Adaptive Multi-Sensor Interrogation of Targets Embedded in Complex Environments

摘要

This project is critical for distributed sensor and communication network operation where there are multiple sensors whose data must be combined in a computationally efficient manner across a distributed network. This methodology is required in current Air Force C2/ISR networks as requirements for distributed platform data increase as with existing and future Airborne Networks. There are five objectives to this proposal that can be used as a core set of theoretical approaches for content based data refinement in distributed and networked sensors using Markov decision theory. The first is to exploit information from previous sensing and learning, the second is to address incomplete multi-sensor data, the third is to develop POMDP and reinforcement learning sensor-query algorithms, the fourth is to develop information- theoretic algorithms for acquisition of imperfect labels, and the fifth is the development of semi-supervised algorithms.