An Intelligent System Proposal for Providing Driving Data for Autonomous Drive Simulations

Abstract

Simulation technology is being used to reduce the costs of development and testing processes in autonomous driving studies. Autonomous systems gaining driving experience in simulation is faster and more cost-effective than real-world work. However, systems developed through simulation are becoming increasingly distant from reality, and it is uncertain how these systems will respond to situations that may occur in the real world. Therefore, having a simulation environment that is close to reality in which an autonomous driving system will be developed will contribute to the more efficient operation of the systems developed in the simulation environment in the real world. To create a more realistic driving simulation, a large amount of driving data from the real world is needed. In this study, a smart system has been developed that analyzes various driving videos and produces the necessary data for the simulation. The proposed system calculates the position change of the vehicles in each frame of the video. Depending on the position change, the vehicle's speed, acceleration, total displacement, and maneuvering style are revealed. This allows for data on different driving styles to be obtained, transferred to the simulation environment, and enables the simulated vehicles to behave like real drivers. In addition, the vehicles and stationary objects in the video are detected by an intelligent system, and the vehicles in consecutive frames are matched and scaled to made frames identical. By this way we eliminate the effects of extrinsic parameters such as camera transformation, angle change, and zoom in/zoom out to produce the required data. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.