This paper presents an optimal path planning strategy for aerial searching and surveying of a user-designated area using multiple Unmanned Aerial Vehicles (UAVs). The method is designed to deal with a single unseparated polygonal area, regardless of polygonal convexity. By defining the search area into a set of grids, the algorithm enables UAVs to completely search without leaving unsearched space. The presented strategy consists of two main algorithmic steps: cellular decomposition and path planning stages. The cellular decomposition method divides the area to designate a conflict-free subsearch-space to an individual UAV, while accounting the assigned flight velocity, take-off and landing positions. Then, the path planning strategy forms paths based on every point located in end of each grid row. The first waypoint is chosen as the closest point from the vehicle-starting position, and it recursively updates the nearest endpoint set to generate the shortest path. The path planning policy produces four path candidates by alternating the starting point (left or right edge), and the travel direction (vertical or horizontal). The optimal-selection policy is enforced to maximize the search efficiency, which is time dependent; the policy imposes the total path-length and turning number criteria per candidate. The results demonstrate that the proposed cellular decomposition method improves the search-time efficiency. In addition, the candidate selection enhances the algorithmic efficacy toward further mission time-duration reduction. The method shows robustness against both convex and non-convex shaped search area.

Recently, robots have been actively utilized for logistics and delivery services in various places such as restaurants, hotels, and hospitals. In addition, it provides a safer environment, convenience, and cost efficiency to the customers. However, when it comes to autonomous delivery in a multi-floor environment, the task is still challenging. Especially for wheeled mobile robots, it is necessary to deal with elevators to perform the last-mile delivery services. Therefore, we present a multi-floor route planning algorithm that enables a wheeled mobile robot to traverse an elevator for the delivery service. In addition, an elevator boarding mission algorithm was developed to perceive the drivable region within the elevator and generate a feasible path that is collision-free. The algorithm was tested with real-world experiments and was demonstrated to perform autonomous postal delivery service in a multi-floor building. We concluded that our study could contribute to building a stable autonomous driving robot system for a multi-floor environment.

This study introduces a smart wrist band system with pressure measurements using wrist skin curvature variation due to finger motion. It is easy to wear and take off without pre-adaptation or surgery to use. By analyzing the depth variation of wrist skin curvature during each finger motion, we elaborated the most suitable location of each Force Sensitive Resistor (FSR) to be attached in the wristband with anatomical consideration. A 3D depth camera was used to investigate distinctive wrist locations, responsible for the anatomically de-coupled thumb, index, and middle finger, where the variations of wrist skin curvature appear independently. Then sensors within the wristband were attached correspondingly to measure the pressure change of those points and eventually the finger motion. The smart wrist band was validated for its practicality through two demonstrative applications, i.e., one for a real-time control of prosthetic robot hands and the other for natural human-computer interfacing. And hopefully other futuristic human-related applications would be benefited from the proposed smart wrist band system.

This paper shows a design of a gripper with an adjustable working area to depalletize a delivery cardboard box of various sizes. The gripper should pick the box with only one flat surface to lift the boxes stacked close to each other. The lift force of the gripper is the vacuum suction force. To handle boxes of various sizes, the gripper adjusts the working area. The gripper operates four vacuum generators independently. The simultaneous rotation on different axes of four gripper-arms with suction cups moves the position of suction force. The six operation modes of the gripper are divided into the size of the working area. The operation mode is determined according to the size of the top side of the box. Experiments are conducted by lifting the box of various sizes. The gripper can pick the box of various sizes without vacuum leaks from unused cups. Also, the experiments verify the improvement of stability of the box by adjusting the working area of the gripper. The gripper can lift the box without deformation of the box by adjusting the working area.

This paper introduced how to construct Denavit-Hartenberg (DH) parameters from the Unified Robot Description Format (URDF). URDF is convenient for describing a robot even though the robot is very complex. On the other hand, DH convention is not an easy notation for many novices who want to describe a robot. Therefore, most vendors provide URDF and users prefer to use URDF to describe a robot. However, some controllers or algorithms are based on DH parameters to perform kinematics, dynamics, control, etc. To connect URDF and DH parameters, we present a three-step approach to construct DH parameters from URDF. The first step is to define the joint axis for constructing DH parameters. The second step is constructing DH parameters to define joint character. The final step is constructing DH parameters to define the coordinate frame of the child link. This approach is based on intuitive vector calculation and guarantees the uniqueness of DH parameters. To verify our approach, we applied our approach to a simple one-link robot, a manipulator with 6 DOF, and a quadruped robot with 3 DOF per leg. We verified that our approach worked well based on forward kinematic results.

