Online intention recognition applied to real simulated driving maneuvers
DOI: 10.1109/cogsima49017.2020.9216115
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 challenge of real-time human driving intention recognition, a critical component for Advanced Driver Assistance Systems (ADAS) aimed at enhancing traffic safety. Motivated by the high incidence of accidents caused by driver misoperation, the authors propose a method to predict driver maneuvers—specifically lane keeping (LK), lane changing to the left (LCL), and lane changing to the right (LCR)—to enable timely warnings or assistance. The study introduces a novel fuzzy Random Forest (fuzzy-RF) approach that integrates fuzzy logic with machine learning to handle the vagueness inherent in driving behaviors. The methodology utilizes data collected from a driving simulator featuring a two-way highway scenario with three lanes. Eight licensed drivers participated, generating data for both offline training and online testing. The model employs 24 input variables categorized into six types, including ego-vehicle dynamics (velocity, steering angle), surrounding vehicle states (distance, velocity, time-to-collision), and discrete indicators (lane number, turn signals). To improve classification performance, input signals are quantified into fuzzy sets using trapezoidal membership functions. These membership functions are automatically generated via a fuzzy density clustering method called FN-DBSCAN, eliminating the need for manual parameter tuning. The core classifier is a Random Forest algorithm trained on these fuzzified inputs. A key innovation in this work is the redefinition of lane-changing intention labels using the "difference between road angle and heading angle of the ego-vehicle," with thresholds of 0.05°, 0.2°, and 0.5° tested against a reference time-based label. Experimental results demonstrate high efficacy in intention recognition. During the offline training phase, validated via 10-fold cross-validation, the model achieved accuracy rates exceeding 98% and detection rates above 92%, with false-alarm rates below 4%. In the online test phase, where models were applied to real-time driving data from the same participants, the average accuracy remained robust, ranging from 95.0% to 95.8% depending on the labeling threshold used. Detection rates in the online phase averaged between 75.3% and 76.8%, while false-alarm rates stayed between 12.1% and 13.2%. The study found that the new angle-based labeling definition performed comparably to or better than traditional time-based definitions, particularly when using a 0.5° threshold. The significance of this work lies in its successful application of an automated, data-driven fuzzy-RF framework for real-time driver behavior prediction. By automatically generating membership functions and utilizing novel geometric parameters for labeling, the approach reduces manual effort and improves adaptability to individual driving styles. The high accuracy rates suggest that such systems can reliably support ADAS functions, such as providing lateral steer-by-wire or longitudinal brake-by-wire assistance, thereby contributing to safer assisted and conditionally automated driving modes.
Key finding
The fuzzy Random Forest model achieved offline training accuracy greater than 98% and online test accuracy greater than 91% for recognizing driving intentions.
Methodology
simulator
Sample size: 8
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 | — | — | 7 | 2026-06-06 |
| 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).
- Methodological Resource: tool software
- Theoretical Contribution: computational model