서지주요정보
Interaction-aware path planning for autonomous vehicle under mandatory lane changing scenario = 의무 차로 변경 상황에서의 상호작용을 고려한 자율주행 경로 계획
서명 / 저자 Interaction-aware path planning for autonomous vehicle under mandatory lane changing scenario = 의무 차로 변경 상황에서의 상호작용을 고려한 자율주행 경로 계획 / Jaehee Choi.
발행사항 [대전 : 한국과학기술원, 2024].
Online Access 원문보기 원문인쇄

소장정보

등록번호

8043388

소장위치/청구기호

학술문화관(도서관)2층 학위논문

MGT 24007

휴대폰 전송

도서상태

이용가능(대출불가)

사유안내

반납예정일

리뷰정보

초록정보

This study aims to define interaction criteria between vehicles and incorporate them into an autonomous vehicle lane-changing path planning algorithm to minimize the differences in driving behavior between human-driven vehicles and autonomous vehicles. Traditional autonomous vehicle path planning research has mainly focused on physical safety distances and collision avoidance, but this approach does not fully reflect the complex interactions between vehicles in real road environments. Human-driven vehicles continuously interact with surrounding vehicles while changing lanes, necessitating the exploration of autonomous driving path planning that accounts for these interactions. This study focuses on mandatory lane-changing scenarios that require active interaction with surrounding vehicles. For modeling the criteria of interaction, an experimental environment was designed where two driving simulators influence each other while changing lanes. Using data obtained from the experiments, the merging probability is modeled with logistic regression and applied to the cost function of an optimal polynomial path generation algorithm. By involving human drivers in the simulators the proposed model and the baseline one were compared. The proposed model demonstrated, compared to the baseline model, a higher lane change success rate, less speed reduction in human-driven vehicles, and shorter lane change times for autonomous vehicles. This suggests that the proposed model can effectively reduce the heterogeneous driving behaviors between human-driven and autonomous vehicles. The results validate the effectiveness of integrating vehicle interactions into lane-changing path planning.

본 연구에서는 차량 간 상호작용의 기준을 정의하고 자율주행 차로 변경 경로 계획 알고리즘에 반영하여 인간 운전 차량과 자율주행의 이질적 주행 행태를 최소화하는 것을 목표로 한다. 기존의 자율주행 차량 경로 계획 연구는 주로 물리적 안전거리와 충돌 회피의 요소에 집중해 왔으나, 이는 실제 도로 환경에서 차량 간 복잡하게 이루어지는 상호작용을 충분히 반영하지 못한다. 실제 도로상에서 인간 운전 차량들은 주변 차량과 지속적으로 상호작용하며 차로를 변경하기 때문에 이를 반영한 자율주행의 주행 경로 계획에 대한 탐구가 필요하다. 본 연구에서는 주변 차량과의 적극적인 상호작용을 요구하는 의무 차로 변경 상황에 집중하였 다. 상호작용의 기준을 모델링하기 위해 두 대의 운전 시뮬레이터로 서로에게 영향을 주며 차로 변경하는 상황이 되도록 실험 환경을 조성하였다. 실험을 통해 얻은 데이터로부터 차로 변경 확률을 로지스틱 회귀로 모델링하고, 이를 최적 다항식 경로 생성 알고리즘의 비용 함수에 적용하였다. 사람을 운전 시뮬레이터에 개입하여, 제안된 모델과 기존 방식과의 결과를 비교하였다. 제안된 모델은 기존 방식 대비 높은 차로 변경 성공 확률을 보였고, 인간 운전 차량의 속도 저하를 덜 유발하였으며, 자율주행 차량의 차로 변경 시간이 더 짧은 것으로 나타났다. 이는 제안된 모델이 인간 운전 차량과 자율주행의 이질적 주행 행태를 효과적으로 줄일 수 있음을 시사한다. 결과를 통해, 본 연구는 차로 변경 경로 계획에 차량 간 상호작용을 통합하는 접근 방식의 유효성을 입증하였다.

