A Study of an Autonomous Underwater Vehicle’s Navigation Algorithm using an Extended Kalman Filter

강현석 한국해양대학교 해양과학기술전문대학원 2017 국내석사

Navigation technology is vital to determine where Unmanned Underwater Vehicle (UUV) is located. This is essential to complete missions, such as submarine resource development, marine geological survey, marine ecological survey and mine clearance, and make information gathered during the mission more accurate, reliable and valuable. Dead reckoning that commonly uses Inertia Measurement Unit (IMU), Doppler Velocity Logger (DVL) and magnetic compass has position errors due to integrating acceleration and velocity. Moreover, the heading error of magnetic compass based on geodetic north includes declination and sensor noise caused by local magnetic-field effect and characteristics of sensor. This could raise the position error in the North-East-Down (NED) coordinate system in the case of dead reckoning especially using magnetic compass, because it is based on not geodetic north, but magnetic north. This makes it difficult to implement an integrated navigation system or compare the performance of navigation algorithms, such as dead reckoning, satellite navigation using Global Positioning Systems (GPS) and terrain-aided navigation using bathymetry maps. This thesis introduces a GPS-aided navigation algorithm to reduce errors accumulated while using dead reckoning navigation. This will help better estimate the position of UUVs while using dead reckoning in the NED coordinate system. For sensor fusion and measurement noise rejection, the navigation algorithm was designed to use an Extended Kalman Filter (EKF), which has much fewer calculations than an Unscented Kalman Filter (UKF) and a Particle Filter (PF). This algorithm defined the heading bias error of a magnetic compass as the difference between the UUV heading angle based on geodetic north and a magnetic compass’ heading measurement. The magnetic compass’ heading bias error was asymptotically estimated by receiving GPS positional data when it surfaced. When the navigation algorithm estimated the magnetic compass’ heading bias error, the UUV’s position was displayed in the NED coordinate system, even when the UUV was submerged. While using Matlab Simulink, an Autonomous Underwater Vehicle (AUV) dynamic simulation program was built to check the performance of the proposed navigation algorithm. The simulation program consists of a dynamic model, a sensor model, a controller and the navigation algorithm. A Naval Postgraduate School (NPS) AUV called as ARIES was used as the dynamic model because of its detailed dimensions and its precedent research containing large amounts of hydrodynamic coefficients. Furthermore, the sensor model’s characteristics were decided on according specifications and test results of sensors currently in use. Considering the sensor characteristics, the measured values of GPS, magnetic compass, DVL, gyro and pressure sensor are artificially generated on the basis of the position, attitude and velocity of AUV in the simulation. After receiving the data, the navigation algorithm estimates the compass’ heading bias error and the AUV’s position allowing control of the AUV and the ability to perform way-points and heading control simulation. The simulation incorporates three different scenarios. Two of them determine and estimate the AUV’s position and heading bias error after receiving(or not) the GPS positional data. The other uses trajectory and heading bias errors similar to those in the field test which allows comparisons of the field test results. The simulations will show that the navigation algorithm improves the accumulated positional errors of dead reckoning and the magnetic compass’ heading bias errors. In the underwater driving scenario, it was confirmed that the AUV’s position errors were improved. This was accomplished by the navigation algorithm examining the magnetic compass’ heading bias error compared to the conventional dead reckoning method. The GPS-aided navigation algorithm was applied to navigation system of a hovering-type AUV in order to verify the performance of the algorithm through field test. The applied algorithm estimates the position and attitude of the AUV and the heading bias error of Tilt-compensated Compass Module (TCM) based on geodetic north, by receiving the measurements of GPS, DVL, TCM and Attitude & Heading Reference System (AHRS). The monitoring and control system based on LabVIEW was implemented to provide the operator with the information about the AUV’s operation. Also, the AUV operating system includes the propulsion system to perform the heading control experiment or the way-point control experiment, which can be configured by the operator. Unlike the simulation, the application of GPS positional data and the estimation of TCM heading bias error depend on additional conditions for the efficient application of the navigation algorithm in the field test. In other words, the navigation algorithm utilizes GPS positional data to estimate the position and attitude of the AUV and the TCM heading bias error, so long as the positional information is judged to be efficient. Otherwise, the position and attitude of the AUV are estimated by dead reckoning considering the heading bias error of TCM obtained previously. As a result, the field test verified the performance of the navigation algorithm, by checking how precisely and accurately the TCM heading bias error was estimated and comparing the position error with the conventional dead reckoning, which was not considering the heading bias error. This thesis proposes the GPS-aided navigation algorithm for UUV. The algorithm’s performance was verified by the simulation and field test. When there is no positional information provided by acoustic beacon and bathymetry map due to long-term and long-distance voyage, the navigation algorithm can be a crucial part of a UUV’s navigation technology.

