Vehicle Automation and Transportability of Crash Modification Factors

Davis, Gary; Gao, Jingru · 2019 · ROSA P / University of Minnesota

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 report investigates the feasibility of adapting Crash Modification Factors (CMFs) from the Highway Safety Manual (HSM) to future roadway environments characterized by the presence of automated vehicles (AVs). Current HSM tools rely on empirical data reflecting historical driver and vehicle conditions, raising concerns about their validity as AV market share increases. The authors address the problem of CMF "transportability"—the ability to transfer causal effects estimated in one situation to another—by applying formal causal inference methods developed by Judea Pearl and Elias Bareinboim. The study aims to determine if existing safety research can be leveraged to predict safety outcomes in mixed-traffic environments without requiring new, costly empirical studies. The methodology employs graphical causal models and selection diagrams to derive transport formulas for CMFs. The authors define two theoretical scenarios: one where a safety treatment affects pedestrian behavior while vehicle behavior changes due to AVs, and another where the treatment affects driver behavior which is also altered by AVs. Using Pearl’s do-calculus, they derive equations that allow the recalibration of an original CMF based on the distribution of specific variables (such as driver caution) in the new situation. To demonstrate practical application, the authors develop a probabilistic causal model for Pedestrian Hybrid Beacons (PHBs). They simulate vehicle-pedestrian encounters using data from the Pedestrian Crash Data Study (PCDS) and other sources to estimate parameters such as driver reaction times, braking rates, and pedestrian behavior. This explanatory model is then used to assess how the PHB CMF would change in a hypothetical scenario where vehicles possess autonomous braking capabilities. The findings indicate that transportability analysis is theoretically possible provided a causal mechanism explaining the CMF’s operation can be established. In the computational examples, the authors successfully recalibrated CMFs for PHBs. For instance, in a scenario where AVs increase the proportion of careful drivers, the recalibrated CMF showed a greater crash reduction effect than the original estimate, demonstrating that the safety benefit of the beacon interacts with vehicle automation levels. The simulation results provided specific distributions for driver reaction times and braking rates, validating the causal model’s ability to represent crash-generating processes. The study confirms that if the relevant circumstances (e.g., changes in driver behavior due to AVs) are identified and quantified, existing CMFs can be adjusted to reflect new conditions. The significance of this work lies in providing a rigorous framework for maintaining the relevance of safety engineering tools amidst technological change. By establishing that CMFs can be transported via causal modeling, the report suggests that the substantial investment in historical safety data need not become obsolete with the advent of AVs. Instead, engineers can use causal explanations to adjust existing estimates, bridging the gap between current empirical knowledge and future roadway conditions. This approach offers a pathway for rational decision-making in the short-to-medium term, allowing for safety predictions in mixed-traffic environments before extensive new empirical data becomes available.

Key finding

Transportability analysis allows for the recalibration of existing crash modification factors to new conditions, such as the presence of automated vehicles, provided a causal model explaining the treatment mechanism is available.

Methodology

theoretical

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_with_issues.

Topics

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

• incidence prevalence
• induced exposure
• causation analyses
• pre crash contributing factors
• perception reaction time
• telematics crash prediction

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).

• Applied Guidance: countermeasure evaluation
• Empirical Findings: crash risk outcomes
• Theoretical Contribution: computational model