Core Concepts
Accurate and early prediction of truck overtakes is crucial for safe and efficient driving. This study investigates the detection of truck overtakes from real-world CAN bus data using machine learning techniques.
Abstract
The authors present an ongoing work on overtake detection in trucks using real-world CAN bus data. They employ three classifiers - Artificial Neural Networks (ANN), Random Forest (RF), and Support Vector Machines (SVM) - to perform the task.
Key highlights:
The database consists of CAN data from 3 real operating trucks, with labels for overtake and no-overtake events obtained by manual video annotation.
The CAN signals analyzed include vehicle speed, acceleration, distance to the vehicle ahead, and turn indicator status, among others.
The classifiers are evaluated on their ability to predict overtakes up to 10 seconds before the event, using a sliding window approach.
The results show that the prediction scores for the overtake class tend to increase as the overtake trigger is approached, while the no-overtake class remains more stable or oscillates.
The classifiers achieve high recall/TPR (≥93%) in detecting overtakes, but the accuracy in classifying no-overtakes is suboptimal (TNR typically 80-90% and below 60% for one SVM variant).
The authors combine two classifiers (RF and linear SVM) by averaging their output scores, which improves the no-overtake classification (TNR ≥92%) at the expense of reducing overtake accuracy (TPR), but keeps the latter above 91% near the overtake trigger.
The fusion approach provides a more balanced performance in detecting both classes compared to individual classifiers.
Stats
The vehicle speed is more than 50 km/h.
The distance to the vehicle ahead is less than 200 m.
The relative speed difference between the vehicle and the left wheel is more than 0.1 km/h.
Quotes
"Safe overtakes in trucks are crucial to prevent accidents, reduce congestion, and ensure efficient traffic flow, making early prediction essential for timely and informed driving decisions."
"Knowing the driver's intention is an integral part of the system, to determine if the ADAS should activate, providing opportune aids or alerts, or even overriding the driver's inputs."