Modelling Navigation Representations During Naturalistic Driving

Zhang, Tianjiao University of California, Berkeley ProQuest Disser 2021 해외박사(DDOD)

Navigation in the natural world is a complex task that engages many cognitive systems, including vision, attention, motor control, cognitive maps, and planning. These systems recruit many brain regions that form multiple functional networks spanning the brain. Navigation has been the subject of many non-human neurophysiology studies, primarily in rodents, and less frequently in non-human primates.In the recent decades, neuroimaging using functional magnetic resonance imaging has enabled non-invasive brain activity recordings. Neuroimaging studies are able to examine cognitive processes in healthy human subjects. Together, neurophysiology and neuroimaging studies have revealed multiple regions in the human brain that are active during navigation. However, we still know very little about how these regions accomplish the complex task of navigation, and what information is represented by each region. This dissertation describes a next-generation neuroimaging experiment that maps navigational representations across the cortex to gain insight into the complex processes underlying human navigation. Chapter 1 reviews the current neurophysiology and neuroimaging literature on the navigation system of the brain. These current literatures suggest that the brain breaks navigation down into a set of subtasks, each with its associated brain regions. This chapter also describes several limitations in the current understanding of navigation in the human brain. Chapter 2 describes a next-generation, naturalistic neuroimaging paradigm to study human navigation. This paradigm leverages Unreal Engine 4, a modern game engine, to create a realistic and dynamic virtual world. A custom MR-compatible steering wheel and pedal set enables subjects to drive naturally in this virtual world. Chapter 3 describes an active navigation neuroimaging experiment that uses the paradigm developed in Chapter 2. Subjects actively perform a taxi driver task in a virtual world while BOLD activity is recorded from the brain. Voxelwise modelling with banded ridge regression is used to map the cortical representation of over 20,000 features across 33 feature spaces. Chapter 4 examines task-related visual semantic tuning shifts. Visual semantic tunings in the active navigation task are compared with those from a passive movie watching task. Results show that there are significant visual semantic tuning shifts between active navigation and passive movie watching. Chapter 5 describes a route progression model derived from the rodent literature. Results suggest that a topologic representation of route progression in RSC is conserved across species. Furthermore, RSC, unlike other visual navigation areas, strongly prefers the start of routes. Finally, Chapter 6 describes a novel, state-space based method for analyzing high-dimensional brain data. This method treats the activity of the brain as a dynamical system, and finds a low-dimensional subspace in the brain's activity space that is related to the representation of task variables. As a proof-of-concept, this method is applied to a visual semantic attention task and a video game task, and recovers low-dimensional spaces related to the representation of attentional targets and game states.

Wayfinding Design in Hospitals for the Elderly: Enhancing Navigational Support in Healthcare Space

KURBONOVA ZARINA JURAEVNA 경희대학교 대학원 2025 국내석사

Navigating hospital environments poses significant challenges for elderly individuals due to age-related cognitive, visual, and mobility impairments. This study investigates these challenges through two integrated approaches: a systematic literature review (Study A) and an experimental study conducted at a multi-level tertiary hospital in Korea (Study B). Study A synthesizes 23 international studies on wayfinding and highlights recurring issues such as signage clarity, vertical circulation complexity, and cognitive overload. Study B empirically validates these themes by analyzing the navigational behavior of 31 elderly participants using path analysis, heatmaps, think-aloud interviews, and post-task reflections. The findings reveal that vertical transitions, ambiguous signage, and corridor similarity are primary barriers to effective navigation. Comparative discussion across both studies confirms alignment on key themes and emphasizes the importance of inclusive design strategies such as improved environmental zoning, consistent visual cues, and cognitive load reduction to enhance elderly user experience in hospital settings. This research contributes evidence-based recommendations for designing accessible and intuitive healthcare environments for aging populations. Key words Wayfinding, elderly, hospital design, signage, spatial disorientation, vertical navigation, cognitive aging 병원 환경에서의 길찾기(wayfinding)는 노인들에게 인지적, 시각적, 신체적 약화로 인해 상당한 어려움을 초래할 수 있다. 본 연구는 이러한 문제를 해결하기 위해 두 가지 접근법을 통합하였다: 체계적인 문헌고찰(연구 A) 과 국내 다층 구조의 상급종합병원에서 수행된 실험적 연구(연구 B)이다. 연구 A는 2004년부터 2024년까지 발표된 23편의 국제 연구를 분석하여, 병원 내 길찾기에서 반복적으로 나타나는 주요 문제로 ‘표지판명확성 부족’, ‘수직 동선(계단 및 엘리베이터) 복잡성’, ‘인지적 과부하’를 확인하였다. 연구 B는 총 31명의 노인 참여자를 대상으로 경로 분석, 히트맵, 사고 발화법, 사후 인터뷰를 활용하여 실제 길찾기 행동을 실증적으로 검토하였다.

