SCSC2003 Abstract S81292

Mapping ODEs to DEVS: Adaptive Quantization

Mapping ODEs to DEVS: Adaptive Quantization

Submitting Author: Prof. Hans Vangheluwe

Abstract:
We show how Initial Value Problems can be solved using quantization. Quantization is the dual of discretization, in that the dependent variables' space is partitioned (rather than the independent variable's). First, a non-adaptive quantization algorithm, expressed in the DEVS formalism and based on the Forward-Euler approximation, is presented. We show that consistency as well as convergence are respected for autonomous systems, but cannot be guaranteed for nonautonomous problems. Absolute-stability as it is usually defined is generally not achieved. We then introduce an adaptive quantization algorithm, which improves the overall performance of its non-adaptive counterpart.



Back to SCSC2003 Abstracts