REDUCING COMPUTATIONAL DEMAND OF MULTI-STATE CONSTRAINT KALMAN FILTER IN VISUAL-INERTIAL ODOMETRY APPLICATIONS

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2019

Tezin Dili: İngilizce

Öğrenci: Kerem Eyice

Danışman: Tolga Çiloğlu

Özet:

The aim of this study is to reduce the computational load required by Multi-State Constraint Kalman Filter in visual-inertial odometry applications while maintaining the accuracy of the localization solution. In order to accomplish this, a keyframe-based pose selection mechanism is proposed. The proposed method fuses visual measurements with inertial measurements in order to estimate the kinematics of the platform. The contribution of this study is to reduce computational demand of the filtering operations by using visual measurements obtained only from selected keyframes for navigation purposes. Experiments have been performed with visual-inertial datasets collected at different speeds of an aerial platform in order to assess the performance of the proposed method. Results showed that the proposed method can attain accurate navigation solution while decreasing the required computational load with a proper keyframe selection criteria.