Map-Integrated Continuous Autonomous Navigation System for Vineyard Robots under Trellis Cultivation

Ren Fujimoto 국립부경대학교 대학원 2026 국내석사

This study proposes a fruit-centric autonomous navigation framework for vineyard cul- tivation support robots operating in overhead trellis-grown Shine Muscat vineyards. Vineyard tasks such as cluster shaping, thinning, and harvesting require precise po- sitioning directly beneath grape clusters. However, conventional navigation methods relying on QR markers, GNSS, or fixed landmarks are unsuitable for dynamic field environments, where cluster appearance and distribution vary throughout the growing season. To address this challenge, this research develops a navigation system that esti- mates grape-cluster positions using upward-facing visual detection and integrates them with SLAM-based mapping to enable sequential autonomous navigation without exter- nal landmarks. The proposed system incorporates two upward-facing detection models (bunch and flower models based on YOLO11), a TF-based coordinate logging node, Cartographer SLAM, AMCL localization, and the Nav2 navigation stack. During mapping, all de- tected cluster positions are recorded in the map frame. Offline DBSCAN clustering then fuses redundant detections and computes stable cluster centroids. These fused po- sitions are used as navigation goals. During execution, the robot sequentially visits each target while continuously adjusting its final alignment using real-time visual feedback. Three experiments were conducted: mapping accuracy evaluation, indoor navi- gation, and vineyard navigation. The mapping experiment with 10 known targets achieved a mean fusion error of 0.108 m, demonstrating the reliability of the hybrid de- tection–mapping pipeline. In indoor tests, single-goal navigation achieved 94.5% success, an average travel time of 22.3 s, and a mean final-position error of 0.041 m, confirming that the proposed navigation system functions accurately in disturbance-free environ- ments. Sequential four-goal navigation achieved 78.1% success, revealing minor error accumulation. In contrast, vineyard tests exhibited 88.5% single-goal success, 66.7% sequential success, and an average final-position error of 0.073 m, indicating that field- specific disturbances—such as uneven ground, LiDAR ground reflections, and odometry drift—are the primary sources of degradation rather than algorithmic limitations. These results demonstrate that the proposed fruit-centric navigation system enables practical and precise approaches to grape clusters without relying on external markers or GNSS. The hybrid online–offline approach provides a strong foundation for integrating manipulation tasks and ultimately reducing labor demands in viticulture. 본 연구는 샤인머스켓 포도 재배용 덩굴하(棚) 환경에서 작동하는 재배지원 로봇을 대상으로, 외부 표식이나 GNSS에 의존하지 않고 포도 송이를 기준으로 이동할 수 있는 자율주행 시스템을 제안한다. 송이 정리, 적과, 수확과 같은 재배 작업에서는 로봇이 포도 송이 바로 아래 위치해야 하며, 기존의 QR 마커 기반 또는 GNSS 기반 내비게이션은 송이의 형태 ·색 ·배치가 계절에 따라 크게 변하는 포도밭 환경에 충분히 대응하기 어렵다. 이를 해결하기 위해 본 연구는 상향 카메라 기반 YOLO11 송이/화방 검출 모델, TF 기반 좌표 로깅, Cartographer SLAM, AMCL 위치추정, Nav2 내비게이션 스택을 통합한 과실 중심(fruit-centric) 자율주행 프레임워크를 구축하였다. 매핑 과정에서 검출된 송이 좌표는 map 좌표계에서 기록되며, 오프라인 DBSCAN 클러스터링을 통해 중 복 검출을 융합하고 안정적인 송이 중심 좌표를 산출한다. 주행 단계에서는 이 융합 좌표를 목표점으로 활용하며, 상향 카메라의 실시간 검출 결과를 이용해 목표점 바로 아래로 미세 조정(move-to-bottom)을 수행한다. 본 연구는 세 가지 실험(매핑 정확도 평가, 실내 자율주행, 포도밭 자율주행)을 통해 시스템을 검증하였다. 10개의 기준 위치를 사용한 매핑 실험에서는 평균 0.108 m의 융합 오차를 달성하여 제안한 검출–매핑 통합 방 식의 신뢰성이 확인되었다. 실내 단일 목표 이동에서는 94.5% 성공률, 평균 22.3초, 평균 0.041 m 정지 오차를 기록하여 외란이 없는 환경에서는 알고리즘이 안정적으로 동작함을 보였다. 네 개 목표점을 연속 이동하는 실내 실험에서는 78.1% 성공률을 달성하였다. 반면, 실제 포도밭에서는 단일 목표 성공률 88.5%, 연속 이동 성공률 66.7%, 평균 0.073 m 정지 오차를 기록하였으며, 불규칙 지면, LiDAR의 지면 오검출, odometry 드리프트 등 환경 요인이 성능 저하의 주요 원인임을 확인하였다. 종합적으로, 제안한 과실 중심 자율주행 시스템은 외부 마커나 GNSS 없이도 포도 송이 위치를 기반으로 높은 정밀도의 접근이 가능하며, 향후 포도 재배 작업(적과, 수확 등)과 결합될 수 있는 실용적인 기반 기술임을 실험적으로 입증하였다.

