RESCUE-ALPHA is a high-performance, distributed simulation system designed for autonomous earthquake disaster response. It utilizes a state-of-the-art 4-node architecture that synchronizes real-time 3D visualization, high-concurrency telemetry handling, and LLM-driven autonomous orchestration via the Model Context Protocol (MCP).
The system enables a fleet of heterogeneous drones (Scanners and Delivery units) to systematically search for human heat signatures in rubble fields, navigate complex obstacles using A* pathfinding, and deliver life-saving aid—all without human intervention.
RESCUE-ALPHA operates on two primary planes: the Telemetry Plane (Spatial Synchronization) and the Control Plane (Autonomous Orchestration).
%%{init: {'theme': 'base', 'themeVariables': { 'primaryColor': '#f5f5f5', 'primaryTextColor': '#333', 'primaryBorderColor': '#666', 'lineColor': '#888', 'secondaryColor': '#eee', 'tertiaryColor': '#f9f9f9' }}}%%
flowchart TB
subgraph UI_Layer ["1. The Command Center (Frontend)"]
UI[React Three Fiber 3D Interface]
HUD[Tactical HUD & Bento UI]
end
subgraph Hub_Layer ["2. The Map Engine (Golang Hub)"]
Hub[WebSocket Hub]
ThermalEngine[Thermal Physics Engine]
Grid[Occupancy Grid]
end
subgraph Swarm_Layer ["3. The Swarm (Drones)"]
D1[Drone #1: Scanner]
D2[Drone #2: Delivery]
DN[Drone #N: Isolated Processes]
end
subgraph Agent_Layer ["4. The Commander (LLM Agent)"]
Agent[LangChain Agent]
MCP[MCP Server]
end
%% Telemetry Flow
D1 -.->|JSON Telemetry| Hub
D2 -.->|JSON Telemetry| Hub
Hub -.->|Broadcast| UI
%% Physics Queries
D1 <-->|HTTP /scan| ThermalEngine
%% Control Flow
Agent <-->|MCP Protocol| MCP
MCP <-->|HTTP Control| D1
MCP <-->|HTTP Control| D2
%% External AI
Agent <-->|LLM API| LLM[External LLM Provider]
%% Version-specific Styles for High Contrast
classDef layerStyle fill:none,stroke:#888,stroke-width:2px,stroke-dasharray: 5 5;
class UI_Layer,Hub_Layer,Swarm_Layer,Agent_Layer layerStyle;
- The Command Center (Frontend): A React-based 3D digital twin of the disaster zone, featuring a triple-view camera system (Global, Follow, Pilot) and real-time thermal heatmap rendering.
- The Map Engine (drone-sim-server): A high-concurrency Golang WebSocket hub utilizing non-blocking I/O to handle 100Hz+ telemetry streams with zero backpressure.
- The Swarm (drone-processes): Isolated Python processes simulating physical hardware, A* navigation, and 3D conical FOV (Field of View) thermal detection.
- The Commander (backend): An LLM-powered orchestration engine that communicates via the Model Context Protocol (MCP) to manage the fleet as a set of autonomous tools.
- Go 1.21+ (Map Engine)
- Node.js 18+ (Frontend)
- Python 3.11+ (Drones & Agent)
- LLM API Key (DeepSeek, Gemini, OpenAI, or Anthropic)
To ensure proper handshaking between nodes, start the services in the following order:
- Map Engine:
cd drone-sim-server && go run main.go --server - Drone Swarm: Start as many drones as needed in separate terminals.
cd drone && python main.py(Default: Scanner at port 8001)DRONE_TYPE=delivery DRONE_PORT=8002 python main.py
- Commander Agent:
cd backend && uvicorn backend.main:app --port 8000 - Command Center:
cd frontend && npm run dev
.\dev.ps1 -Setup # Install all dependencies and create venvs
.\dev.ps1 # Launch all 5+ terminals automatically| Layer | Stack | Key Features |
|---|---|---|
| Frontend | React, R3F, Drei, Zustand | Transient Ref Pattern, Triple-View Camera, Bento UI |
| Sim Server | Go, lxzan/gws | Non-blocking WebSocket Hub, Non-pressure Telemetry |
| Drone AI | Python, FastAPI, A* | Conical FOV simulation, Battery management, Lawnmower search |
| Command Agent | Python, LangChain, MCP | Model Context Protocol, Chain-of-Thought Reasoning |
- Architecture & Orchestration (Backend)
- 3D Visualization (Frontend)
- Telemetry Hub (Go Sim Server)
- Drone Simulation (Python Swarm)