• Journal of Auto-vehicle Safety Association
• /
• v.14 no.2
• /
• pp.14-19
• /
• 2022

This paper presents motion planning algorithm that yields to intervening side lane vehicles in a congested traffic flow based on vehicle to vehicle (V2V) communication. Autonomous driving in dense traffic situation requires advanced driving performance in terms of vehicle interaction and risk mitigation. One of the most important functions necessary for congested traffic autonomous driving is to predict the lane change intention of the side lane target vehicle. However, implementing this function by using only environmental sensors has limitations. In this study, V2V communication is used to overcome the limitations and determine the intention of cut-in vehicles. Lane change intention of the intervening side lane vehicle is inferred by its longitudinal speed, steering angle, and turn signal light information received by the on-board-unit (OBU). Once the yield decision is made, the subject vehicle decelerates to generate sufficient clearance for the target vehicle to enter. Validation of the algorithm was conducted with actual autonomous test vehicles.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.6
• /
• pp.117-122
• /
• 2021

Autonomous driving allows cars to drive without people and is being studied very actively thanks to the recent development of artificial intelligence technology. Among artificial intelligence technologies, deep reinforcement learning is used most effectively. Deep reinforcement learning requires us to build a neural network using an appropriate activation function. So far, many activation functions have been suggested, but different performances have been shown depending on the field of application. This paper compares and evaluates the performance of which activation function is effective when using deep reinforcement learning to learn autonomous driving on highways. To this end, the performance metrics to be used in the evaluation were defined and the values of the metrics according to each activation function were compared in graphs. As a result, when Mish was used, the reward was higher on average than other activation functions, and the difference from the activation function with the lowest reward was 9.8%.

• International Journal of Internet, Broadcasting and Communication
• /
• v.16 no.1
• /
• pp.63-71
• /
• 2024

Deep reinforcement learning (RL) significantly influences autonomous vehicle development by optimizing decision-making and adaptation to complex driving environments through simulation-based training. In deep RL, an activation function is used, and various activation functions have been proposed, but their performance varies greatly depending on the application environment. Therefore, finding the optimal activation function according to the environment is important for effective learning. In this paper, we analyzed nine commonly used activation functions for RL to compare and evaluate which activation function is most effective when using deep RL for autonomous vehicles to learn highway merging. To do this, we built a performance evaluation environment and compared the average reward of each activation function. The results showed that the highest reward was achieved using Mish, and the lowest using SELU. The difference in reward between the two activation functions was 10.3%.

• Journal of Auto-vehicle Safety Association
• /
• v.15 no.4
• /
• pp.23-31
• /
• 2023

In recent years, the technology for autonomous driving has been advancing rapidly, ADAS (Advanced Driver Assistance System) functions, which improve driver convenience and safety performance, are mostly equipped in recently released vehicles and range from level 0 to level 2 in autonomous driving technology. Among the various functions of ADAS, AEBS (Autonomous Emergency Braking System), which analyzes traffic accidents, is the most closely related to the vehicle's braking. This study developed a simulation technique for reproducing accidents related to AEBS based on real vehicle experimental data, and it was applied to the analysis of actual ADAS vehicle accidents to identify the causes of accidents.

• Journal of Auto-vehicle Safety Association
• /
• v.13 no.3
• /
• pp.102-110
• /
• 2021

As an alternative to solving safety, environments, and aging problems, ADS (Automated driving system) in the global automotive market is actively being developed as a new growth industry. In time for the appearance of ADS, relevant regulations and assessment programs must also be developed. For example, safety standards for the Level 3 automated driving system were promulgated in December 2019 by the Ministry of Land, Infrastructure and Transport of Korean government. However, assessment programs such as KNCAP for autonomous functions of ADS have not yet been introduced in Korea as well as globally. The autonomous driving functions of ADS at Level 3 or higher must be capable to recognize, judge and respond to objects and events in a wide variety of complex situations. In this paper, we examined and studied the complex situations, considerations and assessment items that ADS must respond to in the interest of safety for passengers, pedestrians and other road users. We hope this paper will be helpful to develop an execution program in the future.

