Benson Isaac

Numerical simulation of high fidelity coupled multidisciplinary systems is often hindered by exorbitant expenses associated with the need to perform fixed point iterations during an analysis. In this paper, we present a novel methodology aimed at exploiting the common additive nature of input-output mappings for physical systems and the flexibility that may exist in accuracy requirements for an analysis at hand. The methodology consists of identifying sparse approximations to coupling variable fixed point sets in univariate and bivariate subspaces using l1-minimization with a specified error tolerance. Cut-high dimensional model representations are constructed from these approximations to provide surrogate models of the input to fixed points in the coupling variable space. The result is a fully decoupled system. The methodology is demonstrated on numerical examples and a fire detection satellite model and performs well in all cases.

• For full paper: Click here

In this paper a data-driven approach to producing rapid, online estimates of aircraft capability by associating offline learned scenarios with the current aircraft state is presented. This association is carried out via an online Bayesian classification process that dynamically incorporates online sensed data by incorporating a novel confidence mapping approach that enables rapid online quantification of uncertainty in capability estimates. The problem of optimal offline learning for library construction for given online stochastic mission scenarios is formulated and a demonstration of an exhaustive search solution, as well as a greedy approach is presented. Our methodology and demonstrations are developed in the context of a high altitude, long endurance unmanned aerial vehicle.

• For full paper: Click here

The optimization of expensive black-box models is a challenging task owing to the lack of analytic gradient information and structural information about the underlying function, and also due to the sheer computational expense. A common approach to tackling such problems is the implementation of Bayesian global optimization techniques. However, these techniques often rely on surrogate modeling strategies that endow the approximation of the underlying expensive function with non-existent features. Further, these techniques tend to push new queries away from previously queried design points, making it difficult to locate an optimum point that rests near a previous model evaluation. To overcome these issues, we propose a gold rush policy that relies on purely local information to identify the next best design alternative to query. The method employs a surrogate constructed pointwise, that adds no additional features to the approximation. The result is a policy that performs well in comparison to state of the art Bayesian global optimization methods on several benchmark problems. The policy is also demonstrated on a constrained optimization problem using a penalty method.

• Full paper: Click here for journal paper