Response: Physics dictates a strict trade-off between altitude/endurance and low-level agility/cost. A high-altitude HALE drone (Layer 1) can see 250km with heavy SAR radar but is too expensive to risk near ground fire and cannot see under forest canopies. Conversely, a cheap Layer-3 micro-drone can fly inside urban corridors and under trees to laser-designate a tank for artillery, but lacks the battery and antenna aperture for long-range communication. Stacking Layers 1, 2, and 3 gives us strategic persistence, mid-tier relay/interception, and close-combat precision simultaneously.