Naturalistic Driving Data Baseline for Automated Driving System-Equipped Commercial Motor Vehicles [Technology Brief]

NHTSA · 2021 · ROSA P / United States. Department of Transportation. Federal Motor Carrier Safety Administration. Office of Analysis, Research, and Technology

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This technology brief addresses the need for standardized safety benchmarks to evaluate Automated Driving Systems (ADS) in commercial motor vehicles (CMVs). As ADS technologies like platooning and traffic jam assist advance, regulators and developers require reliable baselines to compare ADS performance against human drivers. The study’s primary objective was to establish unique, public-domain baseline rates for human CMV drivers, specifically focusing on highway environments where these technologies are most mature. The research team developed two categories of safety performance baselines: "lagging" and "leading." The lagging baseline relied on historical crash data derived from 3.44 billion vehicle miles traveled (VMT) and over 3,700 crashes. This data was sourced from SmartDrive Systems, Inc., using onboard monitoring systems that recorded kinematic and video data over a two-year period starting in August 2016. The leading baseline, designed to capture real-time operational performance, utilized over 3.2 million miles of continuous driving data matched to detailed map information. These metrics focused on longitudinal control (e.g., deceleration, following distance, speed relative to limits) and lateral control (e.g., lane deviation, stability). Both baselines were evaluated across specific Operational Design Domains (ODDs), primarily limited-access highways within the interstate system, and included data from certain Canadian corridors. Key findings revealed that 29% of crashes involved driver error, indicating a significant opportunity for ADS to improve safety. The most frequent crash types were conflicts with objects in the roadway and vehicles in adjacent lanes, each occurring at a rate of 28.4 crashes per 100 million VMT. Notably, in the majority of crashes involving Class 8 trucks with interior cameras, drivers were not engaged in secondary tasks or negative driving behaviors. Instead, the most common unexpected events were debris hitting the vehicle or animals/pedestrians entering the highway. Furthermore, deceleration rates and shorter following distances increased in lower speed limit zones, reflecting tighter traffic conditions. This highlights the complexity of designing ADS that can exhibit "roadmanship" and interact smoothly with human drivers, rather than merely complying with posted rules. The study concludes that both lagging measures (crashes) and leading measures (maneuvers) are essential for objectively assessing ADS safety. These baselines provide developers, regulators, and policymakers with reproducible, non-proprietary metrics to compare ADS performance against human drivers. The research tools used to derive these baselines are available to the public, facilitating future verification and the evaluation of how ADS adoption impacts overall transportation safety.

Key finding

Driver error represents 29 percent of crashes in commercial motor vehicles, and the study provides the first public baselines for comparing human and automated driving performance on interstate highways.

Methodology

dataset

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• naturalistic crash near crash
• adas effectiveness
• exposure measurement

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes, observational prevalence
• Methodological Resource: dataset resource