Exploiting Cellular Signals for Navigation: 4G to 5G

Shamaei, Kimia ProQuest Dissertations & Theses University of Cali 2020 해외박사(DDOD)

Global navigation satellite systems (GNSS) have been the main technology used in aerial and ground vehicle navigation systems. As vehicles approach full autonomy, the requirements on the accuracy, reliability, and availability of their navigation systems become very stringent. Due to the limitations of GNSS, namely severe attenuation in deep urban canyons and susceptibility to interference, jamming, and spoofing, alternative sensors and signals are sought. The most common approach to address the limitations of GNSS-based navigation in urban environments is to fuse GNSS receivers with inertial navigation systems (INSs), lidars, cameras, and map matching algorithms. An alternative approach has emerged over the past decade, which is to exploit ambient signals of opportunity (SOPs), such as cellular, digital television, AM/FM, WiFi, and low Earth orbit (LEO) satellite signals. Among SOPs, cellular signals have attracted significant attention due to their inherently desirable attributes, including: abundance, geometric diversity, high received power, and large transmission bandwidth. Cellular systems have gone through five generations. Long-term-evolution (LTE) and new radio (NR) are the standards of the last two generations of wireless technology, namely 4th generation (4G) and 5th generation (5G), respectively. LTE has been developed and standardized in most countries over the past few years and currently has more than four billion users. The structure of NR signals has been finalized in 2019 and since then cellular providers have started rolling 5G out in major cities around the world. Cellular signals are not designed for navigation. In order to exploit cellular signals for navigation purposes, several challenges must be addressed: (1) specialized receivers are required to extract navigation observables from cellular signals, (2) cellular towers typically transmit from low elevation angles, causing multipath signals to be received alongside line-of-sight signals. Multipath can introduce error on the estimated navigation observables, which must be alleviated, (3) the achievable ranging accuracy in multipath-free and multipath-rich environments must be characterized, (4) navigation framework must be developed to localize the receiver using the derived navigation observables, and (5) cellular signals base stations' clock biases must be estimated, since they are not available to the receiver. This dissertation aims to address all of the above challenges for cellular LTE and NR signals. In particular, for LTE, first, a software-defined receiver (SDR) is proposed that is capable of (1) extracting the essential parameters for navigation from received LTE signals, (2) acquiring and tracking LTE signals transmitted from multiple eNodeBs, and (3) producing navigation observables from LTE signals including code and carrier phase and Doppler frequency measurements. Second, the accuracy of the produced measurements are derived as a function of carrier-to-noise ratio and signal transmission bandwidth. It is shown that LTE cell-specific reference signal (CRS) can provide higher precision compared to the LTE secondary synchronization signal (SSS) due to its high transmission bandwidth. Third, standalone and non-standalone navigation frameworks are proposed to localize the receiver using the generated navigation observables. Fourth, it is proposed to exploit the received LTE signal's time-of-arrival (TOA) and direction-of-arrival (DOA) to produce a navigation solution in cold-start applications, where there is no estimate of the receiver's initial state. For this purpose, an SDR is designed to jointly acquire and track TOA and DOA of LTE signals. For NR, first, an SDR is proposed that is capable of (1) acquiring synchronization signal (SS), physical broadcast channel (PBCH) signal, and its associated demodulation reference signal (DM-RS), which are transmitted on a block called SS/PBCH block and (2) tracking SS/PBCH block to produce code and carrier phase and Doppler frequency measurements from NR signals. Second, the precision of the derived code and carrier phase measurements are analyzed as a function of carrier-to-noise ratio and NR numerology. Finally, the statistics of the NR position estimation error are derived for different propagation channels. Throughout the dissertation, numerical and experimental results are provided to validate the theoretical contributions.

