DEV Community

ADHIMULAM LEENA
ADHIMULAM LEENA

Posted on

How MongoDB Enables Edge Intelligence for AI Without Internet

#ai #security #programming #webdev

Written by Leena, Vedanth, Vaibhav, and Laasya

Guided by Chanda Raj Kumar

In contemporary software engineering systems, it has become commonplace to design under the presumption of "High Availability." We build systems around global cloud regions and auto-scaling clusters connected by high-speed fiber optic networks. There is also a standard expectation of at least 99.999% uptime.

But what happens when your connection to the cloud - located thousands of miles away - is lost or disrupted? In military environments, this is known as a DDIL environment (Degraded, Disrupted, Intermittent, and Low Bandwidth).

In such environments, traditional AI architectures do not just slow down - they fail completely. Without connectivity, AI models become isolated and unusable. To solve this, we explore Collaborative Edge Intelligence (CEI) and introduce a concept called a self-healing orchestrator, capable of redistributing computational intelligence across a decentralized network.

🚀 The Role Of MongoDB In Edge Intelligence Systems

Efficient data management is essential in decentralized, low-connectivity environments. MongoDB is a key technology in the Tactical Edge Orchestrator (TEO) for enabling scalable, real-time and flexible data management.

💡 Why Use MongoDB?

MongoDB has a document-based database structure that can support dynamic and unstructured data storage; this provides a key advantage over traditional relational databases when dealing with edge environments where data can vary greatly in format among all the devices (UAV, Sensors, Communication Nodes) that produce it.

🔐 User Authentication With MongoDB

MongoDB is used to facilitate the management of user authentication and session management for each user. User credentials (user names, emails, hashed passwords) are stored securely in MongoDB collections with the associated user data. When a user logs in, their credentials are independently validated against the stored user data in order to provide a secure, efficient means of controlling access to the system.

# MongoDB Connection (PyMongo)
from pymongo import MongoClient
from passlib.context import CryptContext

client = MongoClient("mongodb://localhost:27017/")
db = client["teo_db"]
users_collection = db["users"]

pwd_context = CryptContext(schemes=["bcrypt"], deprecated="auto")

# User Signup
def create_user(username, email, password):
    hashed_password = pwd_context.hash(password)

    user = {
        "username": username,
        "email": email,
        "password": hashed_password
    }

    users_collection.insert_one(user)
    return {"message": "User created successfully"}
Enter fullscreen mode Exit fullscreen mode

Passwords are securely hashed before being stored in MongoDB, ensuring safe authentication.

# Login Verification
def verify_user(email, password):
    user = users_collection.find_one({"email": email})

    if user and pwd_context.verify(password, user["password"]):
        return {"message": "Login successful"}
    return {"message": "Invalid credentials"}
Enter fullscreen mode Exit fullscreen mode

This enables secure login validation using stored MongoDB data.

⚡ Real-time Data Processing

The system continuously processes high frequency data including telemetry, signal strength and node health. MongoDB's support for high performance read/write operations provides an effective means of storing:

• Current telemetry
• Current node status
• Current system state

As a result, the status of the system and the dashboard reflects the most current condition of the system without delay.

# Storing Real-Time Telemetry Data
telemetry_collection = db["telemetry"]

def store_telemetry(node_id, signal_strength, battery):
    data = {
        "node_id": node_id,
        "signal_strength": signal_strength,
        "battery": battery
    }

    telemetry_collection.insert_one(data)
Enter fullscreen mode Exit fullscreen mode

MongoDB efficiently handles continuous incoming data from edge devices like sensors and UAVs.

🔄 Event Loggings & Recovery of System

The need to monitor failures and recoveries in a self-healing architecture requires tracking all failure events along with their recovery event logs. The following events are logged via event logging solutions like MongoDB:

Node Failure Events
Communication Disruptions
Data Re-routing Events

This logging procedure provides the ability to observe how the overall system is performing, in addition to how well the overall system can perform under stress.

# Logging System Events
logs_collection = db["logs"]

def log_event(event_type, description):
    event = {
        "event_type": event_type,
        "description": description
    }

    logs_collection.insert_one(event)
Enter fullscreen mode Exit fullscreen mode

These logs help monitor failures, rerouting decisions, and system recovery in real-time.

📈 The Scalability and Flexibility

MongoDB's ability to scale horizontally enables the addition of multiple additional nodes/environments, while effectively managing increasing volumes of data as the system grows, as well, and provide a flexible record scheme so that the addition of new types of data does not require extensive reorganization.

