PROJECT NUMBER: PSR-26-02

FUNDING SOURCE: Caltrans

TOTAL COST: $110,000

PERFORMANCE PERIOD: JANUARY 2027 - JUNE 2028

Project description

The U.S. goods-movement system is efficient, but one persistent source of inefficiency is the mismatch between where local policy permits logistics development and where the freight network can support it. Agencies can observe warehouses once they are built and bottlenecks once they appear, but the land-use and permitting conditions that govern logistics development are dispersed across municipal codes, specific plans, ordinances, and permitting practices. Without a structured policy dataset, freight land-use planning remains reactive: agencies respond to congestion and entitlement conflicts after they materialize, rather than anticipating where logistics growth and corridor capacity are likely to diverge. In short, current practice does not provide any way to link existing or future freeway/arterial goods movement capacity with information on where local governments are, or are not, open to expansion of warehousing/distribution activity.

This project closes that gap by building an AI-enabled freight-policy data infrastructure that is national in design, demonstrated first in the Southern California port region, and useful for screening decisions. We will use large language models (LLMs) with a retrieval-augmented generation (RAG) framework to convert unstructured local logistics policy into validated records, construct a first-generation Logistics Land-Use Readiness Index, and overlay the index with freight-corridor performance and observed warehouse intensity. The output will allow planning triage screen: it will identify places where local policy signals and observed logistics activity coincide with already strained truck-corridor conditions.

The work builds on the team's experience with transportation data infrastructure, curb-management simulation, and logistics systems modeling (Giuliano & Xiao, 2026). It addresses a literature gap at the intersection of three strands: natural language processing (NLP) of residential-zoning codes provides useful extraction methods but does not target freight or logistics policy (National Zoning Atlas, n.d.; National Zoning Atlas & Cornell Tech, n.d.; Urban Institute, 2023); freight and logistics-zoning research establishes the policy relevance but relies on manual review and a different framing (Fried et al., 2026; Yuan, 2019); and corridor performance measurement is mature but is rarely linked to machine-readable local policy text (NCFRP, 2011; Federal Highway Administration, n.d.). This project contributes the missing intersection: auditable freight-policy extraction connected to highway/arterial corridor-readiness screening.