Fast SimEDaPE: Simulation Estimation by Data Patterns Exploration

Abstract

In the context of smart cities, solving problems such as pollution, congestion, and public transport, regularly faced by large cities like São Paulo, is not trivial. To tackle those problems researchers often rely on simulations. An example of a smart city simulator is InterSCSimulator, which simulates urban traffic. However, this simulator has limitations regarding its performance in large scale scenarios. SimEDaPE, a technique used to improve simulation performance based on the recurrence of patterns from previous simulations, was proposed in this context. SimEDaPE is still under active development and as such has some performance bottlenecks in some stages, such as the temporal mapping stage. In this work, we propose an improvement to this step of SimEDaPE using optimized libraries (written in C instead of Python), and parallelism. As a result, we obtained a considerable relative performance of 156x, running on 8 cores compared to the reference sequential implementation.