Many studies have been conducted to ensure that Visual Place Recognition is reliable in various environments, including edge cases. However, existing approaches use visible imaging sensors, RGB cameras, which are greatly influenced by illumination changes, as is widely known. Thus, in this paper, we use an invisible imaging sensor, a long wave length infrared camera (LWIR) instead of RGB, that is shown to be more reliable in low-light and highly noisy conditions. In addition, although the camera sensor used to solve this problem is an LWIR camera, but since the thermal image is converted into RGB image the proposed method is highly compatible with existing algorithms and databases. We demonstrate that the proposed method outperforms the baseline method by about 0.19 for recall performance.

In general, in order to operate an autonomous underwater vehicle (AUV) in an underwater environment, a navigation system such as a Doppler Log (DVL) using a Doppler phenomenon of ultrasonic waves is used for speed and direction estimation. However, most of the ultrasonic sensors in underwater is large for long-distance sensing and the cost is very high. In this study, not only canal neuromast on the fish's lateral lines but also superficial neuromast are studied on the simulation to obtain pressure values for each pressure sensor, and the obtained pressure data is supervised using spherical CNN. To this end, through supervised learning using pressure data obtained from a pressure sensor attached to an underwater vehicle, we can estimate the speed and angle of the underwater vehicle in a quasi-static underwater environment and propose a method for a non-ultrasonic based navigation system.

This paper proposes an Extended and Adaptive Inverse Perspective Mapping (EA-IPM) model that can obtain an accurate bird's-eye view (BEV) from the forward-looking monocular camera on the sidewalk with various curves. While Inverse Perspective Mapping (IPM) is a good way to obtain ground information, conventional methods assume a fixed relationship between the camera and the ground. Due to the nature of the driving environment of the mobile robot, there are more walking environments with frequent motion changes than flat roads, which have a fatal effect on IPM results. Therefore, we have developed an extended IPM process to be applicable in IPM on sidewalks by adding a formula for complementary Y-derive processes and roll motions to the existing adaptive IPM model that is robust to pitch motions. To convince the performance of the proposed method, we evaluated our results on both synthetic and real road and sidewalk datasets.

As cameras have become primary sensors for mobile robots, vision based Simultaneous Localization and Mapping (SLAM) has achieved impressive results with the recent development of computer vision and deep learning. However, vision information has a disadvantage in that a lot of information disappears in a low-light environment. To overcome the problem, we propose an image enhancement method to perform visual SLAM in a low-light environment. Using the deep generative adversarial models and modified gamma correction, the quality of low-light images were improved. The proposed method is less sharp than the existing method, but it can be applied to ORB-SLAM in real time by dramatically reducing the amount of computation. The experimental results were able to prove the validity of the proposed method by applying to public Dataset TUM and VIVID++.

We introduce tightly-coupled GNSS-LiDAR-Inertial state estimator, which is capable of SLAM (Simultaneously Localization and Mapping) and autonomous driving. Long term drift is one of the main sources of estimation error, and some LiDAR SLAM framework utilize loop closure to overcome this error. However, when loop closing event happens, one's current state could change abruptly and pose some safety issues on drivers. Directly utilizing GNSS (Global Navigation Satellite System) positioning information could help alleviating this problem, but accurate information is not always available and inaccurate vertical positioning issues still exist. We thus propose our method which tightly couples raw GNSS measurements into LiDAR-Inertial SLAM framework which can handle satellite positioning information regardless of its uncertainty. Also, with NLOS (Non-light-of-sight) satellite signal handling, we can estimate our states more smoothly and accurately. With several autonomous driving tests on AGV (Autonomous Ground Vehicle), we verified that our method can be applied to real-world problem.

This paper proposes a disturbance observer-based control for 6-DOF remotely operated underwater vehicles with model uncertainties. The sum of external disturbance and the forces generated from model parameters except for the inertial matrix of the hydrodynamic model is defined as a lumped disturbance in this paper. Then, the lumped disturbance caused by model uncertainties and the external forces is estimated using the disturbance observer. Fortunately, the disturbance observer is constructed as a linear form because all the elements of the inertial matrix of the hydrodynamic model are constants. To verify the proposed control scheme, we show that the actual lumped disturbance is similar to the estimated lumped disturbance obtained by the disturbance observer. Finally, the position tracking performance in the disturbance environment is confirmed through the comparative study with a traditional inverse dynamics PD controller.

Lithium-ion batteries are actively used in various industrial sites such as field robots, drones, and electric vehicles due to their high energy efficiency, light weight, long life span, and low self-discharge rate. When using a lithium-ion battery in a field, it is important to accurately estimate the SoC (State of Charge) of batteries to prevent damage. In recent years, SoC estimation using data-based artificial neural networks has been in the spotlight, but it has been difficult to deploy in the embedded board environment at the actual site because the computation is heavy and complex. To solve this problem, neural network lightening technologies such as network pruning have recently attracted attention. When pruning a neural network, the performance varies depending on which layer and how much pruning is performed. In this paper, we introduce an optimized pruning technique by improving the existing pruning method, and perform a comparative experiment to analyze the results.

