범용 가상환경 프레임워크 NAVER를 제안하고, 이를 케이브기반 가상현실환경에 적용하여 자동차 시제품 평가 실험에 활용한 사례를 소개한다. NAVER는 다양한 가상현실 어플리케이션을 구현하기 위한 가상환경 프레임워크로, 확장성이 뛰어나고 재구성이 가능하다 NAVER는 Render Server, Control Server, 그리고 Device Server로 구성되어 있으며, 각 서버는 네트워크로 상호 통신하여 각각의 기능을 수행한다. NAVER는 XML 기반 스크립팅 언어를 지원하여 사용자가 자유롭게 가상환경의 여러 가지 객체와 인터랙션을 정의할 수 있도록 설계되었다. NAVER를 케이브 기반 가상현실환경에 적용하여 자동자 시제품평가 실험에 활용하였다. KIST의 케이브 기반 가상현실 환경은 4면의 정방형 스테레오 디스플레이 장치, 햅틱 암마스터 장비, 3차원 음향장비 등으로 구성되어 있어, 사용자에서 시각적인 측면에서 뿐만 아니라 촉각, 청각과 같은 여러 가지 측면에서 다중현실감을 제시할 수 있다. 자동차 시제품 평가 실험을 통하여 사용자가 실제 자동차가 아닌 가상의 자동차 시제품을 관찰하고, 만져보고, 주행해 봄으로써 더욱 높은 몰입감과 현실감으로 자동차 조작장치의 조작성을 평가할 수 있음을 입증하였다.

Multi-floor navigation of a mobile robot requires a technology that allows the robot to safely get on and off the elevator. Therefore, in this study, we propose a method of recognizing the elevator from the current position of the robot and estimating the location of the elevator locally so that the robot can safely get on the elevator regardless of the accumulated position error during autonomous navigation. The proposed method uses a deep learning-based image classifier to identify the elevator from the image information obtained from the RGB-D sensor and extract the boundary points between the elevator and the surrounding wall from the point cloud. This enables the robot to estimate the reliable position in real time and boarding direction for general elevators. Various experiments exhibit the effectiveness and accuracy of the proposed method.

Getting on and off an elevator is one of the most important parts for multi-floor navigation of a mobile robot. In this study, we proposed the method for the pose recognition of elevator doors, safe path planning, and motion estimation of a robot using RGB-D sensors in order to safely get on and off the elevator. The accurate pose of the elevator doors is recognized using a particle filter algorithm. After the elevator door is open, the robot builds an occupancy grid map including the internal environments of the elevator to generate a safe path. The safe path prevents collision with obstacles in the elevator. While the robot gets on and off the elevator, the robot uses the optical flow algorithm of the floor image to detect the state that the robot cannot move due to an elevator door sill. The experimental results in various experiments show that the proposed method enables the robot to get on and off the elevator safely.

Home robot arms require a payload of 2 kg to perform various household tasks; at the same time, they should be operated by low-capacity motors and low-cost speed reducers to ensure reasonable product cost. Furthermore, as robot arms on mobile platforms are battery-driven, their energy efficiency should be very high. To satisfy these requirements, we designed a lightweight counterbalance mechanism (CBM) based on a spring and a wire and developed a home robot arm with five degrees of freedom (DOF) based on this CBM. The CBM compensates for gravitational torques applied to the two pitch joints that are most affected by the robot’s weight. The developed counterbalance robot adopts a belt-pulley based parallelogram mechanism for 2-DOF gravity compensation. Experiments using this robot demonstrate that the CBM allows the robot to meet the above-mentioned requirements, even with low-capacity motors and speed reducers.

Direct teaching is an essential function for collaborative robots for easy use by non-experts. For most robots, direct teaching is implemented only in joint space because the realization of Cartesian space direct teaching, in which the orientation of the end-effector is fixed while teaching, requires a measurement of the end-effector force. Thus, it is limited to the robots that are equipped with an expensive force/torque sensor. This study presents a Cartesian space direct teaching method for torque-controlled collaborative robots without either a force/torque sensor or joint torque sensors. The force exerted to the end-effector is obtained from the external torque which is estimated by the disturbance observer-based approach with the friction model. The friction model and the estimated end-effector force were experimentally verified using the robot equipped with joint torque sensors in order to compare the proposed sensorless approach with the method using torque sensors.

