• International Journal of Internet, Broadcasting and Communication
• /
• v.13 no.1
• /
• pp.210-218
• /
• 2021

CNN (Convolutional Neural Network), a type of deep learning algorithm, is a type of artificial neural network used to analyze visual images. In deep learning, it is classified as a deep neural network and is most commonly used for visual image analysis. Accordingly, an AI autonomous driving model was constructed through real-time image processing, and a crosswalk image of a road was used as an obstacle. In this paper, we proposed a low-cost model that can actually implement autonomous driving based on the CNN model. The most well-known deep neural network technique for autonomous driving is investigated and an end-to-end model is applied. In particular, it was shown that training and self-driving on a simulated road is possible through a practical approach to realizing lane detection and keeping.

• Smart Media Journal
• /
• v.11 no.9
• /
• pp.9-20
• /
• 2022

Deep reinforcement learning(RL) is an end-to-end data-driven control method that is widely used in the autonomous driving domain. However, conventional RL approaches have difficulties in applying it to autonomous driving tasks due to problems such as inefficiency, instability, and uncertainty. These issues play an important role in the autonomous driving domain. Although recent studies have attempted to solve these problems, they are computationally expensive and rely on special assumptions. In this paper, we propose a new algorithm MCDT that considers inefficiency, instability, and uncertainty by introducing a method called uncertainty sequence modeling to autonomous driving domain. The sequence modeling method, which views reinforcement learning as a decision making generation problem to obtain high rewards, avoids the disadvantages of exiting studies and guarantees efficiency, stability and also considers safety by integrating uncertainty estimation techniques. The proposed method was tested in the OpenAI Gym CarRacing environment, and the experimental results show that the MCDT algorithm provides efficient, stable and safe performance compared to the existing reinforcement learning method.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.11
• /
• pp.107-114
• /
• 2022

In this paper, proposes a delivery robot that can be autonomous driving learning. The proposed robot is designed to be used in park-type apartments without ground parking facilities. Compared to the existing apartments with complex ground and underground routes, park-type apartments have a standardized movement path, allowing the robot to run stably, making it suitable for students' initial education environment. The delivery robot is configured to enable delivery of parcels through machine learning technology for route learning and autonomous driving using cameras and LiDAR sensors. In addition, the control MCU was designed by separating it into three parts to enable learning by level, and it was confirmed that it can be used as a delivery robot for learning through operation tests such as autonomous driving and obstacle recognition. In the future, we plan to develop it into an educational delivery robot for various delivery services by linking with the precision indoor location information recognition technology and the public technology platform of the apartment.

• Journal of IKEEE
• /
• v.25 no.4
• /
• pp.728-734
• /
• 2021

In recent years, there has been a strong interest in the end-to-end autonomous driving based on deep reinforcement learning. In this paper, we present a reward function of latent SAC deep reinforcement learning to improve the longitudinal driving performance of an agent vehicle. While the existing reward function significantly degrades the driving safety and efficiency, the proposed reward function is shown to maintain an appropriate headway distance while avoiding the front vehicle collision.

• Journal of Auto-vehicle Safety Association
• /
• v.13 no.4
• /
• pp.129-143
• /
• 2021

This paper proposes actuator fault detection and adaptive fault-tolerant control algorithms using performance index and human-like learning for longitudinal autonomous vehicles. Conventional longitudinal controller for autonomous driving consists of supervisory, upper level and lower level controllers. In this paper, feedback control law and PID control algorithm have been used for upper level and lower level controllers, respectively. For actuator fault-tolerant control, adaptive rule has been designed using the gradient descent method with estimated coefficients. In order to adjust the control parameter used for determination of adaptation gain, human-like learning algorithm has been designed based on perceptron learning method using control errors and control parameter. It is designed that the learning algorithm determines current control parameter by saving it in memory and updating based on the cost function-based gradient descent method. Based on the updated control parameter, the longitudinal acceleration has been computed adaptively using feedback law for actuator fault-tolerant control. The finite window-based performance index has been designed for detection and evaluation of actuator performance degradation using control error.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.20 no.6
• /
• pp.313-330
• /
• 2021

In an autonomous driving system equipped with a camera and lidar, depth completion techniques enable dense depth estimation. In particular, using self-supervised learning it is possible to train the depth completion network even without ground truth. In actual autonomous driving, such depth completion should have very short latency as it is the input of other algorithms. So, rather than complicate the network structure to increase the accuracy like previous studies, this paper focuses on network latency. We design a U-Net type network with RegNet encoders optimized for GPU computation. Instead, this paper presents several techniques that can increase accuracy during the process of self-supervised learning. The proposed techniques increase the robustness to unreliable lidar inputs. Also, they improve the depth quality for edge and sky regions based on the semantic information extracted in advance. Our experiments confirm that our model is very lightweight (2.42 ms at 1280x480) but resistant to noise and has qualities close to the latest studies.