In this paper, the development of an autonomous electric vehicle for logistics with a robotic arm is introduced. The manual driving electric vehicle was converted into an electric vehicle platform capable of autonomous driving. For autonomous driving, an encoder is installed on the driving wheels, and an electronic power steering system is applied for automatic steering. The electric vehicle is equipped with a lidar sensor, a depth camera, and an ultrasonic sensor to recognize the surrounding environment, create a map, and recognize the vehicle location. The odometry was calculated using the bicycle motion model, and the map was created using the SLAM algorithm. To estimate the location of the platform based on the generated map, AMCL algorithm using Lidar was applied. A user interface was developed to create and modify a waypoint in order to move a predetermined place according to the logistics process. An A-star-based global path was generated to move to the destination, and a DWA-based local path was generated to trace the global path. The autonomous electric vehicle developed in this paper was tested and its utility was verified in a warehouse.

In this paper, the cinch bag typed fruit harvesting gripper was proposed. This gripper is focused on preventing problems that we found from the related research and setting the breakthrough as a design condition according to the harvest failures of other related studies. The cover part is designed to overcome the surrounding obstacles of target fruits such as tomato, Korean melon, and sweet pepper. The measurement of maximum load showed that the well-grasped target object, such as a spherical object with 65 mm of diameter, is unable to slip in a range of 0 kg to 10 kg. The fact that the gripper allows from 4 cm to 6 cm of positional error was shown in the measurement of positional error tolerance. And the cover part of this gripper showed that the suggested gripper can grab a target object without being obstructed by leaves and stems. Finally, it was proved that the gripper satisfied the design conditions through the measurement of contacting force, which showed it is appropriate for grasping an actual fruit without damage.

In this study describes the design of Exo-suit to assist those who work in unstructured positions. The present study aimed to analyze various types of work, especially those performed in unstructured postures by heavy industry workers. Based on the motion capture analysis results, an attempt was made to develop a shoulder muscle-assistive Exo-suit capable of assisting a wearer who is working using shoulder muscles. In the present study, as the first step of developing a shoulder muscle-assistive Exo-suit, different working scenarios were simulated, and the corresponding motion data were estimated using motion capture devices. The obtained motion data were reflected in the design of the Exo-suit. The main structure of the shoulder muscle-assistive Exo-suit was made of a carbon fiber-reinforced composite to obtain the weight reduction. The shoulder muscle assistive Exo-suit was designed to fully cover the range of motion for workers working in unstructured postures.

Research on walking assistance exoskeletons that provide optimized torque to individuals has been conducted steadily, and these studies aim to help users feel stable when walking and get help that suits their intentions. Because exoskeleton auxiliary efficiency evaluation is based on metabolic cost savings, experiments on real people are needed to evaluate continuously evolving control algorithms. However, experiments with real people always require risks and high costs. Therefore, in this study, we intend to actively utilize human musculoskeletal simulation. First, to improve the accuracy of musculoskeletal models, we propose a body segment mass distribution algorithm using body composition analysis data that reflects body characteristics. Secondly, the efficiency of most exoskeleton torque control algorithms is evaluated as the reduction rate of Metabolic Cost. In this study, we assume that the torque minimizing the Metabolic Cost is the optimal torque and propose a method for obtaining the torque.

An unexpected occlusion in a real life, not in a laboratory, can be more fatal to neural networks than expected. In addition, it is virtually impossible to create a network that learns all the environmental changes as well as occlusions. Therefore, we propose an alternative approach in which the architecture and number of parameters remain unchanged while adapting to occlusion circumstances. Learning method with the term Conversion Learning classifies them more robustly by converting the weights from various occlusion situations. The experiments on MNIST dataset showed a 3.07 [%p] performance improvement over the baseline CNN model in a situation where most objects are occluded and unknowing what occlusion will appear in advance. The experimental results suggest that Conversion Learning is an efficient method to respond to environmental changes such as occluded images.

Recently, various input shapers have been introduced to reduce residual vibrations of flexible robots. However, there have been no studies on the design of an input shaper that can suppress residual vibration at an arbitrarily specified duration. In this paper, a novel input shaper called an SD (specified-duration) shaper is proposed for an undamped or underdamped system, which can suppress residual vibration at an arbitrarily given specified duration. If the specified duration is larger than a half period, a positive SD shaper composed of all positive impulses is designed, and if the specified duration is smaller than a half period, a negative SD shaper including a negative impulse is designed. As the specified duration is increased every half period after a half period, the number of impulses of the positive SD shaper is increased one by one, and the robustness of the SD shaper to modeling errors is increased. The performance of the SD shaper is analyzed through simulation studies for an undamped and underdamped second-order systems. The validity of the SD shapers is demonstrated experimentally using an experimental device that can generate container transport motions.