메뉴 건너뛰기
소속 기관 / 학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
고객센터 ENG
주제분류

논문 기본 정보

저자정보
(국민대학교) (국민대학교) (국민대학교) (국민대학교) (국민대학교) (국민대학교)
저널정보
한국자동차공학회 한국자동차공학회논문집 한국자동차공학회논문집 제33권 제11호
오류 신고하기

피인용 0

검색

    초록·키워드

    Reliable localization in autonomous driving requires real-time, context-aware reweighting of heterogeneous sensors. This paper proposes a hierarchical fusion architecture that treats the outputs and dynamic covariances of two estimators—MSCKF-based VIO (OpenVINS) and EKF-based GNSS/INS (SBG Ellipse-D) —as observation models and fuses them with an Interacting Multiple Model (IMM) filter. In the dynamics layer, the IMM runs Constant Acceleration and Constant Turn models in parallel. In the observation layer, time-varying measurement covariance is injected into the likelihood to adapt mode probabilities and sensor weights without fixed sensor-priority rules. We employ statistical gating (Mahalanobis), covariance capping/scaling, and a gradual weight handover during GNSS outages to increase robustness to outliers and dropouts. In ROS-based road tests, the method consistently reduced Absolute and Relative Trajectory Errors against VIO-only and GNSS/INS-only baselines. Beyond accuracy, we quantify transition behavior and internal reasoning and yield improvements in Transition Stability Error, Re-convergence Time, and Model Probability Accuracy. Leveraging the dynamic reliability of the constituent estimators without prior shadow maps or fixed priorities, the approach jointly addresses maneuver and sensor uncertainties within a single Bayesian framework.

    최근 본 자료 전체보기

      UCI(KEPA) : I410-151-26-02-094580082