• Journal of Cadastre & Land InformatiX
• /
• v.53 no.2
• /
• pp.53-66
• /
• 2023

Recently, there has been a significant interest in the development of autonomous driving simulation environment based on digital twin. In the development of such digital twin-based simulation environment, many researches has been conducted not only performance and functionality validation of autonomous driving, but also generation of virtual training data for deep learning. However, such digital twin-based autonomous driving simulation system has the problem of requiring a significant amount of time and cost for the system development and the data construction. Therefore, in this research, we aim to propose a method for rapidly designing and implementing a digital twin-based autonomous driving simulation system, using only the existing 3D models and high-definition map. Specifically, we propose a method for integrating 3D model of FBX and NGII HD Map for the Busan EDC area into CARLA, and a method for adding and modifying CARLA functions. The results of this research show that it is possible to rapidly design and implement the simulation system at a low cost by using the existing 3D models and NGII HD map. Also, the results show that our system can support various functions such as simulation scenario configuration, user-defined driving, and real-time simulation of traffic light states. We expect that usability of the system will be significantly improved when it is applied to broader geographical area in the future.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.42 no.1
• /
• pp.205-213
• /
• 2017

This paper is a study on the method to derive the functional level required for autonomous unmanned ground vehicle, one of the defense robots. Conventional weapon systems are not significantly affected by the operating environment, while defense robots exhibit different performance depending on the operating environment, even if they are on the same platform. If the performance of defense robot is different depending on operational environment, results of mission performance will be vary significantly. Therefore, it is necessary to clarify the level of function required by the military in order to research and develop most optimal defense robots. In this thesis, we propose a method to derive the required function level of unmanned ground vehicles, focusing on autonomous driving, one of the most vital functions of defense robots. Our results showed that the autonomous driving function depending intervention levels and evaluated functional sensitivity for autonomous driving of the unmanned vehicle using climate and topography as variables.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2017.05a
• /
• pp.246-248
• /
• 2017

Self-driving cars have three main functions. The first recognizes the surrounding environment, judge the risk, and lastly plans the drive path. Therefore, the driving operation is minimized. And it refers to a human friendly car capable of safe driving on its own. The reason for the need for self-driving car was to reduce traffic jams on limited roads and to reduce carbon dioxide emissions. Driving ahead of these self-driving car businesses can be expected to attract and expand the existing business and expand the new business and create new business opportunities for ICT firms. It is urgent for the concerned agencies to establish legal and institutional basis for self-driving cars. By doing so, new services could be provided to consumers. Therefore, this paper introduces the technological development trends for self-driving cars.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.7
• /
• pp.674-684
• /
• 2011

More and more elderly and disabled people are using electric scooters instead of electric wheelchairs because of higher mobility. However, people with high levels of impairment or the elderly still have difficulties in driving the electric scooters safely. Semi-autonomous electric scooter system is one of the solutions for the safety: Either manual driving or autonomous driving can be used selectively. In this paper, we implement a semi-autonomous electric scooter system with functions of localization and path following. In order to recognize the pose of electric scooter in outdoor environments, we design an outdoor localization system based on the extended Kalman filter using DGPS (Differential Global Positioning System) and wheel encoders. We added an accelerometer to make the localization system adaptable to road condition. Also we propose a path following algorithm using two arcs with current pose of the electric scooter and a given path in the map. Simulation results are described to show that the proposed algorithms provide the ability to drive an electric scooter semi-autonomously. Finally, we conduct outdoor experiments to reveal the practicality of the proposed system.

• Journal of IKEEE
• /
• v.24 no.1
• /
• pp.243-247
• /
• 2020

As the autonomous driving market is rapidly growing, research on autonomous driving is being conducted. Self-driving functions should be performed regardless of the weather for the driver's safety. However, misty weather is difficult to autonomous driving because of the lack of visibility, so a defog algorithm should be used. The image obtained through the fog removal algorithm causes the image quality to deteriorate. To improve this problem, HSV color correction is used to increase the sharpness. In this paper, we propose a color correction hardware using HSV that can cope with 4K images. The hardware was designed with Verilog and verified by Modelsim. In addition, the FPGA was implemented with the goal of Xilinx's xc7z045-2ffg900.