Development of an AR indoor navigation app based on a waypoint algorithm

코캐브 무라드 동아대학교 대학원 2026 국내석사

Abstract (Korean) 웨이포인트 알고리즘 기반 AR 실내 네비게이션 애플리케이션 개발 by KOKHAEV MURAD 동아대학교 스마트융합시스템공학과 부산, 한국 본 논문은 증강현실(AR) 시각화와 웨이포인트 기반 경로탐색 알고리즘을 통합한 혁신적인 실내 네비게이션 시스템의 개발 및 검증을 제시한다. 본 연구의 주요 목적은 직관적인 AR 인터페이스를 통해 사용자 경험을 향상시키고, 추가 인프라 없이 동작하는 강건한 솔루션을 구현함으로써 기존 실내 측위 시스템의 한계를 해소하는 데 있다. 구현된 시스템은 실내 공간을 상호 연결된 웨이포인트 네트워크로 모델링하는 그래프 기반 네비게이션 방식을 채택하였다. 핵심 경로탐색 메커니즘으로 사용된 너비 우선 탐색(BFS) 알고리즘은 최소한의 계산 비용으로 최적 경로를 산출한다. AR Foundation 프레임워크를 기반으로 정교한 공간 인식 및 평면 감지가 구현되어, 실제 환경에 3D 네비게이션 큐를 정밀하게 배치할 수 있다. 대학교 건물 환경에서 수행한 종합적 테스트 결과, 본 시스템은 기존 2D 지도 기반 네비게이션 대비 현저한 성능 향상을 보였다. 작업 완료 시간은 38% 단축되었으며, 네비게이션 오류는 71~74% 감소하였다. 시스템 사용성 척도를 활용한 사용자 경험 평가에서 86.4점의 우수한 점수를 기록하여 높은 사용자 만족도와 인터페이스 직관성을 확인하였다. 주요 혁신 요소로는 자동 경로탐색과 사용자 생성 AR 마커를 결합한 하이브리드 상호작용 모델, 기존 평면도를 활용한 확장 가능한 웨이포인트 매핑 방법론, 가상-물리 환경 정합 정확도를 보장하는 최적화된 공간 보정 기법 등이 포함된다. 본 시스템은 추가 인프라 없이 일반적인 모바일 하드웨어에서 효과적으로 동작하므로 다양한 실내 환경에 손쉽게 적용 가능하다. 본 연구는 복잡한 실내 환경에서 웨이포인트 기반 접근법의 실용적 타당성을 입증함으로써 증강현실 네비게이션 분야에 기여한다. 연구 결과는 공간 컴퓨팅 및 인간-컴퓨터 상호작용 분야의 향후 발전을 위한 기초를 마련하며, 신뢰할 수 있는 실내 네비게이션이 필수적인 교육, 의료, 상업 시설 등의 적용을 위한 유용한 통찰을 제공한다. Abstract (English) Development of an AR Indoor Navigation App based on a waypoint algorithm by KOKHAEV MURAD Dept. of Smart Convergence System Engineering Dong-A University Busan, Korea This thesis presents the development and validation of an innovative indoor navigation system that integrates augmented reality (AR) visualization with waypoint-based pathfinding algorithms. The primary objective of this research was to address the limitations of traditional indoor positioning systems by creating a robust, infrastructure-free solution that enhances user experience through intuitive AR interfaces. The implemented system utilizes a graph-based navigation approach, where indoor environments are modeled as networks of interconnected waypoints. The Breadth-First Search (BFS) algorithm serves as the core pathfinding mechanism, ensuring optimal route calculation with minimal computational overhead. The AR Foundation framework enables sophisticated spatial awareness and plane detection, allowing for precise placement of 3D navigation cues in real-world environments. Comprehensive testing was conducted in a university building environment, with experimental results demonstrating significant performance improvements over conventional 2D map-based navigation. The system achieved a 38% reduction in task completion time and 71-74% decrease in navigation errors. User experience assessment using the System Usability Scale yielded an exceptional score of 86.4, indicating high user satisfaction and interface intuitiveness. Key innovations include a hybrid interaction model combining automated pathfinding with user-generated AR markers, a scalable waypoint mapping methodology using existing floor plans, and optimized spatial calibration techniques ensuring accurate virtual-physical environment alignment. The system operates effectively on commodity mobile hardware without requiring additional infrastructure, making it readily deployable across various indoor settings. This research makes significant contributions to the field of augmented reality navigation by demonstrating the practical viability of waypoint-based approaches for complex indoor environments. The findings establish a foundation for future developments in spatial computing and human-computer interaction, offering valuable insights for applications in educational, healthcare, and commercial facilities where reliable indoor navigation is essential.

MEMS 관성센서를 이용한 차량 자세인지 Navigation 구현

李政勳 강원대학교 대학원 2008 국내석사

본 논문은 MEMS 관성센서(가속도 센서, 각속도 센서), 지자기 센서를 사용해 차량의 자세와 회전 방향에 관한 정보를 획득해 이를 GPS Navigation 시스템과 연동하여 사용할 수 있음을 보여주고 있다. 가속도 센서를 사용하여 이동 중인 차량의 Pitch정보를 획득하여 차량이 상승중인지 하강중인지를 알 수 있다. 이 정보를 바탕으로 Navigation 시스템이 차량의 고가도로, 오르막길, 내리막길 진입여부를 빠르게 판단 할 수 있게 된다. 또한 가속도 센서와 지자기 센서를 사용하여 차량의 회전 방향(yaw)정보와 진행방향 정보를 얻을 수 있음을 보여주고 있다. 이를 바탕으로 차량의 운행 중 경로 이탈을 빠르게 확인 할 수 있으며 차선 변경과 같은 작은 범위의 움직임도 알 수 있게 된다. 즉, 다양한 MEMS센서를 이용하면 기존의 GPS만을 이용한 시스템보다 더 지능적이고 효율적인 Navigation이 가능하다. 이를 위한 센서의 센싱 방법과 데이터 획득, 데이터 처리, 시뮬레이션에 관한 방법을 본 논문에서 제시한다. This study has shown that it could acquire information of vehicle posture and rotational direction to be used by the interlock with GPS Navigation system by using MEMS inertial sensor(acceleration sensor and angular velocity sensor) and geomagnetic sensor. By using acceleration sensor like accelerometer this study was to know whether the vehicle was moving upward or downward by acquiring the information of Pitch for vehicle during motion. Based on this information this was to judge it fast whether Navigation system could make the vehicle enter to elevated highway, uphill road, and downhill. In addition, this study has shown that it could acquire information of vehicle’s rotational direction(yaw) and of moving direction by using the accelerometer and geomagnetic sensor. On the basis of it this study could find the escape of route fast during moving vehicle. And, this study was to find out the movement of small range like the lane change of vehicle. After all, when it uses various MEMS sensors, it can be more intelligent and effective Navigation than the system using existing GPS only. Accordingly, this study was to suggest methods of sensing, data acquisition, data processing, and simulation for sensors.

