Modelling stop intersection approaches using Gaussian processes
DOI: 10.1109/itsc.2013.6728466
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the limitation of Advanced Driving Assistance Systems (ADAS) that assume uniform driver behavior, despite significant individual variations in deceleration patterns at road intersections. Since over 40% of car accidents occur at intersections and are largely caused by driver errors, the authors propose a method to model individual driver velocity profiles as vehicles approach stop intersections. The goal is to enable ADAS to detect deviations from a driver’s typical behavior, thereby identifying potential risks more accurately than generic models. The study employs Heteroscedastic Gaussian Processes (GP), a machine learning technique for non-linear regression, to model these velocity profiles. Standard GP assumes constant noise, which is insufficient for capturing the variability in human driving. The authors utilize a heteroscedastic variant where variance is input-dependent, allowing the model to account for greater uncertainty in specific phases of the approach, such as the onset of deceleration. Additionally, the method incorporates input noise correction to handle inaccuracies in GPS localization data. The experimental design involved collecting real-world data from a passenger vehicle driven by four different users. Each driver performed ten runs at five distinct cruise speeds (30, 40, 50, 60, and 70 km/h) approaching a stop intersection. Data recorded included vehicle velocity from the CAN bus and position relative to the intersection from a GPS receiver. The results demonstrate that Gaussian Processes effectively capture individual driving patterns, producing smooth, continuous mean velocity profiles with appropriate confidence intervals. The heteroscedastic approach successfully estimated variance based on training samples, revealing that drivers exhibit significant uncertainty regarding when they begin decelerating and the rate of deceleration. In contrast, homoscedastic GP models would have required arbitrary parameter tuning, leading to over- or under-estimated variance. The models generated for individual drivers were distinct from generic average profiles. The authors conclude that modeling individual patterns allows for more precise risk assessment, as deviations from a driver’s learned profile can indicate dangerous behavior or external influences like fatigue, offering a significant improvement over current ADAS approaches that rely on arbitrary, driver-independent thresholds.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-25 |
| archive | success | unpaywall | — | — | 2 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-26 |
| chunk | success | chunk | — | — | 1 | 2026-06-26 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-26 |
| enrich | success | openalex | — | — | 1 | 2026-06-26 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-26 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; 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).
- Theoretical Contribution: computational model