Automatic Runtime Adaptation for Component-Based Simulation Algorithms

Abstract

The state and structure of a model may vary during a simulation and, thus, also its computational demands. Adapting simulation algorithms to these demands at runtime can therefore improve their performance. While this is a general and cross-cutting concern, only few simulation systems offer reusable support for this kind of runtime adaptation. We present a flexible and generic mechanism for the runtime adaptation of component-based simulation algorithms. It encapsulates simulation algorithms applicable to a given problem and employs reinforcement learning to explore the algorithms’ performance during a simulation run. We evaluate our approach on a modeling formalism from computational biology and on a benchmark model defined in PDEVS, thereby investigating a broad range of options for improving its learning capabilities.