Development of neuromorphic neural network model and robot for spatial navigation

김성현 Graduate School, Korea University 2017 국내석사

Spatial navigation is a fundamental for most mobile navigating robots, including autonomous self-driving robots, exploration robots, and rescue robots. However, previous spatial navigation models for these robots were unable to adapt in unexplored, novel, or complex environments. To overcome these limitations, a neuromorphic spatial navigation model based on the spatial information processing of animals is described in this thesis. This neuromorphic neural network model was constructed using a computational model of spatial information processing neurons, which include boundary vector, head direction, grid, and place cells in the rodent brain. Firstly, spatial navigation was simulated using virtual rats in a virtual environment to verify whether the neuromorphic neural network model could perform spatial navigation, especially collision avoidance and goal-directed navigation task. In order to enhance the spatial navigation efficiency, the role of spatial information processing neurons in the spatial navigation was investigated. Finally, the neuromorphic neural network model was applied to control a mobile robot for spatial navigation. As a result, the mobile robot successfully performed the spatial navigation in unexplored, novel and complex environment. Overall, the neuromorphic neural network model based on the spatial information processing of the rodent brain can be considered as a candidate for the most efficient spatial navigation model. 내비게이션 기술은 다양한 종류의 센서를 이용해 주변 장소에 대한 정보들을 획득하고, 이를 바탕으로 특정 목적을 수행하기 위한 이동 좌표를 결정하는 기술을 말하며, 자율주행 자동차, 탐사로봇 그리고 재난 구조 로봇 등, 많은 로봇 분야에서 응용되고 있다. 하지만, 기존의 내비게이션 기술은 미탐사 지역이나 복잡하고 새로운 지역에서 적용하기 어렵다는 단점이 있다. 이러한 한계점을 극복하기 위해, 본 연구에서는 동물의 뇌에서 일어나는 장소정보처리 원리를 이용한 새로운 내비게이션 모델을 구축하고자 한다. 동물의 해마체(hippocampal formation)은 장소정보를 처리하는데 특화된 영역이다. 특히, 환경의 경계면에 대한 정보를 처리하는 경계면 세포(boundary vector cell), 동물의 머리방향에 대한 정보를 처리하는 머리방향 세포(head direction cell), 공간의 격자 좌표 정보를 처리하는 격자 세포(grid cell), 그리고 특정 장소에 대한 정보를 처리하는 장소 세포(place cell)가 발견되었다. 이러한 장소정보처리 세포는 해마체 내에서 장소정보처리 신경회로를 통해 신경신호를 공유하며 장소정보를 처리하게 된다. 본 연구는, 뇌 장소정보처리 신경세포의 형태학적 구조와 전기생리학적 특성을 정교하게 모사할 수 있는 뉴로모픽 단일신경세포 모델을 구축하고, 해부학적으로 알려진 장소정보처리 신경회로를 기반으로 뉴로모픽 신경망 모델을 구축하는 것을 목적으로 한다. 또한, 구축된 뉴로모픽 신경망 모델을 가상 쥐 행동 시뮬레이터에 연결하여, 실제 쥐처럼 공간에서의 내비게이션을 수행할 수 있는지 연구하였다. 나아가, 쥐의 행동을 모사하는 쥐 로봇을 개발하고, 뉴로모픽 신경망 모델이 실제 환경에서 로봇을 통해 내비게이션을 수행하는지 관찰하였다. 결과적으로, 본 연구를 통해 구축한 뉴로모픽 신경망 모델은 미로 환경, 장애물 환경에서 성공적으로 충돌회피를 수행할 수 있었고, 숨겨진 목적지를 탐색하며 학습하는 목적지 지향 내비게이션도 성공적으로 수행하였다. 또한, 충돌 회피를 위해선 경계면 세포로부터 생성된 경계면 정보와, 회피 학습을 위해선 머리방향 세포로부터 생성된 머리방향 정보가 중요함을 밝혔다. 나아가, 목적지 탐색 내비게이션 학습은 장소세포와 목적에 대한 정보를 표현하는 세포간의 학습이 중요함을 밝혔다. 이러한 결과는 뉴로모픽 내비게이션 모델을 이용해 이전 내비게이션 모델들이 수행하기 어려웠던, 미탐사 지역이나 복잡하고 새로운 지역에서도 실시간으로 학습을 통해 최적의 내비게이션을 수행할 수 있음을 밝혔으며, 실험적으로 연구하기 어려운 뇌 장소정보처리 세포의 역할을 규명하는데 중요한 역할을 할 수 있을 것으로 기대된다.

Football enhances Context–dependent Spatial navigation via the Entorhinal hippocampal circuit plasticity

