Supervised Machine Learning for Input Modelling of an Agent-Based Simulation Model for Autonomous On-Demand Shuttle Services

The quality of simulation-based experimentation is directly related to the estimation of its key input parameters. Yet, especially when it comes to innovative transportation concepts that are characterized by a multiplicity of influencing factors, such as autonomous on-demand transport systems, reliable and realistic input parameters are difficult to obtain. In order to tackle this challenge, we propose an automated machine learning integration to estimate contextualized parameters for an agent-based simulation model. To demonstrate the effectiveness of the proposed approach, a test scenario is conducted on the estimation of mobility patterns for an agent-based simulation model of autonomous on-demand shuttle operations. Here, the results of several simulation experiments prove the viability of the proposed input modelling approach, showing that an automated machine learning integration can generate more accurate estimations of input parameters in innovative, highly uncertain systems in an efficient manner.