Research Areas of Interests
Vehicle Dynamics, Guidance, Navigation and Control
Modeling and analysis of vehicle dynamics, linear/nonlinear control and estimation, adaptive control, dual control, non-conventional filter design, underactuated/nonholonomic systems

Intelligence and Autonomy for Manned/Unmanned Vehicles and Mobile Robots
Robotic system design and analysis, motion planning for manipulators and robotic vehicles, machine learning (supervised/unsupervised), neural network and computer vision techniques

Research Overview

USVs (unmanned surface vehicles) remove the operator from the vehicles and operate on the surface of water. They have advantage of performing hazardous or time-consuming missions in ocean environment.
UUVs (unmanned underwater vehicles) are able to operate underwater without a pilot. They have significantly played an important role in exploratory missions in underwater that are inaccessible to humans.
The underwater glider is a new type of autonomous unmanned vehicles(AUVs). A glider can glide through the ocean while controlling its attitude by using an internal moving mass system and bouyancy.

Automatic detection and tracking of target ship using monocular vision, Vision-based collision avoidance of USVs
Precision navigation and mapping in GPS-denied areas (SLAM), 3D reconstruction by sensor fusion
Visual simultaneous localization and mapping (SLAM), Relative pose estimation from unstructured seafloor images, Feature-based automatic visual mapping
Autonomous underwater manipulation with underwater robot, Vision-based relative pose estimation in underwater environment, Underwater navigation

Geophysical navigation for submarines or UUVs that exploits topographic and geomagnetic information to enable precise long-term and/or long-range navigation
Panel-based bathymetric SLAM which represents the terrain surface as rectangular panels using bathymetric sensors
Parameterized SLAM in an indoor structure through parameterized map-building which represents the surrounding structures using a small number of geometric parameters

Optimization of the fish-like locomotion is a problem of finding a swimming pattern of fish-like creatures using flapping fins. By dividing complex fin motions into a set of motion segments properly, we can obtain the optimal fin motion of the fish by learning each motion easily.
Path Opmimization for Marine Vehicles is finding an optimal path in complex ocean current field applying dynamic characteristic. Q-learning algorithm which is one of the reinforcement learning algorithm was employed for path optimization algorithm.
Ship Performance Analysis & Modeling
There are many variables which affects ship performance in real ocean such as ship speed, loading conditions, weather and the relationship between those are quite complicated. Many empirical formula and the equation based on model-test data had been developed, however, accuracy and reliability of the method is not satisfactory. Therefore, machine learning algorithms is applied to form the model based on the measured data in real ocean.

Copyright ⓒ Mobile Robotics and Intelligence Lab., 2018
Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology
291 Daehak-ro, Yuseong-gu, Daejeon 305-701, Republic of Korea