English | 中文
Xianbiao (XB) Hu, PhD — Associate Professor, Department of Civil and Environmental Engineering, Pennsylvania State University · xbhu@psu.edu
Published in Transportation Research Today, Volume 1, November 2026, Article 100009 (Elsevier) · Open access, CC BY-NC-ND 4.0 📄 Read the full article · DOI: 10.1016/j.trt.2026.100009
📰 Coverage: WeChat article (Chinese) · LinkedIn post
Highly automated vehicle (HAV) deployments are expanding fast: Waymo robotaxis operate in multiple U.S. cities, Aurora's autonomous trucks haul freight on the Dallas–Houston corridor, and the Autonomous Vehicle Industry Association reports over 145 million autonomous miles logged on U.S. roads. Yet no vehicle a consumer can currently purchase operates beyond SAE Level 2 — the consumer "self-driving car" does not exist in any legally meaningful sense. Even where it might, many states allow only corporations and institutions, not individuals, to hold operating certificates. The gap between what people assume and what the law permits is wider than most realize.
This paper argues that the gap is not an information deficit. It is the product of three structural forces:
- American federalism — 50 states have developed 50 independent regulatory regimes with no automatic reciprocity between them
- Market incentives — companies are rewarded for blurring the line between driver-assistance and genuine autonomy (e.g., Tesla naming a Level 2 system "Full Self-Driving")
- Legislative lag — state legislatures enact HAV statutes faster than enforcement agencies can build the training, protocols, and institutional capacity to implement them
Drawing on expert focus groups with 35 participants from 29 organizations, a legislative review of all 50 states and the District of Columbia, and the author's direct involvement in evaluating Pennsylvania's Act 130, the paper poses six diagnostic questions that expose how these forces produce consequential policy failures:
- Is your car really "self-driving"?
- Who is allowed to put an HAV on the road?
- Should all levels of automation be regulated the same way?
- What does it take for "driverless" vehicles to scale economically?
- Can an HAV company operate across state lines?
- What happens after the law is signed?
Among the findings: a single interstate freight trip can cross six states with six incompatible regulatory regimes, insurance requirements ranging from zero to five million dollars, and no automatic interstate reciprocity. State-level fragmentation is a durable institutional feature of American HAV governance, not a temporary phase. Autonomous freight is identified as the highest-return near-term policy priority.
Full abstract and plain-language summary: paper/abstract.md
Q1: Is your car really "self-driving"? · Q2: Who can put an HAV on the road? · Q3: Should all automation levels be regulated the same way? · Q4: What does it take for HAVs to scale? · Q5: Can an HAV company operate across state lines? · Q6: What happens after the law is signed? · Policy recommendations
The analysis draws on three evidence streams: expert focus groups (35 participants, 29 organizations) conducted as part of a PennDOT-funded evaluation of Pennsylvania's Act 130; a legislative review covering all 50 states and the District of Columbia; and publicly available industry data on operational economics and safety performance.
Figure 1. Organizations included in the Pennsylvania Act 130 focus groups (left) and their distribution by category (right).
No new vehicle a consumer can currently purchase operates beyond SAE Level 2 — the consumer "self-driving car" does not exist in any legally meaningful sense. Mercedes-Benz DRIVE PILOT was, until January 2026, the only Level 3 system ever certified for U.S. consumer use; Mercedes has since paused it for new models. Every Level 4 system (Waymo, Aurora, Kodiak, Zoox, Gatik) is a corporate fleet operation, unavailable for individual purchase. Tesla's naming of a Level 2 system "Full Self-Driving" is the most consequential example of how market incentives produce public confusion: AAA found only 43% of drivers can accurately define full automation, and nearly 1 in 10 believed they could currently buy a vehicle that drives itself while they sleep. In December 2025, the California DMV ruled Tesla's marketing violated state law; Tesla sued to reverse the ruling in February 2026.
There is no single answer. Only 11 of the 34 states with any HAV-related framework have detailed, AV-specific regulations. Under Pennsylvania's Act 130, certificate holders must be organizations — individuals are explicitly excluded, so a Pennsylvanian could not legally operate a Level 3/4 vehicle at home even if the technology were perfected tomorrow. Texas draws the opposite line: commercial operators need authorization, but individuals using a vehicle for personal use are exempt. Arkansas, Alabama, and Louisiana restrict HAV operation to commercial use only. The rules governing who may even attempt to deploy an HAV vary so fundamentally that no single national business model fits all states.
Most state statutes, including Act 130, group Level 3, 4, and 5 under one "HAV" definition — a product of legislative convenience, not a considered risk-based policy judgment. A Level 3 vehicle with a licensed driver behind the wheel presents a fundamentally different enforcement challenge than a driverless Level 4 freight truck. States already implicitly differentiate through driver requirements even without an explicit tiered framework. The European Union's UNECE Regulation 157 shows that a separate, differentiated type-approval process for Level 3 is both technically and legally feasible — a model no U.S. state or federal agency has adopted.
