Platform Development
The portable driving environment simulator platform is an in-house development designed to simulate Indian traffic conditions using the CARLA open-source simulator.
Features:
- A three-screen, high-resolution display providing a 135-degree forward visual field-of-view
- A cockpit
- Fully operational driving controls (steering wheel and pedals)
- Simulator software package (CARLA)
The Driving Simulator platform initially started with a single monitor connected to a steering wheel and a machine. It was later upgraded to a three-monitor setup in a workspace. Eventually, a fully functional simulator platform was developed, featuring a triple-monitor system and an accurately designed seat to provide a realistic driving experience.
Simulated Critical Traffic Scenarios
Driving Levels: Elementary, Intermediate, Advanced
- We simulate a range of roadway environments, including urban, suburban, rural, and motorway settings, as well as custom events such as hazardous situations (critical and dangerous driving scenarios) that cannot be staged in the real world due to safety concerns. Additionally, various weather conditions, illumination levels, and traffic situations are simulated.
Data Capture and Collection Setup
- Data Capture Setup : The data capture setup is designed to comprehensively analyze driver behavior and gaze patterns using multiple perspectives.
- GoPro Hero 8 Camera: Positioned to capture a clear view of the driver while driving, providing insights into posture, hand movements, and overall behavior.
- Meta Aria Glasses: Used to track and analyze the driver's gaze patterns, helping to understand focus, attention shifts, and reactions in different driving scenarios.
- Screen Recording View: Captures the road ahead, recording traffic conditions, obstacles, and environmental factors that influence the driver's decisions.
Data Collection
We collected driving data with 24 drivers to analyze their behavior and gaze patterns while wearing Aria glasses. The study included various levels of driving difficulty, different weather conditions (sunny, clear, and rainy), and diverse routes.
Each driver participated across three skill levels—elementary, intermediate, and advanced— while navigating through selected towns, each presenting unique driving challenges.
Drivers provided detailed feedback on their experiences, highlighting the challenges faced, ease of navigation, and overall engagement. Many described the study as a driving school- like experience, combining fun with skill-building, making it both an educational and enjoyable process.
Gaze-Based Automatic Segmentation
In this work we focus on utilizing gaze-based technology for automatic segmentation. By tracking the driver’s eye movements, the system identifies areas of interest, analyzes driving behavior, and assesses potential risks in real time. This approach enhances safety by providing insights into driver attention and detecting hazardous situations.
Segmentation of traffic objects using gaze data