For the era of autonomous cars, the accurate and reliable positioning of a vehicle is critical. However, a city is still not easy for them. For example, tall buildings disrupt the GNSS signal. Therefore, it is necessary for a robot to estimate the position using only the surrounding information obtained from equipped sensors. However, the appearance of a place is diverse. Day and night are different. Dynamic objects appear and disappear. A building that existed yesterday could be demolished today. In this thesis, we explore the intrinsic feature of a place that distinguishes that place from others. How does a human recognize a place? In the field of urban design, there has been a concept called isovist. The isovist is an observer’s egocentric visibility and means the openness of a space. The openness that an observer feels in the space also determines the use of that space. For example, in a square, we get the feeling that we are open and that it is closed between high-rise buildings. The openness of the space refers to how robust it is in the presence of dynamic objects and light condition changes. This thesis proposes a robust robot localization method using a LiDAR. Because light goes straight, the shape of the surrounding environment obtained from a LiDAR is the robot’s egocentric visible space’s shape. Using this point cloud, the data-driven three-dimensional (3D) isovist is proposed and employed for robot localization. That is, in this thesis, robot localization meets 3D isovist. Extensive experiments are conducted to cover diverse environments and times, and the results—for example, those related to placeness —might come from the openness.
이 학위논문은 라이다 센서를 이용하여 로봇이 도심 속에서 강인하게 자기 위치를 추정하는 방법에 대해 제안한다. 먼저, 도시 디자인 분야에서 발전되어온 아이소비스트라는 공간의 개방도를 의미하는 개념 을 라이다 센서로부터 얻어지는 정밀한 포인트 클라우드를 이용하여 재정의한다. 공간의 개방도는 사람이 공간을 인식함에 있어 광조건 및 환경의 부분적 구조 변화 등에 강건한 것으로 알려져 있다. 제안하는 데이터 기반의 아이소비스트를 이용하면 다양한 환경변화들이 발생할 때에도 로봇이 강인하게 위치인식을 할 수 있음을 다양한 실험들을 통해 보인다. 결과적으로 이 학위논문은 아이소비스트가 환경 변화에 무관하게 어떤 공간이 가지는 고유한 특징, 즉 장소성을 대표할 수 있음을 시사한다.