Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Maneuver-based Urban Driving Dataset and Model for Cooperative Vehicle Applications

Behrad Toghi, Divas Grover, Mahdi Razzaghpour, Rajat Jain, Rodolfo Valiente, Mahdi Zaman, Ghayoor Shah, Yaser P. Fallah

TL;DR

This paper introduces D2CAV, a real-world, maneuver-labeled driving dataset collected with the Ford OpenXC platform to support cooperative vehicle applications in mixed-autonomy urban settings. It details data collection, labeling, and synchronization of CAN and GPS streams, and evaluates two classifiers for maneuver recognition, finding Random Forest to be the stronger performer. The work demonstrates the value of maneuver-aware data for model-based communication and predictive decision-making in connected and automated vehicle networks, highlighting practical implications for safety and coordination. Overall, D2CAV provides a foundation for maneuver-level analysis and prediction in urban driving scenarios, enabling improved V2V and human-agent collaboration.

Abstract

Short-term future of automated driving can be imagined as a hybrid scenario in which both automated and human-driven vehicles co-exist in the same environment. In order to address the needs of such road configuration, many technology solutions such as vehicular communication and predictive control for automated vehicles have been introduced in the literature. Both aforementioned solutions rely on driving data of the human driver. In this work, we investigate the currently available driving datasets and introduce a real-world maneuver-based driving dataset that is collected during our urban driving data collection campaign. We also provide a model that embeds the patterns in maneuver-specific samples. Such model can be employed for classification and prediction purposes.

A Maneuver-based Urban Driving Dataset and Model for Cooperative Vehicle Applications

TL;DR

This paper introduces D2CAV, a real-world, maneuver-labeled driving dataset collected with the Ford OpenXC platform to support cooperative vehicle applications in mixed-autonomy urban settings. It details data collection, labeling, and synchronization of CAN and GPS streams, and evaluates two classifiers for maneuver recognition, finding Random Forest to be the stronger performer. The work demonstrates the value of maneuver-aware data for model-based communication and predictive decision-making in connected and automated vehicle networks, highlighting practical implications for safety and coordination. Overall, D2CAV provides a foundation for maneuver-level analysis and prediction in urban driving scenarios, enabling improved V2V and human-agent collaboration.

Abstract

Short-term future of automated driving can be imagined as a hybrid scenario in which both automated and human-driven vehicles co-exist in the same environment. In order to address the needs of such road configuration, many technology solutions such as vehicular communication and predictive control for automated vehicles have been introduced in the literature. Both aforementioned solutions rely on driving data of the human driver. In this work, we investigate the currently available driving datasets and introduce a real-world maneuver-based driving dataset that is collected during our urban driving data collection campaign. We also provide a model that embeds the patterns in maneuver-specific samples. Such model can be employed for classification and prediction purposes.

Paper Structure

This paper contains 10 sections, 1 equation, 9 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. In-field Data Collection Campaign
  4. OpenXC Platform
  5. Classification Methodology
  6. Random Forest Classifier
  7. Support Vector Machine
  8. Analysis & Results
  9. Concluding Remarks
  10. Acknowledgement

Figures (9)

  • Figure 1: A comparison between a pair of given u-turn and left-turn maneuvers and their pattern in the steering angle space and time-domain representation (sampled with 10 Hz rate).
  • Figure 2: Imbalanced dataset distribution for containing maneuvers
  • Figure 3: Illustration of a given right-turn maneuver in terms of 4 features, represented in time-domain (sampled with 10 Hz rate).
  • Figure 4: A sample view of the driving path during the data collection campaign in the UCF campus (map courtesy of Google Earth™)
  • Figure 5: Maneuver classification system architecture: Random Forest classifier and Support Vector Machine
  • ...and 4 more figures