Overcoming Data Scarcity in Calibrating SUMO Scenarios with Evolutionary Algorithms

This 21-minute conference talk from the Eclipse Foundation explores innovative approaches to calibrating traffic simulations with limited observational data. Learn how evolutionary algorithms, specifically the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), can optimize route probabilities in SUMO (Simulation of Urban Mobility) scenarios. The presentation introduces the Mannheim SUMO Traffic Model (MaST) as a case study, demonstrating how this methodology significantly improves calibration accuracy compared to baseline approaches for both 3-hour and 24-hour scenarios. Discover practical solutions for urban planning and mobility management professionals facing data scarcity challenges when developing transportation system simulations. The talk, presented by Jakob Kappenberger and co-authored with Heiner Stuckenschmidt, provides valuable insights into enhancing the effectiveness of traffic simulations even when working with limited data resources.