대형 상용 차량에 적용하기 위한 운전자 지원 시스템 개발

Abstract

A semi-trailer truck is a large commercial vehicle consisting of a semi-trailer which carries a load and a tractor towing it. Because the semi-trailer truck has a larger cargo to carry a load than general commercial vehicles and is able to transport large/heavy loads at a time, it is widely used in the logistics transportation field. However, in case of traffic accident, a scale of damage is very large because the semi-trailer truck has larger size and mass than the general passenger vehicles. The major causes of these kinds of accidents consist of lane departure and collision with surrounding vehicles. Recently, the active safety system for collision avoidance and many driver assistant systems for preventing lane departure have been studied actively, and some systems have been adapted in passenger vehicles. Especially, the collision avoidance system through emergency braking is expected to become more widespread in the future because related regulations are being set in many countries. However, since most of these studies are aimed at passenger vehicles, they are difficult to apply to the semi-trailer truck that exhibit different motion from the passenger vehicles. In addition, there may occur more dangerous situation such as unstable motion of the semi-trailer or falling of the stacked cargoes when the semi-trailer truck conducts emergency braking. Therefore, this thesis proposes the lane keeping assist system and forward collision avoidance system for the semi-trailer truck. In order to perform the control considering several characteristics of the semi-trailer truck distinct from the general passenger vehicle, the proposed system is designed by combining the in-vehicle information estimation algorithm, the vehicle surrounding environment recognition algorithm and the steering control algorithm considering the motion of the semi-trailer. First, the proposed in-vehicle information estimation algorithm consists of the vehicle parameter estimation algorithm and the lateral motion estimation algorithm of the semi-trailer truck. The vehicle parameter estimation algorithm is proposed for design of the system considering characteristics that the mass of the semi-trailer truck may change significantly at every run. The proposed algorithm estimates the vehicle mass just by using the information about vehicle motion and powertrain which can be easily obtained by the in-vehicle network of the real commercial vehicles without requiring the additional sensor. In addition, the inclination angle of the road on which the vehicle is traveling is also estimated to improve the mass estimation performance. The lateral motion estimation algorithm of semi-trailer truck was proposed to estimate the articulation angle which is not measured by actual semi-trailer truck in real time. The proposed estimation algorithm estimates the lateral motion of semi-trailer truck such as articulation angle, lateral velocity, yaw-rate, etc. using lateral dynamics model of the articulated vehicle and the vehicle motion information provided by the in-vehicle network in real time. In addition, in order to ensure robust estimation performance for model uncertainty, disturbances that influence the vehicle state estimator are estimated together and compensated. Secondly, the proposed surrounding environment perception algorithm is composed of a backward lane prediction algorithm and a surrounding vehicle trajectory prediction algorithm. The backward lane prediction algorithm was devised to prevent lane departure, as well as tractor, semi-trailer. The proposed estimation algorithm does not require additional attachments of sensors but utilizes the forward lane information acquired from the camera sensor and the provided vehicle motion information from in-vehicle network to determine lane information of semi-trailer parts in real time. The surrounding vehicle trajectory prediction algorithm was proposed to predict whether a collision is occurring or not and perform the collision avoidance maneuver. The proposed algorithm predicts the trajectory of the surrounding vehicles by using the relative motion information of the surrounding vehicles acquired from the environmental sensor and the lane information acquired from the camera sensor. The proposed algorithm integrates prediction results using two models using physically based models and pilot based models. Finally, the steering control algorithm considering the motion of articulated vehicle is proposed to design lane keeping assist system and collision avoidance system for semi-trailer truck. The proposed algorithm is designed based on model predictive control and utilize the estimation results of the in-vehicle