Toward Human-Centered Design of Automated Vehicles: A Naturalistic Brake Policy

Rahmati, Yalda; Abianeh, Arezoo Samimi; Tabesh, Mahmood; Talebpour, Alireza · 2021 · Crossref

DOI: 10.3389/ffutr.2021.683223

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the safety challenges of Connected and Automated Vehicles (CAVs) in mixed traffic environments, specifically focusing on rear-end collisions at intersections. The authors hypothesize that these crashes stem from a mismatch between CAV braking patterns and human drivers’ expectations. Current safety standards rely on miles driven and crash counts, which fail to capture the nuances of human-CAV interaction. To resolve this, the research aims to characterize human braking behavior and develop a CAV braking policy that aligns with human expectations while maintaining safety. The methodology involves a field experiment conducted at the Texas A&M University RELLIS Campus using an autonomous Chevrolet Bolt EV. Data was collected from eight drivers across four days, recording instantaneous speeds via high-precision GPS/IMU systems. The dataset comprises 213 stopping scenarios at intersections, categorized into "free-flow" (no preceding vehicle) and "car-following" (preceding vehicle present) conditions. The authors analyzed the speed profiles during the 10 seconds preceding a full stop. They employed supervised learning techniques to classify these braking patterns. First, they extracted summary features (average speed, maximum/minimum acceleration) and trained Artificial Neural Networks (ANN) and Support Vector Machines (SVM). Second, they applied univariate time series classifiers, including K-Nearest Neighbor (KNN), Time Series Forest (TSF), and Proximity Forests (PF), to the raw speed data. The results demonstrate significant systematic differences in human braking trajectories between free-flow and car-following conditions. Car-following profiles exhibited consecutive deceleration and acceleration maneuvers with lower average speeds, whereas free-flow profiles were smoother. The ANN classifier achieved the highest performance with an F1-score of 0.95, followed by SVM (0.91), confirming that human braking strategies are distinct and distinguishable based on traffic context. Time series classifiers also performed well, with TSF achieving an F1-score of 0.89. These findings validate the hypothesis that human drivers adapt their braking behavior based on the presence of other vehicles, a nuance often missing in standard CAV logic. The significance of this work lies in its contribution to human-centered CAV design. By quantifying the mismatch between human and automated braking behaviors, the study provides a foundation for developing naturalistic brake policies. The authors propose using Markovian decision modeling to design CAV braking profiles that are compatible with human expectations. This approach aims to prevent rear-end collisions caused by unexpected CAV maneuvers, thereby enhancing safety in mixed traffic environments where human-driven vehicles will predominate for decades. The study offers a scalable framework for assessing CAV safety beyond simple crash metrics, facilitating the development of higher levels of vehicle automation.

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.

StageOutcomeToolModelPromptAttemptsCompleted
discover success Crossref 1 2026-06-07
archive success canonical_url 1 2026-06-09
extract success pdftotext 2 2026-06-09
clean success clean 1 2026-06-09
chunk success chunk 1 2026-06-09
embed success embed Qwen/Qwen3-Embedding-8B 1 2026-06-09
promote success 1 2026-06-07
summarize success llm qwen3.6-27b-prismaquant summ-v5 1 2026-06-09
tag success vector_similarity 8 2026-06-11
verify success 1 2026-06-09

Summary generated by qwen3.6-27b-prismaquant on 2026-06-09; verification: verified.

Topics

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

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).