Driver-in-the-Loop for computer-vision based ADAS testing

Abstract

Driver safety is a main concern nowadays. Hence, a real need for systems like Advanced Driver Assistance Systems (ADAS) has arisen. With this need to such systems, many challenges come up. These challenges are including the time consumption of testing of these systems as It needs to be tested in different weather conditions. Besides, the difficulty of covering all the possible scenarios while testing these systems and the carbon dioxide emission from cars in the process of testing the system. Hence, the purpose of the proposed paper is to implement a simulation-based driver in the loop simulation-based system for indoor testing of computer vision based ADAS. A simplified tricycle kinematics model is implemented for this purpose by using Matlab/Simulink. A vehicle virtual environment is implemented in unreal engine environment by MATLAB platform. The interaction of the driver with the virtual car in this virtual environment is made by a simplified hardware for the car. A feedback is taken from the driver through sensors connected to an Arduino microcontroller. Arduino microcontroller interpret the driver feedbacks and feed it as an inputs to the vehicle model to move the car in the virtual environment. The proposed system is tested for computer vision lane detection, departure, and change algorithms. These algorithms relay mainly on Hough transform. The proposed system is tested in real time and the results obtained is very similar to the results obtained from testing a computer vision based ADAS on a real street.