information and surrounding environment perception. Additionally, the proposed control algorithm secure robustness of model predictive control based control algorithm through compensating estimated disturbance and handling infeasible constraint set. In this thesis, the performance of in-vehicle information estimation algorithm and surrounding environment perception algorithm is verified by conducting experiments using the test vehicle, and the performance of the steering control algorithm was verified through simulation using commercial software. As a result of the verification, the proposed estimation algorithm accurately estimated the vehicle parameters and the vehicle motion state, and the surrounding environment perception algorithm appropriately predicted the backward lane information and the motion of the surrounding vehicles. Finally, it is verified that the proposed system is effective for lane keeping and collision avoidance of semi-trailer truck.|Semi-trailer truck은 짐을 적재하는 Semi-trailer와 이를 견인하는 Tractor로 구성된 대형 상용 차량이다. Semi-trailer truck은 일반적인 상용 차량과 비교하여 화물을 적재할 수 있는 공간이 많고, 크기가 크거나 무거운 화물 또한 한번에 운송할 수 있기 때문에, 물류 운송 분야에서 널리 사용되고 있다. 하지만, Semi-trailer Truck은 일반 승용 차량에 비해 그 크기와 무게가 크기 때문에, 교통 사고가 발생할 경우 그 피해 규모 또한 매우 크다. 이러한 사고의 주된 원인은 차선 이탈과 주변 차량과의 충돌이 차지한다. 최근, 차량의 차선 이탈을 방지하기 위한 다양한 운전자 지원 시스템과 충돌을 방지하기 위한 능동 안전 시스템이 활발히 연구되었으며, 일부 시스템은 실제 차량에 양산 적용 되기도 하였다. 특히, 긴급 제동을 통한 충돌 회피 시스템은 전 세계적으로 관련 규제가 마련되고 있기 때문에 향후 더욱 널리 보급될 것으로 보인다. 하지만, 이러한 연구들은 대부분 일반 승용 차량을 대상으로 하고 있기 때문에, 승용 차량과 상이한 움직임을 보이는 Semi-trailer Truck 에 쉽게 적용되기 어렵다. 또한, Semi-trailer Truck이 긴급 제동을 실시할 경우, Semi-trailer의 운동이 불안정해지거나 적재된 화물이 쏟아지는 등 더욱 위험한 상황이 발생할 가능성이 있다. 따라서 본 논문에서는, Semi-trailer Truck에 적용하기 위한 차선 유지 지원 시스템과 전방 충돌 회피 시스템을 제안한다. 제안하는 시스템은 일반 승용 차량과는 차별되는 Semi-trailer Truck의 여러 특징들을 고려한 제어를 수행하기 위하여, 차량 내부 정보 추정 알고리즘과 차량 주변 환경 인지 알고리즘, Semi-trailer의 운동을 고려한 조향 제어 알고리즘을 통합하여 설계되었다. 먼저, 제안하는 차량 내부 정보 추정 알고리즘은 차량 파라미터 추정 알고리즘과 Semi-trailer Truck의 횡 방향 운동 상태 추정 알고리즘으로 구성되었다. 차량 파라미터 추정 알고리즘은, Semi-trailer Truck의 질량이 매 주행 상황 마다 크게 변할 수 있다는 특징을 고려한 시스템을 설계하기 위하여 제안되었다. 제안하는 추가적인 센서 장착을 요구하지 않고, 실제 상용 차량에서In-vehicle Network를 통해 쉽게 획득할 수 있는 차량 운동 정보와 Powertrain에 관한 정보만을 사용하여 차량의 질량을 실시간으로 추정한다. 이때, 질량 추정 성능을 향상시키기 위하여, 차량이 주행 중인 도로의 경사각 또한 함께 추정 하였다. Semi-trailer Truck의 횡 방향 운동 상태 추정 알고리즘은, 실제 Semi-trailer Truck에서 계측되지 않는 굴절각(Articulation angle)을 실시간으로 추정하기 위하여 제안되었다. 제안하는 추정 알고리즘은 굴절 차량의 횡 방향 운동 모델과 In-vehicle network를 통해 제공되는 차량 운동 정보를 사용하여 Articulation angle, Lateral velocity, Yaw-rate등의 횡 방향 운동 정보를 실시간으로 추정한다. 이때, Model uncertainty에 강인한 추정 성능을 확보하기 위하여 차량에 영향을 미치는 Disturbance를 함께 추정하고 이를 보상하도록 하였다. 두 번째로, 제안하는 차량 주변 환경 인지 알고리즘은 후방 차선 정보 예측 알고리즘과 주변 차량 궤적 예측 알고리즘으로 구성되었다. 후방 차선 정보 예측 알고리즘은, Tractor 뿐만 아니라 Semi-trailer 또한 차선 이탈을 방지하기 위해 고안되었다. 제안하는 추정 알고리즘은 추가적인 센서 장착을 요구하지 않고, 카메라 센서로부터 획득되는 전방 차선 정보와 In-vehicle network를 통해 제공되는 차량 운동 정보만을 활용하여 Semi-trailer부분의 차선 정보를 실시간으로 알아낸다. 주변 차량 궤적 예측 알고리즘은, 주변 차량과의 충돌 여부를 예측하고 이를 회피하는 제어를 수행하기 위하여 제안되었다. 제안하는 예측 알고리즘은 환경 센서로부터 획득한 주변 차량의 상대 운동 정보와 카메라 센서로부터 획득한 차선 정보를 이용하여 주변 차량의 미래의 움직임을 예측한다. 제안하는 알고리즘은 주변 차량 경로 예측 과정에 Physics based model과 Maneuver based model을 사용되었으며, 두 모델을 사용한 예측 결과를 통합하는 방법이 제안되었다. 마지막으로, Articulated vehicle의 운동을 고려한 조향 제어 알고리즘은 Semi-trailer Truck을 위한 차선 유지 지원 시스템과 충돌 회피 시스템의 설계를 위해 제안되었다. 제안하는 제어 알고리즘은 Model Predictive Control 기법을 기반으로 설계되었으며, 앞서 언급한 차량 내부 정보 추정 결과와 주변 환경 인지 결과를 모두 활용한다. 제안하는 제어 알고리즘은, 추가적으로, 추정된 Disturbance를 보상하는 방법과 Infeasible Constraint Set Handling을 통해 Model Predictive Control기반 제어 고리즘의 강인성을 확보하였다. 본 연구에서는 차량을 이용한 실험을 수행하여 차량 내부 정보 추정 알고리즘과 주변 환경 인지 알고리즘의 성능을 검증하였으며, 상용 소프트웨어를 이용한 시뮬레이션을 통해 조향 제어 알고리즘의 성능을 확인하였다. 검증 결과, 제안하는 추정 알고리즘은 차량 파라미터와 차량 운동 상태를 정확히 추정하였으며, 주변 환경 인지 알고리즘은 후방 차선 정보 및 주변 차량의 움직임을 적절히 예측하였다. 최종적으로, 제안하는 시스템이 Semi-trailer Truck의 차선 유지 및 충돌 회피에 효과적임을 확인하였다.; A semi-trailer truck is a large commercial vehicle consisting of a semi-trailer which carries a load and a tractor towing it. Because the semi-trailer truck has a larger cargo to carry a load than general commercial vehicles and is able to transport large/heavy loads at a time, it is widely used in the logistics transportation field. However, in case of traffic accident, a scale of damage is very large because the semi-trailer truck has larger size and mass than the general passenger vehicles. The major causes of these kinds of accidents consist of lane departure and collision with surrounding vehicles. Recently, the active safety system for collision avoidance and many driver assistant systems for preventing lane departure have been studied actively, and some systems have been adapted in passenger vehicles. Especially, the collision avoidance system through emergency braking is expected to become more widespread in