Judo

Judo is a high-performance web application designed for visualizing large-scale traffic simulation data generated by QarSUMO, QCRI’s distributed traffic simulation platform built on top of the open-source SUMO microsimulator.

Architecture

The application features a full-stack architecture that ingests QarSUMO/SUMO simulation outputs into a PostGIS geospatial database, served through a FastAPI backend and rendered via an Angular frontend with PIXI.js for hardware-accelerated graphics.

Key Contributions

System Implementation Implemented full-stack video playback of simulations on browser, including server-side rendering to reduce dependance on client system.

Performance Optimization Engineered a web-based visualization system capable of handling over 1 million vehicles simultaneously, implementing critical performance features including:

• Intelligent vehicle clustering algorithms
• Adaptive frame dropping mechanisms
• Memory-efficient buffering strategies
• Achieved 10x performance improvement (3 FPS → 30 FPS at scale)

Infrastructure & Deployment Containerized QarSUMO using Docker and developed a RESTful API interface to enable seamless deployment and horizontal scalability. Also implemented CI/CD pipelines in Microsoft Azure to deploy the system using Docker Swarm.