Information Discriminant for Global Exploitation (INDIGO): Information Filtering and Recommender Application on Sensor Network
Abstract
In the current era of perpetual advancing digital technology, ubiquitous smart sensors and devices are widely adopted to inform situational awareness, courses of actions, and decision making. Seismic, acoustic, chem-bio sensors, passive infrared, mounted cameras, and multi-modal sensors on unmanned ground and aerial vehicles are just a few examples. While the technology has greatly enhanced decisionmakers ability to understand, forecast, decide, and act fast, it has heightened the risk for drowning in information as they are inundated with growing sources of data. This is further exasperated by scarce communication resources in the field where transporting data is already expensive. Managing an excessive volume of information remains a challenge that continues to burden cognitive ability and stress communication networks. Motivated by this, we posit a sensor-to-user information prioritization framework to deliver the right information to the right person at the right time. We explore a paradigm called Information Discriminant for Global Exploitation (INDIGO) to dynamically filter, prioritize, and recommend relevant information based on the context-and-user aware reinforcement learning approach. Initial evaluation of INDIGO in the lab simulation and a field experiment showed promising results and a potential toward advanced capabilities of information mediation with further development.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 21, 2024
- Accession Number
- AD1222237
Entities
People
- Jade Freeman
- Mark Dennison
- Michael D Lee
- Tim Gregory
Organizations
- United States Army Research Laboratory