A robot usually adopts ANN (artificial neural network)-based object detection and instance segmentation algorithms to recognize objects but creating datasets for these algorithms requires high labeling costs because the dataset should be manually labeled. In order to lower the labeling cost, a new scheme is proposed that can automatically generate a training images and label them for specific objects. This scheme uses an instance segmentation algorithm trained to give the masks of unknown objects, so that they can be obtained in a simple environment. The RGB images of objects can be obtained by using these masks, and it is necessary to label the classes of objects through a human supervision. After obtaining object images, they are synthesized with various background images to create new images. Labeling the synthesized images is performed automatically using the masks and previously input object classes. In addition, human intervention is further reduced by using the robot arm to collect object images. The experiments show that the performance of instance segmentation trained through the proposed method is equivalent to that of the real dataset and that the time required to generate the dataset can be significantly reduced.

Obstacle avoidance is one of the most important parts of autonomous mobile robot. In this study, we proposed safe and efficient local path planning of robot for obstacle avoidance. The proposed method detects and tracks obstacles using the 3D depth information of an RGB-D sensor for path prediction. Based on the tracked information of obstacles, the paths of the obstacles are predicted with probability circle-based spatial search (PCSS) method and Gaussian modeling is performed to reduce uncertainty and to create the cost function of caution. The possibility of collision with the robot is considered through the predicted path of the obstacles, and a local path is generated. This enables safe and efficient navigation of the robot. The results in various experiments show that the proposed method enables robots to navigate safely and effectively.

Recent studies on automatic parking have actively adopted the technology developed for mobile robots. Among them, the path planning scheme plans a route for a vehicle to reach a target parking position while satisfying the kinematic constraints of the vehicle. However, previous methods require a large amount of computation and/or cannot be easily applied to different environmental conditions. Therefore, there is a need for a path planning scheme that is fast, efficient, and versatile. In this study, we use a multi-dimensional path grid map to solve the above problem. This multi-dimensional path grid map contains a route which has taken a vehicle's kinematic constraints into account; it can be used with the A* algorithm to plan an efficient path. The proposed method was verified using Prescan which is a simulation program based on MATLAB. It is shown that the proposed scheme can successfully be applied to both parallel and vertical parking in an efficient manner.

Static balance of an articulated robot arm at various configurations requires a torque compensating for the gravitational torque of each joint due to the robot mass. Such compensation torque can be provided by a spring-based counterbalance mechanism. However, simple installation of a counterbalance mechanism at each pitch joint does not work because the gravitational torque at each joint is dependent on other joints. In this paper, a 6 DOF industrial robot arm based on the parallelogram for multi-DOF counterbalancing is proposed to cope with this problem. Two passive counterbalance mechanisms are applied to pitch joints, which reduces the required torque at each joint by compensating the gravitational torque. The performance of this mechanism is evaluated experimentally.

A robot manipulator handling a heavy weight requires high-capacity motors and speed reducers, which increases the cost of a robot and the risk of injury when a human worker is in collaboration with a robot. To cope with this problem, we propose a collaborative manipulator equipped with a counterbalance mechanism which compensates mechanically for a gravitational torque due to the robot mass. The prototype of the manipulator was designed on the basis of a four-bar linkage structure which contains active and passive pitch joints. Experimental performance evaluation shows that the proposed robot works effectively as a collaborative robot.

Conventional path tracking methods designed for two-wheeled differential drive robots are not suitable for omni-directional robots. In this study, we present a controller which can accomplish more accurate path tracking and orientation correction by exploiting the unconstrained movement capability of omni-directional robots. The proposed controller is proven to be stable using a Lyapunov stability criterion. Various experiments in real environments show that performance of path tracking and orientation correction has improved in the proposed controller.

This paper proposes a pose-graph based SLAM method using an upward-looking camera and artificial landmarks for AGVs in factory environments. The proposed method provides a way to acquire the camera extrinsic matrix and improves the accuracy of feature observation using a low-costcamera. SLAM is conducted by optimizing AGV’s explored path using the artificial landmarks installed on the ceiling at various locations. As the AGV explores, the pose nodes are added based on the certain distance from odometry and the landmark nodes are registered when AGV recognizes the fiducial marks. As a result of the proposed scheme, a graph network is created and optimized through a G2O optimization tool so that the accumulated error due to the slip is minimized. The experiment shows that the proposed method is robust for SLAM in real factory environments.

An adaptive robot hand (AR-Hand) has a stable grasp of different objects in unstructured environments. In this study, we propose an AR-Hand based on a tendon-driven mechanism which consists of 4 fingers and 12 DOFs. It weighs 0.5 kg and can grasp an object up to 1 kg. This hand based on the adaptive grasp mechanism is able to provide a stable grasp without a complex control algorithm or sensor system. The fingers are driven by simple tendon structures with each finger capable of adaptively grasping the objects. This paper presents a method to decide the joint stiffness. The adaptive grasping is verified by various grasping experiments involving objects with different shapes and sizes.

