Skip to main navigation menu Skip to main content Skip to site footer

Articles

CJICT: VOL. 9, NO. 2, DECEMBER 2021

A STUDY ON AUTONOMOUS DRIVING ADAPTIVE SIMULATION SYSTEM USING DEEP LEARNING MODEL YOLOV3

Submitted
December 23, 2021
Published
2021-12-20

Abstract

For the safety of autonomous vehicles, it is not necessary that the human driver does not have much trouble detecting other vehicles and maintaining a certain distance between them, but in the case of autonomous vehicles, that's not an easy task. The problem of detecting and recognizing the front state of autonomous vehicles is known as object detection by Yolov3 bounding boxes. Therefore, we propose this study to avoid accidents before they occur due to autonomous driving on the road and for a better future.  Our purpose in this study is to put autonomous vehicles on the road in practice using Simulink Matlab, and it is a reflection on the ability of autonomous vehicles to ensure curve road safety And to quickly determine responses on curve road situations such as acceleration/deceleration, stopping, and keeping the same speed direction so that better decisions can be made quickly. Simulation represents a possible solution by enabling the creation of reliable bounding boxes, as a first step, in this study, we discuss the feasibility of a simulation framework to detect the speed of different autonomous vehicles using Yolov3 in the real world. We first developed the YOLOV3 algorithm for autonomous vehicle image recognition using the dataset from the Matlab site. The YOLO v3 model, with an optimal performance compared to the performances of deep learning algorithms, is applied. The training parameters are refined through experiments and in the second part we proposed an effective system using "Vision Vehicle Detector test brake adapter" adaptive HighwayLaneFollowingTestBench/Simulation 3D Scenario to prepare Matlab Simulink simulation environment and sensors, Vision Vehicle Detector. The training parameters are refined through experiments. The vehicle detection rate is approximately 95.8% As per our best knowledge, as a result of the experiment, the proposed system has shown favorable results.For the safety of autonomous vehicles, it is not necessary that the human driver does not have much trouble detecting other vehicles and maintaining a certain distance between them, but in the case of autonomous vehicles, that's not an easy task. The problem of detecting and recognizing the front state of autonomous vehicles is known as object detection by Yolov3 bounding boxes. Therefore, we propose this study to avoid accidents before they occur due to autonomous driving on the road and for a better future.  Our purpose in this study is to put autonomous vehicles on the road in practice using Simulink Matlab, and it is a reflection on the ability of autonomous vehicles to ensure curve road safety And to quickly determine responses on curve road situations such as acceleration/deceleration, stopping, and keeping the same speed direction so that better decisions can be made quickly. Simulation represents a possible solution by enabling the creation of reliable bounding boxes, as a first step, in this study, we discuss the feasibility of a simulation framework to detect the speed of different autonomous vehicles using Yolov3 in the real world. We first developed the YOLOV3 algorithm for autonomous vehicle image recognition using the dataset from the Matlab site. The YOLO v3 model, with an optimal performance compared to the performances of deep learning algorithms, is applied. The training parameters are refined through experiments and in the second part we proposed an effective system using "Vision Vehicle Detector test brake adapter" adaptive HighwayLaneFollowingTestBench/Simulation 3D Scenario to prepare Matlab Simulink simulation environment and sensors, Vision Vehicle Detector. The training parameters are refined through experiments. The vehicle detection rate is approximately 95.8% As per our best knowledge, as a result of the experiment, the proposed system has shown favorable results.