Applications of intelligent robot or vehicle systems have become more and more important and popular in our daily life. This thesis can be divided into three main directions for the design of intelligent vehicle, including trajectory planning of mobile vehicle, position control of robot arm system and application of image recognition. For the trajectory planning of vehicle system design, the camera installed on the clip of robot arm is applied to capture images with appropriate correction. The moving angle of vehicle is calculated by analyzing the captured images. And then, the vehicle is able to be manipulated in order to achieve the purpose of object tracking.

For the position control of robot arm, the mathematical model is developed by understanding the movement pattern. The basic movement of robot arm system is constructed by using the method of forward and inverse kinematics. The information of object obtained by appropriate image analysis could be employed to obtain the angle of rotation for each joint, so the desired goal is achieved by correct motor operation. In the part of image recognition, useful images are captured by the camera installed on the clip of robot arm. These images must be effectively analyzed to obtain the position, shape and color information of the object. The utilization of image analysis is applied for vehicle moving, object grabbing and object placing. The part of vehicle moving is to find the desired angle to track objects. The parts of grabbing and placing objects are related to the analysis of the information of bricks and sockets, respectively.

As a result, the intelligent vehicle has the potentials to move and stop in front of the target wooden box. And then, after the robot arm grabs the brick placed on the wooden box, the vehicle would turn back to its starting point. Finally, the robot arm places the brick into the socket with correct color and shape. In addition, the implementation of intelligent vehicle design is given to illustrate the successful achievement of the pick-and-place control objective. 

The Autonomous Underwater Vehicles (AUVs) play an important role of tracing submarine cables, surveying submarines, and inspecting underwater structures, all of which require considerable operation time. However, the battery capacity limits AUV operations to a specific duration of time. Even through modern power devices allow long operational periods, AUVs must return to the mother ship for recharging. To solve this problem, an underwater docking function is employed to recharge the battery. Most AUV docking operation studies using visual information are based on monocular camera to calculate the pose between a target and an AUV. The drawback of these studies is that the accuracy of the distance measurement in the depth direction is not sufficient for high homing accuracy applications. Therefore, a stereo-vision-based docking system is designed to fulfill the operations. In this system, the relative pose, i.e., position and orientation, between an AUV and a target object is estimated by using a Real-time Multi-step Genetic Algorithm (RM-GA), which makes estimation of the real-time 3D pose.

This thesis proposes an innovative stereo-vision-based docking system for battery recharging in the real-sea environment. In the proposed method, visual information is directly used for real-time feedback control. In addition, the optimization method RM-GA is developed and implemented according to the concept of dynamic image optimization for real-time target tracking. Moreover, when the AUV operation is conducted in an unstructured environment such as near shore or on the seabed, the environmental disturbances cause the most challenging and inevitable problem with the operation. These disturbances include changes in ambient illumination, high turbidity in the water and the external forces of ocean currents. To overcome such disturbances, the author designed a stereo-vision-based docking system adaptable to different environments. The docking system utilizes a luminous 3D marker, a current-adaptive docking station, and a Remotely Operated Vehicle (ROV) with stereo-vision camera to experiment in a real-sea environment. The results of the experiment verify that the proposed system can provide high homing accuracy and robustness against environmental disturbances. Finally, the proposed system was applied to the AUV to conduct the battery recharging operation, and the result of the operation proves the effectiveness of the proposed AUV docking system.

In the first part of the thesis, both hardware and software of the docking system are discussed. The hovering-type ROV (Delta-150, manufactured by QI co. ltd) with a stereo-vision camera is used to perform the docking operation. The luminous 3D marker installed on the docking station is used as the target object for the docking operation that can be used at different illumination environment. The current-adaptive docking station is produced to solve the problem of the change of the ocean current direction. In the second part of the thesis, a long-term continuous recognition experiment conducted in a simulation pool from daytime to nighttime at Okayama University. Experimental result shows the proposed system can recognized the target object in various illumination environments. In the third part of the thesis, a real-sea experiment, a long-term continuous iterative docking experiment, is performed in the private sea area of Okayama University Ushimado Nearshore Laboratory in Ushimado Town, Setouchi City, Okayama Prefecture, Japan. The experimental result shows that the proposed docking system successfully completes the long-term continuous iterative docking operation in the real-sea environment. This demonstrates the effectiveness of the proposed docking system in the real-sea environment. The final part of the thesis is to confirm the effectiveness of the proposed system with an AUV. This experiment uses the hovering-AUV named as Hobalin, which has been developed by the National Maritime Research Institute Japan. The experiment was a battery recharging experiment, conducted in the test tank with clear still water. In the experiment, the AUV approached the docking station with the normal navigation system by Inertial Navigation System (INS) and Doppler Velocity Log (DVL), and automatically performed the docking operation with only visual information from the stereo-vision camera. The experimental results show that the proposed system applied to the AUV works properly, and the integrated INS error is reset after the AUV completes the docking operation.

This thesis presents an effective docking system to conduct the docking operation in a real-sea environment and to perform the battery recharging operation with the AUV. To demonstrate the practical usefulness of the proposed system, the docking experiment was conducted in the private sea area of Okayama University Ushimado Nearshore Laboratory, and the battery recharging experiment was carried out with a hovering-AUV. The experimental results verify the robustness of the proposed system under different conditions of illumination, turbidity, and external force, thereby providing a reliable solution to extend the operation time of the AUV.

Coral reefs are among the most important marine ecosystems on Earth, primarily found in tropical and subtropical waters. They provide rich biodiversity and crucial ecological services, significantly impacting coastal economies, tourism, and fisheries. However, global climate change, ocean acidification, and human activities such as overfishing and marine pollution have placed immense pressure on coral reefs, leading to frequent occurrences of coral bleaching and death. This makes the protection and restoration of coral reefs a major global environmental and scientific challenge. To better understand and safeguard these valuable ecosystems, there is a need to develop innovative research methods. This paper presents a novel approach that combines active ultrasonic technology with continuous data detection for coral identification and classification. The core of this method involves emitting high-frequency ultrasonic pulses toward the corals and utilizing continuous data detection techniques to analyze the reflected waveforms in real time, thereby obtaining detailed information about the internal structure of the corals. Our findings reveal significant differences in the reflected waveforms among various coral species, providing a new basis for accurate coral identification and classification.