NASA Scientific and Technical Aerospace Reports

near-optimal strategy with DBN serving as a fitness evaluator. The probability of achieving the desired effects (namely, the
probability of success) at a specified terminal time is a random variable due to uncertainties in the environment. Consequently,
the authors focus on signal-to-noise ratio (SNR), a measure of mean and variance of the probability of success, to gauge the
goodness of a strategy. The resulting strategy will not only have a relatively high probability of inducing the desired effects,
but also be robust to environmental uncertainties.
DTIC
Bayes Theorem; Mathematical Models; Optimization; Organizations
20060001859 Connecticut Univ., Storrs, CT USA
Goal Management in Organizations: A Markov Decision Process (MDP) Approach
Meirina, Candra; Levchuk, Yuri N.; Levchuk, Georgiy M.; Pattipati, Krishna R.; Kleinman, David L.; Jan. 1, 2005; 25 pp.;
In English; Original contains color illustrations
Contract(s)/Grant(s): N00014-00-1-0101
Report No.(s): AD-A440393; No Copyright; Avail.: Defense Technical Information Center (DTIC)
Goal management is the process of recognizing or inferring goals of individual team members; abandoning goals that are
no longer relevant; identifying and resolving conflicts among goals; and prioritizing goals consistently for optimal team
collaboration and effective operations. A Markov decision process (MDP) approach is employed to maximize the probability
of achieving the primary goals (a subset of all goals). The authors seek to address the computational adequacy of an MDP as
a planning model by introducing novel problem domain-specific heuristic evaluation functions (HEF) to aid the search
process. They employ the optimal AO* search and two suboptimal greedy search algorithms to solve the MDP problem. A
comparison of these algorithms to the dynamic programming algorithm shows that computational complexity can be reduced
substantially. In addition, they recognize that embedded in the MDP solution there are a number of different action sequences
by which a team’s goals can be realized. That is, in achieving the aforementioned optimality criterion, they identify alternate
sequences for accomplishing the primary goals.
DTIC
Decision Making; Heuristic Methods; Markov Processes; Military Operations; Optimization; Planning
20060001895 North Dakota Univ., Grand Forks, ND USA
Characteristics, Nonlinearity of Statistical Control and Relations with Dynamic Game Theory
Won, Chang-Hee; Nov. 7, 2005; 6 pp.; In English
Contract(s)/Grant(s): W911NF-05-1-0212
Report No.(s): AD-A440472; ARO-47574.3-CI-II; No Copyright; Avail.: CASI: A02, Hardcopy
The PI was awarded a short-term innovative research grant through the University of North Dakota. The grant period was
15 April 2005 to 14 January 2006. However, during the grant period the PI transferred to Temple University in August 2005.
So, the PI is submitting the final progress report with the results that the PI obtained during 15 April 2005 to 31 August 2005.
The main objective of this proposal was to develop statistical control theory. Statistical control is a generalization of Kalman’s
linear-quadratic-Gaussian control, where one optimizes the probability density of the performance index by controlling the
cumulants. A statistical controller will have better performance and stability margin than the linear-quadratic-Gaussian
controller. In this project, the authors investigated the characteristics of linear statistical controllers, developed nonlinear
statistical control theory, and utilized the statistical control paradigm in dynamic game theory.
DTIC
Dynamic Programming; Game Theory; Nonlinear Systems; Nonlinearity; Optimization
20060001896 North Carolina State Univ., Raleigh, NC USA
Active Contours for Multispectral Images With Non-Homogeneous Sub-Regions
Snyder, Wesley E.; Sep. 16, 2005; 114 pp.; In English
Contract(s)/Grant(s): W911NF-04-1-0432
Report No.(s): AD-A440479; ARO-47072.1-C1; No Copyright; Avail.: Defense Technical Information Center (DTIC)
In this work, we develop a framework for image segmentation which partitions an image based on the statistics of image
intensity where the statistical information is represented as a mixture of probability density functions defined in a
multi-dimensional image intensity space. Depending on the method to estimate the mixture density functions, three active
contour models are proposed: unsupervised multi-dimensional histogram method, half-supervised multivariate Gaussian
mixture density method, and supervised multivariate Gaussian mixture density method. The implementation of active contours
176