# Creating Index for Faster Queries
users_collection.create_index("email", unique=True)
Enter fullscreen mode Exit fullscreen mode

Indexing improves performance, especially for authentication and frequent queries.

Integrating MongoDB into an architecture enables a system that provides a balance of performance, flexibility and resiliency for distributed digital intelligence (DDIL)-based-edge-intelligence applications.

⚠️ 1. The Critical Flaw: The Cloud Dependency Trap

Most current Edge AI systems rely heavily on cloud computing:

Single point of failure: If a device (e.g., drone) loses cloud connection, it loses intelligence.
Hardware limitations: Large AI models cannot run on lightweight edge devices.

🛰️ Tactical Edge Orchestrator (TEO)

The TEO organizes operations across a decentralized structure:

Three-tier architecture:

Local Edge: Sensors, wearables, handheld devices
Tactical Edge: Cluster servers on vehicles
Remote Cloud: Used for long-term analysis, not real-time survival

🧩 Key Processes:

Neural Network Sharding:
AI models are split into smaller parts and distributed across multiple nodes. Like splitting a library into pages carried by different people.

Pipeline Parallelism:
Different nodes process different layers of data simultaneously, improving speed and efficiency.

🛡️ 2. Digital Twin Technology: A Virtual Safety Net

Each physical node has a Digital Twin (DT) that monitors:

Battery level
Signal strength
Hardware health

🔁 Self-Healing Mechanism:

If a node fails, the system automatically redistributes its workload to other healthy nodes without human intervention.

⚙️ 3. Engineering Stack

This engineering stack is designed specifically for decentralized, real-time AI operations at the tactical edge.

By combining lightweight frontend frameworks, a high-performance backend, and real-time communication systems, the platform ensures that data can be processed, shared, and acted upon instantly - even in low-connectivity environments.

With technologies like FastAPI powering the backend, Redis and WebSockets enabling real-time communication, and MongoDB handling user authentication and session management, the system achieves both speed and resilience making it well-suited for mission-critical scenarios.

🔌 4. Communication Protocol Efficiency

Different protocols were tested for efficiency:

Protobuf (REST): High accuracy but resource-heavy
gRPC: Balanced performance and energy usage
WebRTC: Best option (low latency + low power, peer-to-peer)

📊 5. Stress Testing Results

0% failure: 98% accuracy
20% failure: 95% accuracy (system self-healed)
40% failure: 90% accuracy (system still functional via alternate routing)

🧪 Code Snippets

1. FastAPI app setup 

app = FastAPI(title="BridgeIT CTI Orchestrator v2")
orchestrator = Orchestrator()
Enter fullscreen mode Exit fullscreen mode

2. WebSocket live updates 

@app.websocket("/ws")
async def websocket_endpoint(ws: WebSocket):
    await ws_manager.connect(ws)
    await ws.send_json({"type":"state","data":orchestrator.get_full_status()})
Enter fullscreen mode Exit fullscreen mode

3. Status API 

@app.get("/status")
async def status():
    return orchestrator.get_full_status()
Enter fullscreen mode Exit fullscreen mode

4. Kill chain simulation 

@app.post("/kill-chain/advance")
async def advance():
    result = orchestrator.advance_kill_chain()
    await _broadcast_state()
    return result
Enter fullscreen mode Exit fullscreen mode

5. DDIL failure simulation 

@app.post("/simulate-failure")
async def ddil():
    msg = orchestrator.simulate_failure()
    await _broadcast_state()
    return {"message": msg}
Enter fullscreen mode Exit fullscreen mode

6. SOC action API 

@app.post("/soc/action")
async def soc(req: SOCActionRequest):
    result = orchestrator.soc_response(req.action, req.target_node)
    await _broadcast_state()
    return result
Enter fullscreen mode Exit fullscreen mode

🔮 6. Future Scope

Federated Multi-Agent Reinforcement Learning (MARL): Integration across air, ground, and sea
Neuromorphic Computing: Brain-like chips using minimal power
Aerial Edge Networks: Communication via high-altitude balloons

🧾 Final Conclusion: The Human Factor

The goal is to reduce human burden in critical situations. By enabling self-healing systems, we shift from relying on the cloud to becoming the cloud ourselves.

🔗 Project Links

Github Link:- https://github.com/aleena-1/Battlefield-Orchestrator/tree/main
Render Link:- https://battlefield-orchestrator.onrender.com/

Top comments (0)