"Driverless" does not mean "humanless." Every HAV deployment on public roads today requires remote human operators. Aurora's own data shows remote-operations cost of $0.64/mile at a 1:1 truck-to-operator ratio, dropping to $0.06/mile at 25:1 — a ratio no commercial operation has yet achieved at scale. Goldman Sachs estimates the current all-in cost of autonomous trucking at $6.15/mile, projected to fall to $1.89/mile by 2030. Freight is the application most likely to close this gap first, thanks to a simpler operational design domain — but technology working is not the same as a company surviving: TuSimple, Embark Trucks, and Argo AI (a $12.4B company) all folded despite demonstrated technical capability.
Federal unification has failed three times in a decade and shows no sign of arriving. The SELF DRIVE Act passed the House in 2017 and died in the Senate; its 2021 version never got a floor vote; the 2026 version is given a 1% chance of enactment by GovTrack. Meanwhile, a single autonomous freight run from Pennsylvania to Texas along the I-81/I-40 corridor crosses six states with six incompatible regulatory regimes — insurance requirements alone range from $0 to $5 million, with no automatic interstate reciprocity.
Figure 2. Regulatory transitions along the I-81/I-40 corridor from Pennsylvania to Texas.
Legislatures can pass an HAV statute in a single session; the enforcement infrastructure to implement it takes years to build. Pennsylvania's Act 130 was signed in November 2022, but certificate-of-compliance guidelines weren't issued until October 2024. No standardized national protocol exists for how a police officer conducts a traffic stop on a vehicle with no driver; first responders largely lack training to safely disable a crashed HAV's high-voltage battery and automated systems simultaneously. Reporting requirements are similarly fragmented: California requires public disengagement and collision reports, while Texas and Arizona keep most operational data confidential.
Each recommendation targets one of the three structural forces identified above:
- Standardized consumer-facing labeling of automation levels, to counter market incentives that reward ambiguity (Q1)
- Differentiated regulatory frameworks for Level 3 vs. Level 4/5, following the EU's UNECE Regulation 157 model (Q3)
- Interstate compacts or mutual recognition agreements, prioritizing eastern freight corridors where small states create frequent regulatory transitions (Q2, Q5)
- Mandatory standardized law enforcement and emergency response protocols as a condition of HAV certification (Q6)
- Transparent public disclosure of remote-operator ratios and operational economics from HAV certificate holders (Q4)
- Autonomous freight corridor harmonization as the highest-return near-term priority, given freight's comparative path to economic viability (Q4, Q5)
paper/
HAV_policy_paper_final.pdf Full paper (author's version, OA)
HAV_Policy_Slides.pdf Conference/presentation slide deck
abstract.md Abstract + plain-language summary + author bio
tables.md All 7 tables, transcribed as Markdown
references.md Full reference list (~80 sources)
figures/ Figures 1–2 and Tables 1–7 as images
reference/ Underlying PennDOT Act 130 evaluation materials
ACT 130 Draft_Final report.pdf
ACT 130 QA document_v2_final.pdf
ATMA Policy Reivew.pdf
CITATION.cff Machine-readable citation metadata
APA
Hu, X. (2026). Highly automated vehicle policy in the United States: Six questions exposing the structural misalignment between perception, regulation, and enforcement. Transportation Research Today, 1, 100009. https://doi.org/10.1016/j.trt.2026.100009
BibTeX
@article{Hu2026HAVPolicy,
title = {Highly Automated Vehicle Policy in the United States: Six Questions Exposing the Structural Misalignment Between Perception, Regulation, and Enforcement},
author = {Hu, Xianbiao},
journal = {Transportation Research Today},
volume = {1},
pages = {100009},
year = {2026},
doi = {10.1016/j.trt.2026.100009},
publisher = {Elsevier}
}GitHub will also generate a citation via the Cite this repository button, powered by
CITATION.cff.
Dr. Xianbiao (XB) Hu is an Associate Professor of Civil and Environmental Engineering at Penn State and a Managing Editor of the Journal of Intelligent Transportation Systems. His research spans smart mobility systems, automated vehicles, Physical AI, and transportation electrification. He partnered with PennDOT to evaluate Act 130 and leads research on autonomous truck-mounted attenuator deployment through a national pooled fund study with 17 state DOTs.
This study was partially funded by the Pennsylvania Department of Transportation (PennDOT). The views, findings, and conclusions are solely those of the author and do not necessarily reflect the official policy or position of PennDOT or the Commonwealth of Pennsylvania.
The article is distributed under CC BY-NC-ND 4.0: reuse and redistribution are permitted for non-commercial purposes with attribution, without modification. Please cite the published version linked above.