서지기타정보

서지기타정보
청구기호 {MGT 24007
형태사항 iv, 51 p. : 삽도 ; 30 cm
언어 영어
일반주기 저자명의 한글표기 : 최재희
지도교수의 영문표기 : Kitae Jang
지도교수의 한글표기 : 장기태
Including appendix
학위논문 학위논문(석사) - 한국과학기술원 : 조천식모빌리티대학원,
서지주기 References : p. 46-50
주제 자율주행
경로 계획
상호작용
차로 변경
운전 시뮬레이터
로지스틱 회귀
Autonomous driving
Path planning
Interaction
Lane-changing
Driving simulator
Logistic regression
QR CODE

책소개

전체보기

목차

전체보기

이 주제의 인기대출도서

Autonomous vehicle sales, fleet, travel and benefit projections from 1

Mandatory lane-changing (MLC) traffic scenario. Merging vehicle attempts to merge between the leading vehicle (LV) and the following vehicle (FV) in the target lane.

Research framework

Three hypothetical vehicle trajectories to illustrate the identification of anticipation, relax- ation in [36]

US-101 cite of NGSIM

3D surface plot showing the merging probability as a function of X1 and X2, based on results from [39]. Red dots indicate Table 2.1

Comparative analysis of an interaction case (a) and a non-interaction case (b) in terms of lateral position, speed, and merging probability over time. Each plot illustrates the dynamics between a merging vehicle (blue line) and a following vehicle (yellow line) with the merging probability depicted in the bottom graph

Statistics of situational variables in interaction case and non-interaction case

Histograms comparing interaction and non-interaction cases duringmandatory lane changes considering various situational variables. The histograms represent the distribution of:(a) Duration of lane change, (b) Duration ofinteraction, (c) Maximum lateral velocity, (d) Maximum lateral acceleration (e) Maximum follow gap, and (f) Gap between the FV and the LV when the MV passes the FV. Blue bars indic

Cumulative Distribution Functions (CDFs) comparing interaction and non-interaction cases during mandatory lane changes considering various situational variables. The CDFs represent the distribution of: (a) Duration oflane change, (b) Duration ofinteraction, (c) Maximum lateral velocity (d) Maximum lateral acceleration, (e) Maximum follow gap, and (f) Gap between the FV and the LV when the MV passe

Multi-driver simulator experiment schematic diagram

Frequency of maximum speed difference (a), maximum speed of merging vehicle (b), maximum follow gap (c) during interaction duration

Interfaces of the simulators: FV driver (left) and MV driver (right). The side mirror is visible for the MV driver.

Multi-driver simulator setup with left-turn indicator

3D surface plot showing the merging probability as a function of X1 and X2, based on results from the multi-driver simulator experiment

Diagnostics of logistic regression models

Odds ratios of each model

Collision avoidance diagram. The dotted line indicates the initial state and the solid line indicates the final state.

Transforming coordinates from the Frenet frame (a) to the global frame (b). All generated paths based on Frenet coordinates are shown in grey in (a). In (b), red paths are unfeasible, black paths are feasible candidate paths, and the green path is the optimal path with the minimum cost.

Summary of Cost Functions Used in Path Planning

Generation of path candidates according to vehicle state.

Parameter settings for each state

(a) Decision criteria of finite state machine (FSM), (b) Y> p region for various p values.

Planned and actual trajectories. These plots demonstrate the accuracy of the trajectory planning model in real-time execution.

Speed and yaw degree of baseline model (a) and proposed model (b) when the initial condition is MV speed=40 km/h, follow gap=20m, speed difference-20km/h

Histograms comparing the proposed model with the baseline model in a driving simulator experiment. The histograms illustrate three key performance metrics: The speed drop of the FV(a), required merging time of the AV(b), and distance traveled by the AV in the current lane during the merging process(c)