• 한국ITS학회 논문지
• /
• 제22권5호
• /
• pp.30-42
• /
• 2023

자율주행시대에 맞춰 다양한 연구에서 미시교통시뮬레이션(VISSIM)을 활용하여 자율주행차 도입시 교통류 안전성 분석을 수행중이다. 그러나, 일반자동차의 주행행태를 VISSIM 내 파라미터로 반영하여 혼재시의 교통류 안전성을 분석한 연구는 미비하였다. 따라서 본 연구에서는 실제 주행행태와 유사한 주행행태를 구현하기 위하여 일반자동차의 VISSIM 입력변수를 유전알고리즘을 통해 최적화 한 후, 자율주행차 도입률에 따른 교통류 안전성 분석을 수행하는 것을 목적으로 한다. US I-101 고속도로의 640m 구간을 대상으로 후행차량이 일반자동차인 경우에 대해서 상충횟수 분석을 수행한 결과, 전체 상충횟수는 자율주행차 도입률 20%까지 증가하였으며 20%를 초과한 이후부터는 지속적으로 감소하였다. 일반자동차와 자율주행차 사이의 상충횟수는 자율주행차 도입률 60%까지 증가하는 것으로 분석되었다. 그러나, 자율주행차의 주행행태를 기존 문헌결과를 바탕으로 하여 실제 주행행태를 표현하지 못했다는 한계가 존재한다. 그러므로 보다 정확한 분석을 위해 향후 연구에서는 자율주행차의 실제 주행행태를 반영하여 연구를 수행할 필요가 있다.

• 자동차안전학회지
• /
• 제10권4호
• /
• pp.40-49
• /
• 2018

Currently, a lot of researches about high risk test scenarios for autonomous vehicle and advanced driver assistance systems have been carried out to evaluate driving safety. This study proposes new type of test scenario that evaluate the driving safety for autonomous vehicle by reconstructing accident database of national automotive sampling system crashworthiness data system (NASS-CDS). NASS-CDS has a lot of detailed accident data in real fields, but there is no data of accurate velocity in accident moments. So in order to propose scenario generation method from accident database, we try to reconstruct accident moment from accident sketch diagram. At the same step, we propose an accident of occurrence frequency which is based on accident codes and road shapes. The reconstruction paths from accident database are integrated into evaluation of simulation environment. Our proposed methods and processor are applied to MILS (Model In the Loop Simulation) and VILS (Vehicle In the Loop Simulation) test environments. In this paper, a reasonable method of accident reconstruction typology for autonomous vehicle evaluation of feasibility is proposed.

• 품질경영학회지
• /
• 제46권4호
• /
• pp.1001-1014
• /
• 2018

Purpose: Autonomous driving cars, which are often represent the new convergence product, have been researched since the early years of 1900 but their safety assurance policies are yet to be implemented for real world practices. The primary purpose of this paper is to propose a modular concept based on which a safety assurance system can be designed and implemented for operating autonomous driving cars. Methods: We combine a set of key attributes of CE mark (European Assurance standard), E-Mark (Automobile safety assurance system), and A-SPICE (Automobile software assurance standard) into a modular approach. Results: Autonomous vehicles are emphasizing software safety, but there is no integrated safety certification standard for products and software. As such, there is complexity in the product and software safety certification process during the development phase. Using the concept of module, we were able to come up with an integrated safety certification system of product and software for practical uses in the future. Conclusion: Through the modular concept, both international and domestic standards policy stakeholders are expected to consider a new structure that can help the autonomous driving industries expedite their commercialization for the technology advanced market in the era of Industry 4.0.

• 로봇학회논문지
• /
• 제18권1호
• /
• pp.72-81
• /
• 2023

We introduce tightly-coupled GNSS-LiDAR-Inertial state estimator, which is capable of SLAM (Simultaneously Localization and Mapping) and autonomous driving. Long term drift is one of the main sources of estimation error, and some LiDAR SLAM framework utilize loop closure to overcome this error. However, when loop closing event happens, one's current state could change abruptly and pose some safety issues on drivers. Directly utilizing GNSS (Global Navigation Satellite System) positioning information could help alleviating this problem, but accurate information is not always available and inaccurate vertical positioning issues still exist. We thus propose our method which tightly couples raw GNSS measurements into LiDAR-Inertial SLAM framework which can handle satellite positioning information regardless of its uncertainty. Also, with NLOS (Non-light-of-sight) satellite signal handling, we can estimate our states more smoothly and accurately. With several autonomous driving tests on AGV (Autonomous Ground Vehicle), we verified that our method can be applied to real-world problem.

• 자동차안전학회지
• /
• 제16권1호
• /
• pp.35-41
• /
• 2024

In order to verify autonomous driving scenarios and safety, a lot of driving and accident data is needed, so various organizations are conducting classification and analysis of traffic accident types. In this study, it was determined that accident recording devices such as EDR (Event Data Recorder) and DSSAD (Data Storage System for Automated Driving) would become an objective standard for analyzing the causes of autonomous vehicle accidents, and traffic accidents that occurred from 2015 to 2020 were analyzed. Using the database system of IGLAD (Initiative for the Global Harmonization of Accident Data), approximately 360 accident data of EDR-equipped vehicles were classified and their characteristics were analyzed by comparing them with accident types of ADAS (Advanced Driver Assistance System)-equipped vehicles. It will be used to develop autonomous vehicle accident investigation guidelines in the future.