김상윤 서울대학교 대학원 2026 국내석사

생물체는 생존을 위해 그들의 환경을 탐색하며, 이러한 공간 탐색 행동은 생존 활동의 기반이 되는 인지 기능이다. 포유류의 경우, 내후각피질-해마 회로(entorhinal-hippocampal circuit)가 공간 탐색 행동의 주요 기전으로 알려져 있으며, 회로 내의 격자세포(grid cell)와 장소세포(place cell)는 환경에 대한 공간적 신경 표상을 형성한다. 인간 역시 이 회로에 의존하여 일상생활에서 공간 탐색을 수행해왔으나, 최근 자동화된 내비게이션 시스템의 보편화로 인해 능동적인 공간 탐색 경험이 감소하면서, 인간 고유의 공간 탐색 능력이 저하될 가능성에 대한 우려가 제기되고 있다. 이에 따라 인간의 공간 탐색 능력을 유지하고 향상시킬 수 있는 전략을 규명할 필요성이 대두되고 있다. 축구는 경기 특성상 지속적인 공간 탐색을 요구하는 스포츠 종목 중 하나로, 축구 경험이 내후각피질-해마 회로의 가소성을 유도할 가능성을 시사한다. 이러한 관점에서 축구는 현대 사회에서 공간 탐색 능력을 유지하고 향상시킨는 데 기여할 수 있는 잠재적 활동이 될 수 있을 것이다. 그러나, 최근 공간 탐색이 수행되는 환경에 따라 다르게 나타날 수 있다는 연구들이 제시되고 있으며, 이러한 공간 탐색의 맥락 의존적 특성에 대한 신경학적 메커니즘을 규명하기 위한 연구들이 계속해서 진행되고 있다. 따라서 축구 경험이 서로 다른 맥락 전반에 걸쳐 공간 탐색 능력을 향상시킬 수 있는지에 대해서는 아직 불분명하다. 본 연구는 프로축구선수 집단과 대조군을 대상으로 기능적 자기공명영상(functional magnetic resonance imaging, fMRI)을 활용하여 공간 탐색 능력과 내후각피질-해마 회로의 구조적, 기능적 가소성을 분석하였다. 그 결과, 프로축구선수 집단은 축구 특이적인 맥락에서만 대조군에 비해 더 높은 공간 기억의 정확도와 일관성을 보였으며, 이는 우측 해마 영역의 활성도 증가와 함께 나타났다. 또한, 프로축구선수 집단은 구조적으로 내측 내후각피질의 부피가 더 컸으며, 해마에서 더 높은 분획이방성(fractional anisotrophy, FA)을 보였다. 그러나 이러한 구조적 우수성이 일반적인 맥락에서 공간 탐색 수행 능력이나 뇌 활성도의 차이로 이어지지는 않았다. 따라서, 해당 결과는 축구 경험이 내후각피질-해마 회로의 구조적 및 기능적 가소성을 유도하지만, 맥락 의존적으로 공간 탐색 능력이 향상된다는 것을 보여준다. 이는 인간의 공간 탐색이 본질적으로 맥락에 따라 나타나는 특성을 지님을 시사한다. 나아가 스포츠를 통해 공간 탐색 능력을 향상시키기 위해서는 다양한 맥락에서의 공간 탐색 경험에 참여하는 것이 중요할 수 있음을 제안한다. Mobile organisms navigate their environment to optimize survival, and this navigational behavior is a fundamental cognitive function conserved across species. In mammals, this capacity is supported by the entorhinal–hippocampal circuit, where spatially tuned cells such as grid cells and place cells form neural representations of the environment. In recent decades, the pervasive use of automated navigation systems in human society has raised concerns that reduced engagement in active spatial navigation may contribute to the deterioration of intrinsic navigational abilities. Therefore, it is necessary to identify strategies that preserve and enhance human navigational abilities. Football is a sport that inherently demands navigation, and it suggests that football experience may induce plasticity within the entorhinal–hippocampal circuit. From this perspective, football may contribute to the preservation and enhancement of navigational abilities in human society. However, accumulating evidence suggests that spatial navigation is sensitive to contextual factors, and the neural mechanisms underlying these context-dependent characteristics remain insufficiently understood. Therefore, whether football experience can lead to improvements in navigation that extend across different contexts remains enigmatic. To explore this question, we conducted a functional MRI (Magnetic Resonance Imaging) experiment comparing the navigational performance and entorhinal–hippocampal activation of professional football players and a control group. We further examined how these behavioral and neural patterns differed across two contexts: a football-specific environment and a general navigation environment. Football players showed higher accuracy and greater consistency of spatial memory compared with the control group, but only in the football-specific context. This behavioral superiority was accompanied by increased activation in right hippocampal subregions during navigation. Additionally, players exhibited larger medial entorhinal cortical volume and higher fractional anisotropy (FA) values, indicating a more organized hippocampal microstructure. However, these structural characteristics did not translate into differences in navigational performance or neural activation in the general navigation context. Taken together, the findings indicate that football experience induces structural and functional plasticity within the entorhinal–hippocampal circuit and enhances spatial navigation in a context-dependent manner. This study validates that football experience induces structural and functional plasticity within the entorhinal–hippocampal circuit and enhances spatial navigation in a context-dependent manner, and further suggests that this circuit inherently reflects the context-dependent characteristics of human navigation. Therefore, to enhance broader navigational abilities through sport, it may be necessary to engage in experiences across diverse contexts.

