Researchers at Colorado State University have created ARMOR, a new system that significantly improves how we pinpoint locations indoors using Wi-Fi. ARMOR is engineered to deliver accurate location tracking even in challenging conditions: 1) it handles Wi-Fi signal distortions caused by different mobile devices, 2) it adapts to distortions in Wi-Fi signals over time due to human movement or furniture rearrangement, and 3) it defends against the often-overlooked risk of cyber-attacks designed to mislead indoor localization systems. ARMOR features a dual-AI architecture: the first AI model performs real-time location estimation, while a second AI model continuously monitors the system’s integrity, detecting errors and applying corrections on the fly. This approach ensures reliable and safe indoor localization all in a lightweight package that can be deployed on smartphones.
Pinpointing exact indoor locations is becoming highly valuable for many everyday uses, from asset tracking to personalized services. A popular way to do this is by using Wi-Fi signals collected from all available routers at a given location, known as a “Wi-Fi fingerprint.” Wi-Fi is widely accessible and compatible with most mobile devices, making it a convenient choice for indoor localization / navigation. But this method can get thrown off by small differences in devices or changes in the room, like moving furniture. While a new approach called federated learning lets different devices help train a central location model without sharing private data. But deploying federated learning in real-time, ever-changing indoor spaces introduces new risks: the system can become corrupted due to mobile device, indoor environmental changes, or even deliberate cyber-attacks
ARMOR first learns what a normal, healthy update to its central location model should look like by using a special “state-space model” (SSM). This SSM acts like a smart predictor, anticipating the next expected changes in the model’s “brain.” When a mobile device sends its updated location data, ARMOR checks if these updates match what the SSM predicted. If the updates deviate too much from the prediction, ARMOR steps in. It figures out exactly which parts of the update are problematic and adjusts them, essentially cleaning up the data before it can corrupt the main location model. This technique means only reliable and corrected information is used to improve the overall location system. In real-world tests, ARMOR significantly outperformed existing systems, reducing common location errors by up to 8 times and drastically cutting down on the worst-case errors by nearly 5 times. Plus, it’s designed to run efficiently on mobile devices, keeping user data private.
Available for Exclusive Licensing
TRL: 4
US 19/741,853
Sudeep Pasricha
Danish Gufran
Akhil Singampalli
Jessy McGowan
Jessy.McGowan@colostate.edu
970-491-7100