This paper proposes a novel upward-looking camera-based global localization using a ceiling image map. The ceiling images obtained through the SLAM process are integrated into the ceiling image map using a particle filter. Global localization is performed by matching the ceiling image map with the current ceiling image using SURF keypoint correspondences. The robot pose is then estimated by the coordinate transformation from the ceiling image map to the global coordinate system. A series of experiments show that the proposed method is robust in real environments.

Localization is one of the essential tasks necessary to achieve autonomous navigation of a mobile robot. One such localization technique, Monte Carlo Localization (MCL) is often applied to a digital surface model. However, there are differences between range data from laser rangefinders and the data predicted using a map. In this study, commonly observed from air and ground (COAG) features and candidate selection based on the shape of sensor data are incorporated to improve localization accuracy. COAG features are used to classify points consistent with both the range sensor data and the predicted data, and the sample candidates are classified according to their shape constructed from sensor data. Comparisons of local tracking and global localization accuracy show the improved accuracy of the proposed method over conventional methods.

Human-robot co-operation becomes increasingly frequent due to the widespread use of service robots. However, during such co-operation, robots have a high chance of colliding with humans, which may result in serious injury. Thus, many solutions were proposed to ensure collision safety, and among them, collision detection algorithms are regarded as one of the most practical solutions. They allow a robot to quickly detect a collision so that the robot can perform a proper reaction to minimize the impact. However, conventional collision detection algorithms required the precise model of a robot, which is difficult to obtain and is subjected to change. Also, expensive sensors, such as torque sensors, are often required. In this study, we propose a novel collision detection algorithm which only requires motor encoders. It detects collisions by monitoring the high-pass filtered version of the velocity error. The proposed algorithm can be easily implemented to any robots, and its performance was verified through various tests.

Global localization is one of the essential issues for mobile robot navigation. In this study, an indoor global localization method is proposed which uses a Kinect sensor and a monocular upward-looking camera. The proposed method generates an environment map which consists of a grid map, a ceiling feature map from the upward-looking camera, and a spatial feature map obtained from the Kinect sensor. The method selects robot pose candidates using the spatial feature map and updates sample poses by particle filter based on the grid map. Localization success is determined by calculating the matching error from the ceiling feature map. In various experiments, the proposed method achieved a position accuracy of 0.12m and a position update speed of 10.4s, which is robust enough for real-world applications.

This paper describes hierarchical semantic map building using the classified area information in home environments. The hierarchical semantic map consists of a grid, CAIG (Classified Area Information in Grid), and topological map. The grid and CAIG maps are used for navigation and motion selection, respectively. The topological map provides the intuitive information on the environment, which can be used for the communication between robots and users. The proposed semantic map building algorithm can greatly improve the capabilities of a mobile robot in various domains, including localization, path-planning and HRI (Human-Robot Interaction). In the home environment, a door can be used to divide an area into various sections, such as a room, a kitchen, and so on. Therefore, we used not only the grid map of the home environment, but also the door information as a main clue to classify the area and to build the hierarchical semantic map. The proposed method was verified through various experiments and it was found that the algorithm guarantees autonomous map building in the home environment.

Global positioning system (GPS) is widely used to measure the position of a vehicle. However, the accuracy of the GPS can be severely affected by surrounding environmental conditions. To deal with this problem, the GPS and odometry data can be combined using an extended Kalman filter. For stable navigation of an outdoor mobile robot using the GPS, this paper proposes two methods to evaluate the reliability of the GPS data. The first method is to calculate the standard deviation of the GPS data and reflect it to deal with the uncertainty of the GPS data. The second method is to match the GPS data to the traversability map which can be obtained by classifying outdoor terrain data. By matching of the GPS data with the traversability map, we can determine whether to use the GPS data or not. The experimental results show that the proposed methods can enhance the performance of the GPS‐based outdoor localization.

It is very important for a mobile robot to recognize and model its environments for navigation. However, the grid map constructed by sonar sensors cannot accurately represent the environment, especially the narrow environment, due to the angular uncertainty of sonar data. Therefore, we propose a map building scheme which combines sonar sensors and IR sensors. The maps built by sonar sensors and IR sensors are combined with different weights which are determined by the degree of translational and rotational motion of a robot. To increase the effectiveness of sensor fusion, we also propose optimal sensor arrangement through various experiments. The experimental results show that the proposed method can represent the environment such as narrow corridor and open door more accurately than conventional sonar sensor-based map building methods.