Towards Formal Specification of Agent-based Transport Models

Agent-Based Transport Models (ABTMs) are powerful tools for simulating complex transport systems, yet they often lack formal specifications, limiting their transparency, reproducibility, and automation. While behavioral (sub-) models such as discrete choice models have a formal foundation, their integration into broader simulation systems is usually informally documented or solely defined in source code, making it difficult to replicate findings, reuse components, or enforce best practices. To address this, this paper conducts a comprehensive analysis of various aspects observed within existing different ABTM frameworks, highlighting the variation of their purposes, resolution, agent types, and modeling methods. Building on this analysis, we propose a metamodel tailored for ABTM specification, defining common concepts in an abstract syntax for defining agent attributes, behaviors, and interactions. Our proposed metamodel emphasizes a method-agnostic design, enabling integration and reuse of already existing behavioral models. This paves the way for automatic model validation, code generation, and sensitivity analysis. Additionally, we outline a conceptual mapping to Discrete Event System Specification (DEVS) to formalize the dynamic semantics, along with a graphical syntax for intuitive model creation by transport engineers.