Integrated Framework of Autonomous Vehicle with Traffic Sign Recognition in Simulation Environment

This study proposes an integrated framework for autonomous vehicle research in a simulated environment. There has been much research on the simulations of vehicles and environments, the recognition of traffic signs, and the lateral/longitudinal controls of a vehicle. Yet, not many systems are available for autonomous vehicle researchers to test and improve their algorithms in a realistic simulated environment with sensor suites in their own car model. We aim to provide an integrated framework for a programmable autonomous vehicle in a simulated environment. The simulated vehicle is capable of autonomous driving with traffic sign recognition using deep learning-based object detection capability as well as lateral and longitudinal controllers. To show the feasibility of the proposed system, we built a simulated robotic vehicle with an environment where traffic signs are placed alongside a road. We also integrated a module for object detection and recognition to determine the longitudinal behavior of the vehicle. In addition, the current study implemented a lateral controller based on a convolutional neural network for the vehicle to make it drive by itself. We believe that the proposed integrated framework can be utilized by researchers and educators and lower entry barriers in the prosperous autonomous vehicle research.