Integrating Microsimulation in a Multi-scale Local Freight Activity Model

Term Start:

June 1, 2026

Term End:

May 31, 2027

Budget:

$200,000

Keywords:

Microsimulation, Travel Behavior

Thrust Area(s):

Data Modeling and Analytic Tools

University Lead:

City College of New York

Researcher(s):

Michael Grossberg; Camille Kamga; Alison Conway

This project builds on results from prior TBD Center projects to advance the development of a multi-scale, synthetic data-based simulation framework for evaluating urban freight policy and infrastructure interventions. In Phase 1, we conducted a comprehensive review of existing freight data sources and machine learning- and generative-AI based approaches to address persistent data gaps for local freight decision-making. In Phase 2, we are developing a novel simulation framework to synthesize detailed truck path segments at a traffic analysis zone (TAZ) scale, and testing the utility of this synthetic data for evaluating before-after implementation scenarios for urban freight policies. While this TAZ scale simulation enables us to represent the total activity within a zone – including time spent in specific vehicle states (e.g. moving, stopping, idling) – it does not yet represent location- or segment-specific activity within the zone. However, many urban freight interventions of interest to local cities – such as street design changes, lane or signal prioritization, parking and loading regulations, and segment-specific access controls – require this more detailed granularity. This Phase 3 study aims to address this gap. Many recent freight microsimulation studies have used agent-based modeling approaches to simulate freight stakeholder decisions. Within the US, these are typically implemented at a national, state, or regional scale, and focus on individual decisions such as mode choice, route choice, or parking location choice. These studies rely on proxies like land use or industry to generate freight trips or commodity flows, which are then assigned using traditional traffic modeling approaches (e.g., user equilibrium) or statistics from literature or surveys, limiting adaptability to other decision types. Other studies assess city logistics interventions (e.g. vehicle electrification, consolidation center placement) using optimization-based approaches to fixed vehicle fleets, with objective functions minimizing logistics-related costs. Reinforcement learning appears frequently in the transportation literature but with limited freight applications. This approach could potentially be implemented to represent truck behavior as an agent that learns from repeated experience – choosing, at the segment-level, which street to traverse or where to stop, and receiving feedback based on travel or dwell time – with TAZ-scale synthetic activity defining how much activity occurs in a zone and the agent resolving where it occurs. The aim of this project is to define a simulation framework that could integrate TAZ-scale vehicle activity inputs, be implemented within a TAZ at a population scale to simulate segment-level activities, and be adaptable to simulate multiple local decision types. This project phase has three objectives: (1) to identify best-available microsimulation approaches for evaluating micro-scale freight interventions; (2) to assess the utility of machine learning approaches – in particular reinforcement learning – to improve upon static decision-assignment processes within these simulations; and (3) to develop a scalable and adaptable framework for microsimulation of freight interventions that integrates TAZ-scale truck paths with high-resolution segment-level activity.

Project scope

Scroll to Top