• Sorted by Date • Classified by Publication Type • Classified by Topic • Grouped by Student (current) • Grouped by Former Students •
Dorit Avrahami-Zilberbrand. Efficient Hybrid Algorithms for Plan Recognition
and Detection of Suspicious and Anomalous Behavior . Ph.D. Thesis, Bar Ilan University, 2009.
Plan recognition is the process of inferring other agents' plans and goals based on their observable actions. Modern applications of plan recognition, in particular in surveillance and security raise several challenges. First, a number of key capabilities are missing from all but a handful of plan recognizers: (a) handling complex multi-featured observations; (b) dealing with plan execution duration constraints; (c) handling lossy observations (where an observation is intermittently lost); and (d) handling interleaved plans. Second, essentially all previous work in plan recognition has focused on recognition accuracy itself, with no regard to the use of the information in the recognizing agent. As a result, low-likelihood recognition hypotheses that may imply significant meaning to the observer, are ignored in existing work. In this work we present set of efficient plan recognition algorithms that are capable of handling the variety of features required of realistic recognition tasks. We also present novel efficient algorithms that allow the observer to incorporate her own biases and preferences---in the form of a utility function---into the plan recognition process. This allows choosing recognition hypotheses based on their expected utility to the observer. We call this Utility-based Plan Recognition (UPR). We demonstrate the efficacy of the techniques described above, by applying them to the problem of detecting anomalous and suspicious behavior. The system contains the symbolic plan recognition algorithm, which detects anomalous behavior, and the utility-based plan recognizer which reasons about the expected cost of hypotheses. These two components form a highly efficient hybrid plan recognizer capable of recognizing abnormal and potentially dangerous activities. We evaluate the system with extensive experiments, using real-world and simulated activity data, from a variety of sources.
@PhdThesis{dorit-phd,
author = {Dorit Avrahami-Zilberbrand},
title = {Efficient Hybrid Algorithms for Plan Recognition and Detection of Suspicious and Anomalous Behavior },
school = {{B}ar {I}lan {U}niversity},
year = {2009},
OPTkey = {},
OPTtype = {},
OPTaddress = {},
OPTmonth = {},
OPTnote = {},
abstract = {
Plan recognition is the process of inferring other agents' plans
and goals based on their observable actions. Modern applications
of plan recognition, in particular in surveillance and security
raise several challenges. First, a number of key capabilities are
missing from all but a handful of plan recognizers: (a) handling
complex multi-featured observations; (b) dealing with plan
execution duration constraints; (c) handling lossy observations
(where an observation is intermittently lost); and (d) handling
interleaved plans. Second, essentially all previous work in plan
recognition has focused on recognition accuracy itself, with no
regard to the use of the information in the recognizing agent. As
a result, low-likelihood recognition hypotheses that may imply
significant meaning to the observer, are ignored in existing work.
In this work we present set of efficient plan recognition
algorithms that are capable of handling the variety of features
required of realistic recognition tasks. We also present novel
efficient algorithms that allow the observer to incorporate her
own biases and preferences---in the form of a utility
function---into the plan recognition process. This allows choosing
recognition hypotheses based on their expected utility to the
observer. We call this Utility-based Plan Recognition (UPR). We
demonstrate the efficacy of the techniques described above, by
applying them to the problem of detecting anomalous and suspicious
behavior. The system contains the symbolic plan recognition
algorithm, which detects anomalous behavior, and the utility-based
plan recognizer which reasons about the expected cost of
hypotheses. These two components form a highly efficient hybrid
plan recognizer capable of recognizing abnormal and potentially
dangerous activities. We evaluate the system with extensive
experiments, using real-world and simulated activity data, from a
variety of sources.
},
wwwnote = {},
OPTannote = {}
}
Generated by bib2html.pl (written by Patrick Riley ) on Fri Aug 30, 2024 17:29:51