Partially observable Markov decision processes of sensor detection

Research output: Contribution to conferencePaperpeer-review

Abstract

This paper considers a situation in which a sensor network detects signals from approaching objects (e.g., small boats traveling in a harbor) while protecting a stationary target (e.g., a large battleship at anchor). Once a sufficient number of signals are detected by the network (for example, through video surveillance), it may classify the object's intention as hostile and act accordingly by informing responders. The objective is to accurately and timely recognize intruders. We show that the optimal policy is of a Control Limit Threshold (CLT).

Original languageEnglish
StatePublished - 2010
Event20th Annual Workshop on Information Technologies and Systems, WITS 2010 - St. Louis, MO, United States
Duration: 11 Dec 201012 Dec 2010

Conference

Conference20th Annual Workshop on Information Technologies and Systems, WITS 2010
Country/TerritoryUnited States
CitySt. Louis, MO
Period11/12/1012/12/10

Fingerprint

Dive into the research topics of 'Partially observable Markov decision processes of sensor detection'. Together they form a unique fingerprint.

Cite this