GTA: Generative Traffic Agents for Simulating Realistic Mobility Behavior

TL;DR

This work introduces GTA, a framework that combines census-grounded agent personas with LLM-driven planning to generate city-scale mobility simulations integrated in SUMO. By grounding synthetic populations in population statistics and embedding them in realistic networks, GTA yields emergent behavior at both the individual (persona) and aggregate (traffic) levels, enabling rapid, low-cost evaluation of mobility innovations in early design stages. Empirical validation against Mobility in Germany 2017 data and Berlin traffic counts demonstrates that GTA can reproduce modal splits and general mobility trends, while revealing systematic biases in trip lengths and mode preferences that point to directions for improvement. The framework offers a practical tool for researchers and designers to prototype mobility concepts, policy interventions, and user-experience ideas before costly field studies, with open-source availability to support open science in HCI and urban mobility research.

Abstract

People's transportation choices reflect complex trade-offs shaped by personal preferences, social norms, and technology acceptance. Predicting such behavior at scale is a critical challenge with major implications for urban planning and sustainable transport. Traditional methods use handcrafted assumptions and costly data collection, making them impractical for early-stage evaluations of new technologies or policies. We introduce Generative Traffic Agents (GTA) for simulating large-scale, context-sensitive transportation choices using LLM-powered, persona-based agents. GTA generates artificial populations from census-based sociodemographic data. It simulates activity schedules and mode choices, enabling scalable, human-like simulations without handcrafted rules. We evaluate GTA in Berlin-scale experiments, comparing simulation results against empirical data. While agents replicate patterns, such as modal split by socioeconomic status, they show systematic biases in trip length and mode preference. GTA offers new opportunities for modeling how future innovations, from bike lanes to transit apps, shape mobility decisions.

Figures (7)

• Figure 1: GTA's modularized architecture.
• Figure 2: Information flow diagram between model components.
• Figure 3: Modal split recorded in the survey ermes_mobilitat_2020 and the simulation.
• Figure 4: Modal split recorded in survey and simulation by (a) economic status and (b) occupation.
• Figure 5: Route length and route durations for survey and simulation.