the future because related regulations are being set in many countries. However, since most of these studies are aimed at passenger vehicles, they are difficult to apply to the semi-trailer truck that exhibit different motion from the passenger vehicles. In addition, there may occur more dangerous situation such as unstable motion of the semi-trailer or falling of the stacked cargoes when the semi-trailer truck conducts emergency braking. Therefore, this thesis proposes the lane keeping assist system and forward collision avoidance system for the semi-trailer truck. In order to perform the control considering several characteristics of the semi-trailer truck distinct from the general passenger vehicle, the proposed system is designed by combining the in-vehicle information estimation algorithm, the vehicle surrounding environment recognition algorithm and the steering control algorithm considering the motion of the semi-trailer. First, the proposed in-vehicle information estimation algorithm consists of the vehicle parameter estimation algorithm and the lateral motion estimation algorithm of the semi-trailer truck. The vehicle parameter estimation algorithm is proposed for design of the system considering characteristics that the mass of the semi-trailer truck may change significantly at every run. The proposed algorithm estimates the vehicle mass just by using the information about vehicle motion and powertrain which can be easily obtained by the in-vehicle network of the real commercial vehicles without requiring the additional sensor. In addition, the inclination angle of the road on which the vehicle is traveling is also estimated to improve the mass estimation performance. The lateral motion estimation algorithm of semi-trailer truck was proposed to estimate the articulation angle which is not measured by actual semi-trailer truck in real time. The proposed estimation algorithm estimates the lateral motion of semi-trailer truck such as articulation angle, lateral velocity, yaw-rate, etc. using lateral dynamics model of the articulated vehicle and the vehicle motion information provided by the in-vehicle network in real time. In addition, in order to ensure robust estimation performance for model uncertainty, disturbances that influence the vehicle state estimator are estimated together and compensated. Secondly, the proposed surrounding environment perception algorithm is composed of a backward lane prediction algorithm and a surrounding vehicle trajectory prediction algorithm. The backward lane prediction algorithm was devised to prevent lane departure, as well as tractor, semi-trailer. The proposed estimation algorithm does not require additional attachments of sensors but utilizes the forward lane information acquired from the camera sensor and the provided vehicle motion information from in-vehicle network to determine lane information of semi-trailer parts in real time. The surrounding vehicle trajectory prediction algorithm was proposed to predict whether a collision is occurring or not and perform the collision avoidance maneuver. The proposed algorithm predicts the trajectory of the surrounding vehicles by using the relative motion information of the surrounding vehicles acquired from the environmental sensor and the lane information acquired from the camera sensor. The proposed algorithm integrates prediction results using two models using physically based models and pilot based models. Finally, the steering control algorithm considering the motion of articulated vehicle is proposed to design lane keeping assist system and collision avoidance system for semi-trailer truck. The proposed algorithm is designed based on model predictive control and utilize the estimation results of the in-vehicle information and surrounding environment perception. Additionally, the proposed control algorithm secure robustness of model predictive control based control algorithm through compensating estimated disturbance and handling infeasible constraint set. In this thesis, the performance of in-vehicle information estimation algorithm and surrounding environment perception algorithm is verified by conducting experiments using the test vehicle, and the performance of the steering control algorithm was verified through simulation using commercial software. As a result of the verification, the proposed estimation algorithm accurately estimated the vehicle parameters and the vehicle motion state, and the surrounding environment perception algorithm appropriately predicted the backward lane information and the motion of the surrounding vehicles. Finally, it is verified that the proposed system is effective for lane keeping and collision avoidance of semi-trailer truck.