Autonomous Navigation of a Two-wheeled Mobile Robot with Timed Elastic Band Planner in Greenhouse Environment

CHEN TEAN 전남대학교 2024 국내박사

첨단 대형 온실은 농업 발전의 중요한 발전을 나타내며, 운영 효율성 개선과 노동력 감소를 갖어왔습니다. 온실 환경에 맞춤형으로 개발된 자율이동 로봇은 기존 노동력 기반의 농업과 단순 자동화 기술의 제약을 해결하는 데 중요한 역 할을 하고 있습니다. 하지만, 온실 내 GPS 신호의 이용이 불가능하여 위치 추정 이 어려운 점은 온실 자율이동 로봇의 한계로 나타나고 있습니다. 이로 인해 온 실 내 자율주행 및 위치 추정의 구현이 복잡해지기 때문에 온실 내 자율주행 내비게이션과 정밀 위치추정 기술은 온실 무인 자동화를 위한 핵심기술입니다. 본 연구의 목표는 온실 내 도로 및 레일 표면에 적응할 수 있는 새로운 바 퀴 메커니즘을 개발하고, 센서 시스템과 알고리즘을 통합하여 정밀한 내비게이션 을 보장하며, 이동 로봇의 운동학적 모델링을 수행하고, Hector SLAM을 사용해 고해상도 온실 지도를 작성하며, 적응형 몬테카를로 위치 추정(AMCL)과 비주얼 SLAM 기술을 통해 위치 추정 정확도를 향상시키는 것입니다. 본 연구의 결과는 제한된 농업 환경인 온실 내에서 SLAM 기반의 이동 로 봇 내비게이션을 통해 실증되었습니다. 먼저, 마커를 레일 중앙에 위치시켜 레일 중앙 포인트를 추출하고, 컴퓨터비전 기술을 사용해 정확한 중앙 포인트를 결정 합니다. 온실 환경이 맵핑된 후 마커를 제거하고, 인식된 레일 중앙 포인트를 이 용하여 레일 인식의 계산 복잡성을 줄일 수 있습니다. 이동로봇의 위치 추정은 라이다 스캔을 2D 그리드 지도와 정렬한 후 적응형 몬테카를로 위치 추정 (AMCL) 알고리즘을 통해 실시간으로 인지됩니다. 글로벌 내비게이션 경로 계획 은 Dijkstra 알고리즘을 사용하며, 로컬 내비게이션 경로 계획은 TEB 접근 방식 을 사용합니다. 실험 결과, 온실 환경 내에서 로봇의 위치 누적 오류를 최소화하는 AMCL 및 비주얼 SLAM 기반의 통합 위치 시스템의 효과가 입증되었습니다. 0.4, 0.6, 0.8 m/s의 로봇 주행속도에서 최대 측면 오차는 0.55cm 미만이었으며, 평균 제 곱근 오차(RMSE)는 0.2cm 미만이었습니다. x 및 y 축 방향의 평균 글로벌 위치 오류는 5cm 미만이었으며, 궤적 추적 정확도는 96%를 초과했습니다. 이러한 실 험결과는 제안된 시스템을 통해 온실 환경 내 자율주행 로봇의 정밀한 내비게이 션이 가능함을 보여주었습니다. The large greenhouses signify a crucial advancement in agricultural development, improved operational efficiency, and decreased labor requirements. Autonomous mobile robots tailored for greenhouse environments are essential for addressing the challenges associated with manual labor and the limitations of traditional automation technologies. Moreover, indoor localization presents a unique challenge, as conventional GPS signals are unavailable within greenhouse settings. This limitation complicates efforts to implement autonomous navigation and precise localization for indoor agricultural robots. Addressing these complexities is vital for achieving the full potential of greenhouse automation to enhance productivity and sustainability in agriculture. The objectives encompass the development of a novel wheel mechanism capable of adapting to road and rail surfaces within greenhouses, along with integrating advanced sensor systems and algorithms to ensure precise navigation, the kinematic modeling of mobile robots, using Hector Simultaneous Localization and Mapping (SLAM) for high- fidelity greenhouse mapping, and improving localization using adaptive Monte Carlo localization (AMCL) and visual SLAM techniques for enhanced localization accuracy. The results were demonstrated in a greenhouse using SLAM based mobile robot navigation within confined agricultural environments, specifically in greenhouses. Our methodology involves extracting rail center points with markers; the markers are strategically placed at the center of rails, and lidar technology is utilized for accurate center point determination. Once the environment is mapped, the markers are removed, and the precise rail center points are employed for subsequent robot navigation. During localization, known positions of rail center points in a 2D space prevent the need for additional recognition system, thereby reducing computational complexity. The mobile robot localization algorithm aligns lidar scans with a 2D occupancy grid map, providing real-time position information through the Adaptive Monte Carlo Localization (AMCL) algorithm. This integration of lidar data with the existing 2D map achieves global positioning. Global navigation path planning employs the Dijkstra algorithm, while local navigation path planning utilizes the Timed-Elastic-Band (TEB) approach. Experimental results demonstrate the effectiveness of the integrated positioning system based on AMCL and visual SLAM in minimizing the robot's positioning cumulative error within the greenhouse environment. Across three travel speeds of 0.4, 0.6, and 0.8 m/s, the maximum lateral error remains under 0.55 cm with a root mean square error (RMSE) of less than 0.2 cm. The average global positioning error in the x and y-axis directions is less than 5 cm, and the trajectory tracking accuracy rate exceeds 96%. The autonomous robot's precise and secure navigation within the greenhouse environment showcases the accuracy achieved through the overall proposed system. Keywords: Mobile Robot, Greenhouse, Autonomous Navigation, Obstacle Avoidance, Simultaneous Localization and Mapping (SLAM)