http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
차량 횡방향 모델 정확도 개선을 위한 타이어 조향각 추정
조건희(Kunhee Cho),나호용(Hoyong Na),조완기(Wanki Cho),유승한(Seung-Han You) 한국자동차공학회 2021 한국 자동차공학회논문집 Vol.29 No.7
In this paper, new estimation methods for vehicle tire steering angles were proposed to improve the accuracy of a vehicle lateral model. There are several error factors in the conventional, single-track model as a result of certain assumptions in the derivation process. The front and rear steering angles can be considered influential input signals to a vehicle lateral model, as well as very large error factors. In conventional vehicle models, the steering angles are determined only by the driver’s steering angle. However, in the real world, they are also affected by roll steer and compliance steer. Therefore, the proposed methods, in which roll and compliance are considered, were developed efficiently in the form of an open-loop model by using a lateral acceleration sensor measurement signal. The proposed steering angle estimation models are validated through CARSIM simulations and real-vehicle experiments under different driving conditions.
다양한 주행 상황에 견실한 차량 롤 각 추정 로직 개발
조건희(Kunhee Cho),나호용(Hoyong Na),유승한(Seung-Han You) 대한기계학회 2019 大韓機械學會論文集A Vol.43 No.10
본 논문에서는 별도의 고가 센서 장착 없이 양산 차량에서 가용한 센서 정보만을 이용하여 롤 전복 상황 및 종/횡 경사로 주행과 같이 기존 추정 방법으로는 추정이 어려운 악의적 주행 조건에서도 강건한 롤 각 추정 로직을 개발하였다. 견실한 차량의 롤 각 추정을 위해 정적인 주행 상황에서는 횡 가속도, 요 레이트 센서 계측치를 주로 이용하여 롤 각을 추정하고, 횡 방향 동적 거동이 큰 상황에 대해서는 롤 레이트 적분에 가중치를 더 부여하여 롤 각을 추정하였다. 제안한 추정 알고리즘은 평지 주행상황뿐만 아니라 횡 경사로 주행, 주차타워 주행, 롤 전복 상황과 같은 다양한 실차 주행 시나리오들에서 검증되었다. 그 결과, 제안된 롤 각 추정 로직이 여러 악의 주행 조건들에서 추정 성능을 견실하게 유지하는 것을 확인하였다. This paper presents a vehicle roll angle estimator that is robust to challenging driving situations such as rollover and driving on longitudinal/lateral slopes without installing separate high-cost sensors. To estimate vehicle roll angle in static situation, the lateral acceleration and yaw rate are mainly used to calculate static roll angle. Moreover, the integration of roll rate is chiefly used when the vehicle transient lateral motion is quite large. This proposed algorithm is validated via vehicle experiments in various scenarios such as slalom on a flat road, driving on a banked road and parking tower, and rollover situation. The experimental results indicate that the proposed algorithm robustly maintains the accuracy of estimation in various driving situations.
자율주행 횡방향 제어를 위한 Gaussian Processes 기반의 차량 모델 개발
조건희(Kunhee Cho),전명환(Myungwhan Jeon),이형철(Hyeongcheol Lee) 한국자동차공학회 2023 한국자동차공학회 학술대회 및 전시회 Vol.2023 No.11
Various methods have been proposed for the development of lateral controller for autonomous vehicle over decades. Model predictive control (MPC), which utilize vehicle models and optimization methods considering future information and state constraints, are widely used for designing a lateral controller of autonomous vehicle. However, the vehicle models used in MPC typically employ linear models to reduce computational complexity, which results in prediction errors and a degradation in controller performance for highly nonlinear regions of the vehicle’s behavior. In this paper, we designed a gaussian process (GP)-based vehicle model to improve the prediction performance of lateral controllers for autonomous vehicle. First, we conducted signal measurements necessary for the model construction through real-world vehicle experiments and preprocessing for GP design. Especially, the lateral velocity, which is the output of the model, was measured via RTK/GNSS and utilized for optimizing the GP for the model construction. Through offline optimization process, we derived the types of kernels and hyperparameters of the GP to improve model accuracy. The proposed vehicle model was evaluated for various driving scenarios, such as lane changes and steady circular turns.
[응용논문] Data-Driven 차량 횡방향 모델 정확도 개선
조건희(Kunhee Cho),김좌헌(Joahun Kim),이형철(Hyeongcheol Lee),유승한(Seung-Han You),조완기(Wanki Cho) 한국자동차공학회 2022 한국 자동차공학회논문집 Vol.30 No.2
In this paper, a data-driven based method was proposed to improve the accuracy of the vehicle lateral model. In the derivation of the conventional single-track model, several error factors occurred as a result of the simplification of the model. Among them, front and rear tire cornering stiffness was the factor that was most related to the accuracy of the vehicle lateral model. In general, the conventional model uses nominal cornering stiffness without considering its nonlinearity and the effect of lateral load transfer. The proposed method was developed to compensate sufficiently for the error factors in cornering stiffness with a nonlinear map, which was designed by the parameter optimization method through the measurement data from real vehicle tests. This method was designed and validated with real vehicle experiments under various driving scenarios.
송은아(Euna Song),조건희(Kunhee Cho),이형철(Hyeongcheol Lee) 한국자동차공학회 2023 한국자동차공학회 학술대회 및 전시회 Vol.2023 No.11
Currently, wheel torque is calculated by receiving the estimated torque of the drive system controller via CAN. But, if there is a problem with the controllers calculation, the hazard of the vehicle cannot be detected, so a separate wheel torque estimation method is required for functional safety response. Therefore, this paper developed a wheel torque estimation algorithm to improve torque output fault diagnosis. The algorithm is based on vehicle dynamics (longitudinal model) and is constructed through a sliding mode observer (SMO), and the consistency of the developed algorithm is verified through vehicle dynamics simulation.