Design and Implementation of Event-based DEVS Execution Environment for Faster Execution of Iterative Simulation

摘要

Discrete EVent systems Specification (DEVS) formalism supports the specifications of discrete event models in a hierarchical and modular manner. Diverse models have been developed applying the DEVS formalism. In our empirical studies, to analyze the behavior of models with random variables-e.g., Monte Carlo simulations-the number of simulation results is significant because more results raise the reliability of the analysis. However, the existing execution environment we developed is not focused on faster simulation for collecting plenty of data. In order to speed up the DEVS simulation, the proposed execution environment treats events instead of atomic models as basic simulation units, i.e. event scheduling world view applied to DEVS execution for faster execution. This paper proposes an event-oriented DEVS model for simple event scheduling and suggests mediation processes to simulate the DEVS-compliant model with improved performance, similar to simple event scheduling. This paper discusses a performance test to verify the execution environment. Experimental results show that the proposed environment has a significantly shorter execution time. The expectation is that this improved simulation engine will be applicable to faster DEVS execution for collecting simulation results.