AI-Driven Indoor Navigation
Floorplan2Guide: LLM-Guided Floorplan Parsing for BLV Indoor Navigation
Floorplan2Guide introduces an LLM-guided pipeline for BLV indoor navigation, transforming floorplans into spatial knowledge graphs. It significantly improves navigation accuracy by leveraging few-shot learning and graph-based reasoning, outperforming direct visual reasoning. Real-world evaluations confirm its precision and accessibility benefits for visually impaired users.
Executive Impact
This research delivers significant advancements for assistive technology, offering more reliable and accessible indoor navigation for individuals with Blind and Low Vision (BLV).
0
Short Route Accuracy (Claude 3.7 Sonnet)
0
Accuracy Improvement over Direct Visual Reasoning
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
15.4%
Higher Success Rate with Graph-based Spatial Structure
Enterprise Process Flow
Map Preprocessing
→
Knowledge Graph Construction
→
Navigation Instruction Generation
→
Localization
| Model | Short Routes | Medium Routes | Long Routes |
|---|---|---|---|
| Claude 3.7 Sonnet | 92.31% (12/13) | 76.92% (10/13) | 61.54% (8/13) |
| GPT-4o | 84.62% (11/13) | 69.23% (9/13) | 53.85% (7/13) |
| LLaMA 3.2 Vision-Instruct | 53.85% (7/13) | 38.46% (5/13) | 30.77% (4/13) |
Real-World Evaluation: UMBC Math & Psychology Building
The system was rigorously evaluated in a real-world environment at the Department of Math & Psychology Building (MP-1) at UMBC. The results confirmed that instructions generated by Floorplan2Guide accurately guided users without collisions, validating the system's practical utility. The key advantage is its reliance solely on floorplans and a camera-equipped device with ArUco markers, circumventing the need for costly infrastructure like BLE beacons or RFID systems. This infrastructure-agnostic approach enhances robustness and scalability in dynamic indoor settings, providing significant benefits for BLV users. Graph-based reasoning was shown to improve navigation accuracy by up to 15.4% compared to direct visual reasoning on zero-shot prompting, reinforcing the value of explicit topological encoding.
Advanced ROI Calculator
Estimate the potential return on investment for integrating AI-powered floorplan parsing into your operations.
Estimated Annual Savings
$0
Hours Reclaimed Annually
0
Implementation Timeline
A phased approach ensures seamless integration and optimal performance of your new AI-driven navigation system.
Phase 1: Foundation Model Integration
Integrate LLMs for spatial information extraction and knowledge graph generation.
Phase 2: Data Preprocessing & Graph Construction
Implement OCR and visual analysis for floorplan features, then build the spatial knowledge graph with ArUco marker linking.
Phase 3: Navigation Instruction Generation
Develop algorithms for generating context-aware, human-readable navigation instructions from the knowledge graph.
Phase 4: Real-World Validation
Conduct extensive testing with BLV participants in diverse architectural layouts to refine and validate system performance.
Ready to Transform Your Indoor Navigation?
Discover how Floorplan2Guide can enhance accessibility and efficiency in your building. Connect with our experts today.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat