agent-based transit simulation

Existing agent-based travel demand models suffer from dual limitations in behavioral realism and computational efficiency. This paper introduces, for the first time, large language model (LLM)-driven agents into transportation system modeling, establishing a learnable and adaptive individual travel behavior simulation framework. The framework leverages LLMs’ reasoning and interactive capabilities to capture fine-grained human decision-making, social learning, and dynamic strategy adaptation, while adhering to key behavioral principles. Through simulation experiments on canonical traffic bottleneck scenarios, we demonstrate that LLM agents exhibit realistic path choice, responsive congestion avoidance, and continuous behavioral optimization. Critically, the approach improves behavioral fidelity without sacrificing scalability, while substantially enhancing model generalizability and interpretability. This work establishes a novel paradigm for next-generation high-fidelity, low-computational-overhead traffic simulation.

This study addresses the challenges of modeling human mobility behavior and generating personalized travel solutions in multimodal transportation systems. We propose the first framework integrating large language model (LLM)-driven generative agents into agent-based traffic simulation. Methodologically, the framework synergizes the GAMA simulation platform, GTFS transit data, and the OpenTripPlanner routing engine, enabling agents to autonomously learn and evolve travel decisions within dynamic urban environments. Unlike static rule-based or statistical models, it supports the emergence of context-aware, stable individual travel habits over time. Evaluated in a one-month simulation for Toulouse, France, the framework demonstrates high behavioral fidelity and robust personalization capability. Results confirm enhanced ecological validity in mobility behavior modeling and improved utility for policy analysis and scenario forecasting.

This study addresses the design of public transit fare policies that balance equity and efficiency within complex multimodal transportation systems, explicitly accounting for traveler behavior, socioeconomic heterogeneity, and network interaction effects. To this end, we develop a scalable, data-driven simulation framework integrating synthetic populations, agent-based modeling, multimodal travel time estimation, and a fare-sensitive mode choice model. We further introduce a sampling-accelerated algorithm that achieves a favorable trade-off between computational efficiency and aggregate accuracy, enabling, for the first time at an urban scale, a fine-grained evaluation of how fare policies mediate trade-offs among equity, ridership composition, and fiscal revenue. Empirical results indicate that fare adjustments have limited impact on total ridership but substantially alter modal shares and generate heterogeneous effects across income groups; notably, fare-free transit significantly alleviates mobility burdens for low-income populations, albeit at considerable fiscal cost.

This study addresses the implementation effectiveness and mechanism design of Tradable Credit Schemes (TCS) in transportation systems. We develop the first multi-agent simulation framework integrating travelers, regulators, and credit markets, and implement end-to-end microscopic TCS simulation on the SimMobility platform for the first time. Methodologically, we combine activity-based travel demand modeling, mesoscopic multimodal network representation, and Bayesian optimization–driven automated parameter calibration. Key contributions include: (i) proposing a flexible TCS design paradigm balancing equity and efficiency; (ii) empirically validating daily convergence and theoretical consistency of TCS; (iii) quantifying heterogeneous impacts across user groups, travel behaviors, and local road segments; and (iv) identifying and mitigating market failures such as price manipulation and strategic hoarding. Results demonstrate that TCS significantly alleviates congestion and provides policymakers with a verifiable, tunable simulation foundation for evidence-based decision-making.

This paper addresses the challenge of constructing a reproducible, data-driven activity-based travel demand model for the Melbourne metropolitan area—specifically, how to generate high-fidelity synthetic populations and realistically simulate individual activity chains and travel behavior solely from publicly available data. Method: We propose an “activity-first, trip-derived” paradigm: first generating activity chains (including type, sequence, and duration) via hierarchical clustering and probabilistic modeling; then allocating activity locations and travel modes using a gravity model integrated with spatial decay effects; and finally enforcing a constraint on remaining activities to ensure logically consistent return trips. All components are empirically calibrated against observed data and fully compatible with the MATSim simulation framework. Contribution: We introduce the first activity-chain-oriented modeling framework that is entirely open-source, requires no proprietary data, and guarantees cross-platform reproducibility—thereby advancing transparency, accessibility, and scientific rigor in urban transport modeling.

Traditional transportation models struggle to capture individuals’ behavioral adaptability in dynamic traffic environments. To address this limitation, this study proposes a Multi-Activity Travel and Mobility model based on agent-based modeling (MATraM), which, for the first time, integrates dynamic activity scheduling and rescheduling mechanisms into an agent-based traffic modeling framework, enabling agents to flexibly adjust their daily plans in response to real-time travel conditions. By combining agent-based modeling (ABM), the ODD protocol, dynamic route planning, and activity scheduling algorithms, the model jointly simulates individual behavioral flexibility and system-level congestion dynamics, thereby generating more realistic travel patterns. This approach establishes a novel paradigm for next-generation, scalable transportation models that better reflect the adaptive nature of human mobility.

This study addresses the limitations of traditional traffic behavior modeling, which relies on handcrafted rules and expensive data, making it ill-suited for predicting large-scale human mobility responses during the early stages of emerging technologies or policies. The authors propose a large language model (LLM)-driven framework that generates synthetic populations from census data and simulates context-sensitive activity schedules and mode choices without manual rule specification. This work represents the first application of LLM-based personalized agents to city-scale transportation simulation, enabling flexible evaluation of novel interventions such as bike lanes or mobility apps. Experiments in Berlin demonstrate the model’s ability to reproduce observed travel mode distributions across socioeconomic groups, despite systematic biases in trip distance and mode preference, thereby validating its potential and scalability for urban mobility analysis.

This study addresses the challenge of holistically evaluating urban traffic control policies, where direct effects—such as changes in traffic flow and emissions—are intricately intertwined with indirect effects, including behavioral responses and shifts in economic accessibility. To this end, the authors propose a multilayer urban mobility simulation framework that integrates a physical layer (modeling traffic dynamics and emissions) with a social layer (capturing user behavioral responses). The framework leverages real-world data to instantiate scenarios, encode policy parameters, and formalize behavioral assumptions, thereby enabling systematic comparison and forward-looking assessment of diverse “what-if” policy scenarios. Applied to vehicle restriction policies, the approach effectively uncovers the interactive mechanisms between policy design and user feedback, offering actionable insights for developing more anticipatory and coordinated transportation policies.

This study addresses the optimization of investment strategies in public transportation systems under limited indivisible resource constraints, balancing multi-agent utility and fairness-based welfare. Focusing on two topological settings—linear routes (station selection) and weighted networks (edge traversal time optimization)—the work formulates a multi-agent resource allocation model. Theoretical analysis demonstrates that computing an approximately fair-optimal solution is NP-complete and inapproximable on general graphs. However, for a fixed number of agents—particularly in single- or two-agent scenarios—the authors propose polynomial-time algorithms combining Dijkstra’s shortest-path method with dynamic programming. These results extend to railway network design, offering both theoretical rigor and practical relevance for infrastructure planning under fairness considerations.