Feasibility of Driver Judgment as Basis for a Crash Avoidance Database

Smith, David L.; Najm, Wassim G.; Glassco, Richard A. · 2002 · Crossref

DOI: 10.3141/1784-02

archive: archived pipeline: cataloged

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

Summary

This paper addresses the need for a standardized crash avoidance database structure to support the development of Intelligent Vehicle Initiative countermeasures. The authors propose a framework based on driver judgments that categorizes driving scenarios into four distinct conflict states: low risk, conflict, near-crash, and crash. Each state corresponds to specific countermeasures, such as advisory warnings for conflicts and crash mitigation for imminent crashes. The primary research questions investigated were whether these driving states can be reliably quantified and whether the quantified states can form the basis of a useful database. The study focused on the specific scenario of a following vehicle braking to avoid a rear-end collision with a stopped lead vehicle, a common high-priority crash type. The methodology involved estimating state boundaries using data from controlled experiments. The crash state boundary was derived from rear-end crash data collected in the Iowa Driving Simulator (IDS), where ten subjects responded to a stopped vehicle revealed after a distraction task. The boundary was approximated by the equation Range = (Range Rate)² / (2 × 0.65g), reflecting the deceleration limits of drivers who crashed or steered off-road. The conflict and near-crash boundaries were estimated using data from the GM-Ford Crash Avoidance Metrics Partnership (CAMP) test track study, which involved 108 subjects performing "last-second" comfortable and hard braking maneuvers at speeds of 48, 72, and 97 km/h. To validate the feasibility of this structure, the authors compared these controlled estimates against on-road naturalistic driving data collected by NHTSA’s Vehicle Research Test Center, which recorded braking responses of following vehicles to stopped lead cars using radar and video. The results demonstrated that the proposed database structure is feasible. The naturalistic data confirmed the rough quantitative locations of the state boundaries identified in the controlled experiments. Specifically, 56% of naturalistic braking events occurred in the low-risk state, while a smaller percentage initiated braking in the conflict state, aligning with the CAMP test track findings. The analysis of simulator data identified three distinct crash epoch scenarios: drivers braking at onset, drivers braking after entering the crash state, and drivers who braked and then steered off-road. The study further illustrated the utility of the database by evaluating warning algorithms, showing that time-to-collision (TTC) based alerts at 3 and 5 seconds were often perceived as "too early" or "too late" relative to driver expectations, highlighting the need for kinematic-based rather than purely time-based warning criteria. The significance of this work lies in its provision of a robust data structure that allows for the integration of disparate data sources, including naturalistic driving, test track, and simulator studies. This unified approach enables the evaluation of crash countermeasures, identification of data gaps, and guidance for experimental design. The authors conclude that while the current boundaries are rough estimates, the method is reliable enough to support further development. They recommend extending this approach to other crash scenarios, such as lane changes and run-off-road events, and developing automated processes to convert raw multimedia data into searchable discrete variables for broader safety analysis.

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.

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

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

• naturalistic crash near crash
• crash typology
• driver post crash behavior
• telematics crash prediction
• pre crash contributing factors

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
• Methodological Resource: dataset resource
• Theoretical Contribution: computational model