• 자동차안전학회지
• /
• 제13권1호
• /
• pp.38-44
• /
• 2021

This study aimed to verify the criteria of the driver monitoring systems proposed by UNECE ACSF informal working group and the ministry of land, infrastructure, and transport of South Korea using driving behavior data. In order to verify the criteria, we investigated the safety regulations of driver monitoring systems in a conditional autonomous vehicle and found that the driver monitoring measures were related to eye blinks times, head movements, and eye closed duration. Thus, we took two different experimental data including real-world driving and simulator-based drowsy driving behaviors in previous studies. The real-world driving data were used for analyzing blink times and head movement intervals, and the drowsiness data were used for eye closed duration. In the real-world driving study, 52 drivers drove approximately 11.0 km of rural road (about 20 min), 7.9 km of urban road (about 25 min), and 20.8 km of highway (about 20 min). The results suggested that the appropriate number of blinks during the last 60 seconds was 4 times, and the head movement interval was 35 seconds. The results from drowsy driving data will be presented in another paper - part 2.

• 자동차안전학회지
• /
• 제14권2호
• /
• pp.26-38
• /
• 2022

This paper proposes multiple RLS and actuator performance index-based adaptive actuator fault-tolerant control and detection algorithms for longitudinal autonomous driving. The proposed algorithm computes the desired acceleration using feedback law for longitudinal autonomous driving. When actuator fault or performance degradation exists, it is designed that the desired acceleration is adjusted with the calculated feedback gains based on multiple RLS and gradient descent method for fault-tolerant control. In order to define the performance index, the error between the desired and actual accelerations is used. The window-based weighted error standard deviation is computed with the design parameters. Fault level decision algorithm that can represent three fault levels such as normal, warning, emergency levels is proposed in this study. Performance evaluation under various driving scenarios with actuator fault was conducted based on co-simulation of Matlab/Simulink and commercial software (CarMaker).

• ETRI Journal
• /
• 제44권6호
• /
• pp.1004-1019
• /
• 2022

For autonomous driving without high-definition maps, we present a model capable of generating multiple plausible paths from egocentric images for autonomous vehicles. Our generative model comprises two neural networks: feature extraction network (FEN) and path generation network (PGN). The FEN extracts meaningful features from an egocentric image, whereas the PGN generates multiple paths from the features, given a driving intention and speed. To ensure that the paths generated are plausible and consistent with the intention, we introduce an attentive discriminator and train it with the PGN under a generative adversarial network framework. Furthermore, we devise an interaction model between the positions in the paths and the intentions hidden in the positions and design a novel PGN architecture that reflects the interaction model for improving the accuracy and diversity of the generated paths. Finally, we introduce ETRIDriving, a dataset for autonomous driving, in which the recorded sensor data are labeled with discrete high-level driving actions, and demonstrate the state-of-the-art performance of the proposed model on ETRIDriving in terms of accuracy and diversity.

• 한국인터넷방송통신학회논문지
• /
• 제22권6호
• /
• pp.43-49
• /
• 2022

FMCW레이다는 일반 차량의 충돌방지용도 뿐만 아니라 자율주행시스템에서 활발히 센서로서 사용이 되고 있다. 본 논문에서는 자율주행센서로서 개발한 2-chip 기반의 FMCW MIMO(Multi Input Multi Output) 레이다 설계 및 구현에 대해서 설명하였다. 사용 칩을 이용하여 48채널의 가상배열을 이용하여 방위각 해상도가 우수하게 설계하였으며, 특히 Frame 기반과 Chirp 기반의 파형발생 및 신호처리를 혼합하여 최대탐지 가능 속도와 속도 보상에 대해 강점을 보유할 수 있도록 제작하였으며, 구현된 시스템은 실험실 내 시험과 실제 주행시험을 통하여 성능 및 상용화 가능성에 대한 분석을 진행하였다.

• 대한임베디드공학회논문지
• /
• 제19권5호
• /
• pp.235-242
• /
• 2024

In the context of urban autonomous driving, where various types of traffic lights are encountered, traffic light recognition technology is of paramount importance. We have designed a high-performance traffic light recognition module tailored to scenarios encountered in domestic urban driving and devised a dataset construction process. In this paper, we focus on minimizing the camera's dependency to enhance traffic light recognition performance. The camera is used solely to distinguish the color information of traffic lights, while accurate location information of the traffic lights is obtained through localization and a map. Based on the information from these components, camera RoIs (Region of Interest) are extracted and transmitted to the embedded board. The transmitted images are then sent back to the main system for autonomous driving control. The processing time for one traffic light RoI averages 43.2 ms. We achieve processing times of average 93.4 ms through batch inference to meet real-time requirements. Additionally, we design a data construction process for collecting, refining, and storing traffic light datasets, including semi-annotation-based corrections.