Intelligent multi-sensor fusion and anomaly detection in vehicles via deep learning
Bibliographic record
Abstract
Deep learning techniques are predominantly used to identify vehicular events such as harsh cornering, harsh braking, and rapid acceleration by analyzing signal data. However, deploying deep learning models demands high-quality, large-scale data, and the processes of data acquisition, labeling, extraction, and processing are often overlooked in the literature. In this article, we focus on detailed dataset creation, including labeling and feature analysis, alongside the development of AI models. Real-time data collection is conducted on experimental roads using numerous vehicles equipped with AI-enabled edge units. The raw data collected, however, is unsuitable for training deep learning models due to redundant features, noisy attributes, and a lack of labeled anomalous events. To address this, we employ multiple preprocessing and postprocessing techniques to generate high-quality datasets, analyzing the specific impacts of each signal feature on anomalous events. Since real-time collected data lacks labels, a thorough labeling process is required for each data point. An autoencoder-based labeling process is applied to the final dataset, where the autoencoder detects and labels anomalous behaviors based on data timestamps. Following the labeling, a hybrid deep learning model incorporating Convolutional Neural Networks (CNN), Long Short Term Memory (LSTM), attention, and Fully Connected Neural Networks (FCDNN) layers is trained and tested for detecting anomalous driving events. The results demonstrate that the proposed method outperforms the state-of-the-art solutions by reaching high accuracy rates: 99.69% for harsh cornering events and 98.24% for rapid acceleration and harsh braking events, with corresponding F1 scores of 90.14% and 81.22%, respectively.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".