• ETRI Journal
• /
• v.45 no.5
• /
• pp.847-861
• /
• 2023

Dynamic object detection is essential for ensuring safe and reliable autonomous driving. Recently, light detection and ranging (LiDAR)-based object detection has been introduced and shown excellent performance on various benchmarks. Although LiDAR sensors have excellent accuracy in estimating distance, they lack texture or color information and have a lower resolution than conventional cameras. In addition, performance degradation occurs when a LiDAR-based object detection model is applied to different driving environments or when sensors from different LiDAR manufacturers are utilized owing to the domain gap phenomenon. To address these issues, a sensor-fusion-based object detection and classification method is proposed. The proposed method operates in real time, making it suitable for integration into autonomous vehicles. It performs well on our custom dataset and on publicly available datasets, demonstrating its effectiveness in real-world road environments. In addition, we will make available a novel three-dimensional moving object detection dataset called ETRI 3D MOD.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.5
• /
• pp.1223-1237
• /
• 2024

Autonomous vehicles use onboard sensors to sense the surrounding environment. In complex autonomous driving scenarios, the detection and recognition capabilities are constrained, which may result in serious accidents. An efficient way to enhance the detection and recognition capabilities is establishing collaborations with the neighbor vehicles. However, the collaborations introduce additional challenges in terms of the data heterogeneity, communication cost, and data privacy. In this paper, a novel personalized federated learning framework is proposed for addressing the challenges and enabling efficient collaborations in autonomous driving environment. For obtaining a global model, vehicles perform local training and transmit logits to a central unit instead of the entire model, and thus the communication cost is minimized, and the data privacy is protected. Then, the inference similarity is derived for capturing the characteristics of data heterogeneity. The vehicles are divided into clusters based on the inference similarity and a weighted aggregation is performed within a cluster. Finally, the vehicles download the corresponding aggregated global model and train a personalized model which is personalized for the cluster that has similar data distribution, so that accuracy is not affected by heterogeneous data. Experimental results demonstrate significant advantages of our proposed method in improving the efficiency of collaborative perception and reducing communication cost.

• The Journal of Korea Robotics Society
• /
• v.12 no.3
• /
• pp.297-305
• /
• 2017

As the development of autonomous vehicles becomes realistic, many automobile manufacturers and components producers aim to develop 'completely autonomous driving'. ADAS (Advanced Driver Assistance Systems) which has been applied in automobile recently, supports the driver in controlling lane maintenance, speed and direction in a single lane based on limited road environment. Although technologies of obstacles avoidance on the obstacle environment have been developed, they concentrates on simple obstacle avoidances, not considering the control of the actual vehicle in the real situation which makes drivers feel unsafe from the sudden change of the wheel and the speed of the vehicle. In order to develop the 'completely autonomous driving' automobile which perceives the surrounding environment by itself and operates, ability of the vehicle should be enhanced in a way human driver does. In this sense, this paper intends to establish a strategy with which autonomous vehicles behave human-friendly based on vehicle dynamics through the reinforcement learning that is based on Q-learning, a type of machine learning. The obstacle avoidance reinforcement learning proceeded in 5 simulations. The reward rule has been set in the experiment so that the car can learn by itself with recurring events, allowing the experiment to have the similar environment to the one when humans drive. Driving Simulator has been used to verify results of the reinforcement learning. The ultimate goal of this study is to enable autonomous vehicles avoid obstacles in a human-friendly way when obstacles appear in their sight, using controlling methods that have previously been learned in various conditions through the reinforcement learning.

• Journal of Drive and Control
• /
• v.19 no.4
• /
• pp.104-109
• /
• 2022

For autonomous driving research using AI, datasets collected from road environments play an important role. In other countries, various datasets such as CityScapes, A2D2, and BDD have already been released, but datasets suitable for the domestic road environment still need to be provided. This paper analyzed and verified the dataset reflecting the Korean driving environment. In order to verify the training dataset, the class imbalance was confirmed by comparing the number of pixels and instances of the dataset. A similar A2D2 dataset was trained with the same deep learning model, ConvNeXt, to compare and verify the constructed dataset. IoU was compared for the same class between two datasets with ConvNeXt and mIoU was compared. In this paper, it was confirmed that the collected dataset reflecting the driving environment of Korea is suitable for learning.