Creating Driving Scenarios from Recorded Vehicle Data for Validating Lane Centering System in Highway Traffic

Traffic Jam Assist system helps the driver to follow the preceding vehicle automatically with a predefined time interval in a dense traffic condition while controlling steering for keeping the current lane. This requires a combination of longitudinal control with adaptive cruise control with stop & go and lateral control with lane following control. In order to validate the performance of the automated driving system, it normally requires tens of thousands of miles of driving the vehicle on the road. This is a time-consuming process and adoption of the virtual simulation tool is inevitable to reduce the development time and enhance the robustness. The challenge is how to preciously reproduce the real-world driving scenarios with the virtual driving environment. This paper introduces a workflow to create the driving scenarios from HERE HD map. The HERE HD map provides the road and lane information for the recorded GPS data. MathWorks Automated Driving toolbox™ is used to create the driving scenario based on the extracted road network information. The ego and non-ego vehicles are added to the driving scenario using the recorded vehicle data and onboard vision and radar sensors. A simple localization technique is used to improve the accuracy of ego vehicle's GPS position based on a lane detection sensor. A probabilistic sensor model is applied to synthesize the vision and radar sensor detections. Then, the driving scenario and sensor model are integrated into the closed-loop Simulink model consisting of ego vehicle dynamics and traffic jam assist controller. With this approach, the real-world driving scenario can be reproduced in a virtual simulation environment with closed-loop Simulink model. This virtual simulation environment can be efficiently and conveniently used for validating the automated driving system without spending too much time on the road.