A framework for testing independence between lane change and cooperative intelligent transportation system

Area of Science:

• Transportation Engineering
• Intelligent Transportation Systems
• Road Safety

Background:

• Cooperative Intelligent Transportation Systems (C-ITS) are increasingly deployed globally to enhance road safety.
• Evaluating the safety benefits of C-ITS, particularly their interaction with driver behavior like lane changes during critical events, is essential.
• Existing methods for lane change detection require multiple sensors, which may not be available in all C-ITS Field Operational Tests (FOTs).

Purpose of the Study:

• To propose and validate a novel framework for assessing the dependency between lane changes and C-ITS warnings using limited sensor data.
• To enable the evaluation of C-ITS safety benefits in the context of driver actions during safety-critical events.
• To address the challenge of lane change detection in C-ITS FOTs where vehicles are equipped only with C-ITS and Inertial Measurement Units (IMU).

Main Methods:

• Development of a framework utilizing IMU data to detect lane changes via a trained random forest classifier.
• Construction of a 2x2 contingency table using the classifier's output probabilities for C-ITS and control conditions.
• Application of a permutation test to statistically evaluate the independence between lane changes and C-ITS warnings during safety events.

Main Results:

• Successfully trained a random forest classifier using IMU data to accurately detect lane changes.
• Demonstrated the framework's capability to generate contingency tables and perform permutation tests for dependency analysis.
• Established a method to quantify the relationship between driver's lane change behavior and C-ITS warning system activation.

Conclusions:

• The proposed framework effectively tests the dependency between lane changes and C-ITS warnings using only IMU data.
• This approach provides a viable solution for evaluating C-ITS safety benefits in FOTs with limited sensor configurations.
• The methodology contributes to a better understanding of driver-system interaction in intelligent transportation environments.