Improved Driving Behaviors Prediction Based on Fuzzy Logic-Hidden Markov Model (FL-HMM)
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 accurately predicting human driving behaviors to enhance Advanced Driver Assistance Systems (ADAS). Since most traffic accidents result from driver misoperations, reliable prediction of maneuvers such as lane changes is critical for providing timely warnings. The authors identify that standard Hidden Markov Models (HMM) often yield suboptimal recognition results due to the variability of individual driver decisions and differing driving scenarios. To overcome this, the study proposes a Fuzzy Logic-Hidden Markov Model (FL-HMM) approach that combines scenario-specific HMMs with fuzzy logic classification and optimizes model parameters using a genetic algorithm. The methodology involves a two-stage prediction process. First, Fuzzy Logic (FL) classifies the current driving scene into "Very Safe," "Safe," or "Dangerous" categories based on two inputs: the distance to the vehicle in front and the Time to Collision (TTC). Trapezoidal membership functions define these fuzzy sets. Second, a specific HMM is selected and trained for the identified scenario. The HMM models three hidden states: lane keeping, left lane change, and right lane change. Observations for the scene-based HMM include TTC values relative to surrounding vehicles, while a supplementary operation-based HMM uses driver inputs such as steering angle, accelerator position, and brake pressure. A prefilter quantizes these signals into observation sequences. The final prediction fuses probabilities from both HMMs using a weighted sum. To optimize the FL thresholds, HMM prefilters, and fusion weight, the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed to maximize accuracy (ACC) and detection rate (DR) while minimizing the false alarm rate (FAR). Experiments were conducted using a driving simulator with seven participants performing overtaking maneuvers on a four-lane highway. Data from 40-minute drives were used for training, and 10-minute drives for testing. The results demonstrate that the optimized FL-HMM significantly outperforms alternative methods, including Artificial Neural Networks (ANN), Support Vector Machines (SVM), and standard HMMs. The proposed model achieved an average ACC, DR, and (1-FAR) values exceeding 80%. Specifically, the FL-HMM improved the detection rate to 82% and generated the lowest false alarm rate among all tested models. Receiver Operating Characteristic (ROC) analysis confirmed the superior performance of the FL-HMM. Additionally, the model successfully predicted lane changes approximately 1.8 to 3.9 seconds before the maneuver execution, providing sufficient lead time for assistance systems. The significance of this work lies in its ability to improve the reliability of driving behavior prediction by accounting for both environmental context and driver-specific operations. By integrating fuzzy logic for scenario distinction and using NSGA-II for parameter optimization, the FL-HMM offers a robust framework for ADAS development. The findings suggest that optimizing prediction models through multi-objective optimization can substantially enhance detection rates and reduce false alarms, thereby increasing the safety and effectiveness of automated driver assistance technologies.
Key finding
The optimized FL-HMM approach significantly outperforms standard HMM, ANN, and SVM models in predicting driving behaviors, achieving accuracy and detection rates above 80% with the lowest false alarm rate.
Methodology
simulator
Sample size: 7
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).
- Theoretical Contribution: computational model