What is it about?
Imagine trying to navigate a massive university building or shopping mall without getting lost. Our system, called Tesseract, takes simple floor plan images (like the fire evacuation maps you see on walls) and automatically converts them into digital navigation maps that computers and robots can understand. Think of it like teaching a computer to "read" a building layout. The system identifies room numbers, doors, and hallways, then creates a connected map showing how to get from any room to any other room. It's similar to how Google Maps works for city streets, but for indoor spaces. The best part? It only needs a basic floor plan image. No expensive 3D scanners, no manual data entry, and no specialized software. The system does everything automatically, making it possible to create navigation maps for thousands of buildings quickly and affordably.
Featured Image
Photo by Kier in Sight Archives on Unsplash
Why is it important?
Accessibility at Scale: Until now, creating digital indoor maps required expensive equipment or countless hours of manual work. This meant most buildings (especially in developing regions or smaller institutions) simply couldn't afford to have smart navigation systems. Tesseract democratizes this technology by working with simple images that already exist. Safety-Critical Applications: Our system ensures that every room can reach an exit, making it invaluable for emergency evacuation planning. In crisis situations, having accurate, immediately available building layouts can save lives. Privacy-First Design: Unlike systems that require cameras or sensors constantly collecting data, Tesseract works with static floor plans offline. This makes it perfect for sensitive environments like hospitals, government buildings, or corporate offices where privacy is paramount. Future-Ready Infrastructure: As autonomous robots become more common in buildings (for delivery, cleaning, or assistance), they'll need accurate indoor maps to navigate. Tesseract provides the foundation for this future, creating maps that are 90%+ accurate while being 75% more compact than traditional approaches. Timely Solution: With the rapid advancement of AI and indoor robotics, the gap between outdoor (GPS-based) and indoor navigation technology has become critical. Tesseract bridges this gap at exactly the moment when indoor autonomous systems are becoming mainstream.
Perspectives
Working on Tesseract has been rewarding because it addresses a problem I encounter regularly across the different institutions I work with. The frustration of navigating unfamiliar buildings is something everyone can relate to, whether you're a student, visitor, or facility manager. What made this research particularly enjoyable was seeing how a seemingly simple task (reading a floor plan) required combining multiple AI techniques in creative ways. The challenge wasn't just about accuracy, but about making something that could actually be deployed in real buildings without requiring specialized expertise or equipment. I'm excited about the potential applications beyond what we initially envisioned. Since publishing, we've had interest from facility managers, emergency response planners, and robotics teams who see value in having affordable, privacy-respecting indoor mapping solutions. It's gratifying to work on research that bridges the gap between academic innovation and practical, real-world utility. The collaborative nature of this project, working with colleagues across different campuses and time zones, also reinforced how important it is to build systems that are truly accessible and can work anywhere in the world.
Yaqoob Ansari
Carnegie Mellon University Department of Computer Science
Read the Original
This page is a summary of: Tesseract: Unfolding Navigable Graph Representations from Low-Semantic Floor Plans, November 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3748636.3762771.
You can read the full text:
Contributors
The following have contributed to this page
