Exploring Driver Responses to Unexpected and Expected Events Using Probabilistic Topic Models
DOI: 10.17077/drivingassessment.1661
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 investigates how driver expectations influence lateral vehicle control responses to sustained hazards, specifically crosswinds. While previous research has examined expectations in discrete events like collision avoidance, less is known about how drivers adapt to continuous disturbances. The authors hypothesize that traditional aggregate metrics (e.g., maximum steering angle) may fail to capture the temporal evolution of steering behavior. To address this, they applied probabilistic topic modeling, a machine learning technique originally used for text analysis, to granular time-series driving data. The researchers analyzed data from 41 participants in a fixed-base driving simulator. Participants drove on rural highways and encountered repeated high-intensity lateral crosswind gusts. The first exposure was defined as an "unexpected" event, while the final exposure in the block served as an "expected" event, assuming drivers had developed ad-hoc expectations through prior trials. Steering wheel angle and rate of change were recorded at 5 Hz. These numeric values were converted into symbolic "words" using the Symbolic Aggregate Approximation (SAX) algorithm, binning data into percentiles over 200-millisecond windows. Each trial’s sequence of symbols formed a "document" for topic modeling. Using the stm package in R, the authors identified four distinct steering topics. Topic 1 characterized large rates of steering change, while Topic 2 represented negligible or slow steering rates. Results showed that expected events contained a higher proportion of Topic 1, indicating drivers made continuous, faster steering corrections when anticipating the crosswind. Conversely, unexpected events showed a higher proportion of Topic 2, suggesting fewer abrupt movements initially, with larger corrections appearing later as drivers began to adapt. To validate the method, the authors built Random Forest predictive models. The topic model approach achieved an Area Under the Curve (AUC) of 0.67 in distinguishing expected from unexpected trials, outperforming traditional models using aggregated maximum values, which achieved an AUC of 0.6. The findings demonstrate that probabilistic topic modeling effectively captures nuanced differences in driver behavior that aggregate statistics miss. Drivers adapt to expected crosswinds with smoother, continuous corrections, whereas unexpected events elicit more hesitant or abrupt responses. This approach offers a viable method for analyzing complex driving behaviors across multiple timescales, with implications for developing driver models and improving vehicle safety technologies such as crosswind assist systems and electronic stability control.
Key finding
Drivers exhibit continuous, faster steering corrections during expected crosswind events compared to the fewer, more abrupt movements observed during unexpected events, and probabilistic topic modeling classifies these behavioral differences more effectively than traditional aggregated metrics.
Methodology
simulator
Sample size: 41
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 author_sweep_intake on 2026-05-28.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | canonical_url | — | — | 1 | 2026-06-04 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | — | — | — | 1 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
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.
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: behavioral performance data
- Methodological Resource: tool software
- Theoretical Contribution: computational model