Development and Experimental Validation of an Intelligent Camera Model for Automated Driving

Area of Science:

• Automotive Engineering
• Perception Systems
• Sensor Modeling

Background:

• Virtual testing of advanced driver assistance systems (ADAS) and automated driving (AD) functions necessitates realistic perception sensor models.
• Simulating limitations and measurement errors of real-world sensors is crucial for generating valid data for ADAS/AD system testing.

Purpose of the Study:

• To introduce a novel sensor modeling approach for automotive perception sensors, focusing on position measurement errors.
• To enable realistic simulation of sensor limitations for improved virtual testing of ADAS/AD.

Main Methods:

• A novel approach combining kernel density estimation with regression modeling was developed.
• The model is designed for automotive perception sensors providing object-level position estimations.
• A Mobileye 630 camera was used for demonstration and evaluation, with data collected on a Hungarian highway.

Main Results:

• The developed sensor model achieved a pointwise position error of 9.60% in the lateral and 1.57% in the longitudinal direction.
• The model effectively captured the natural scattering and deviations in sensor measurement outputs.
• The approach demonstrated successful modeling of position measurement errors for automotive perception sensors.

Conclusions:

• The novel sensor modeling approach provides a realistic simulation of automotive perception sensor errors.
• This method enhances the virtual testing and validation capabilities for advanced driver assistance and automated driving functions.
• The approach is applicable to various automotive perception sensors providing object-level position data.