Scenario based Simulation Testing of Autonomous Vehicle using Malaysian Road

Abstract

Autonomous vehicles are known as one of the promising technologies to decrease the traffic congestions and road accidents. Generally, autonomous vehicles have been tested to adopt automatically while driving on multiple road conditions with different types of traffic situations via simulation analysis. However, most of the autonomous vehicle simulation testing is conducted in the developed countries environment model and mostly on highway driving scenarios with less pedestrian’s movements. Meanwhile in Malaysia, most of the automotive researchers have initiated researches related to autonomous vehicle based on controlled environment only. The researchers explore this research on theoretical based simulation and then directly implemented in actual vehicle for on road testing. This kind of testing not sufficient enough to optimize the performance of autonomous vehicle based on Malaysian environment. To further enhance the capability of autonomous vehicle in Malaysia, a scenario-based simulation testing is required using virtual testing platform in order to adopt with Malaysian road and traffic environment before on-road testing. As for testing, University of Nottingham Malaysia has been selected as the location for testing. Meanwhile, a deep learning method using YOLOv3 is used in this study to classify critical from recorded video data and used the data for scenario generation and testing autonomous vehicle performance.