In this paper, we consider a regulation problem of nonlinear systems under two triangular conditions where there possibly exist unbounded parameters in the systems. We propose a state feedback controller with dynamic gains in order to deal with unbounded parameters based on the condition of the time-varying rate of the growing parameter. The analysis of our control scheme is carried out by Lyapunov stability method. Our control method is verified by simulation results.

This paper presents a method to design a controller for active suspension with 1-DOF decoupled models. Three 1-DOF decoupled models describing vertical, roll and pitch motions are used to design a controller in order to generate a vertical force, roll and pitch moments, respectively. These control inputs are converted into active suspension forces with geometric relationship. To design a controller, a sliding mode control is adopted. Frequency domain analysis and simulation on vehicle simulation software, CarSim$^{(R)}$, show that the proposed method is effective for ride comfort.

This paper addresses the development of 3D data acquisition system (3D scanner) based laser structured-light image. The 3D scanner consists of a stripe laser generator, a conventional camera, and a rotation table. The stripe laser onto an object has distortion according to 3D shape of an object. By analyzing the distortion of the laser stripe in a camera image, the scanner obtains a group of 3D point data of the object. A simple semiconductor stripe laser diode is adopted instead of an expensive LCD projector for complex structured-light pattern. The camera has an optical filter to remove illumination noise and improve the performance of the distance measurement. Experimental results show the 3D data acquisition performance of the scanner with less than 0.2mm measurement error in 2 minutes. It is possible to reconstruct a 3D shape of an object and to reproduce the object by a commercially available 3D printer.

Motion planning problem is still one of the important issues in robotic applications. In many real-time motion planning problems, it is advisable to find a feasible solution quickly and improve the found solution toward the optimal one before the previously-arranged motion plan ends. For such reasons, sampling-based approaches are becoming popular for real-time application. Especially the use of a rapidly exploring random $tree^*$ ($RRT^*$) algorithm is attractive in real-time application, because it is possible to approach an optimal solution by iterating itself. This paper presents a modified version of informed $RRT^*$ which is an extended version of $RRT^*$ to increase the rate of convergence to optimal solution by improving the sampling method of $RRT^*$. In online motion planning, the robot plans a path while simultaneously moving along the planned path. Therefore, the part of the path near the robot is less likely to be sampled extensively. For a better solution in online motion planning, we modified the sampling method of informed $RRT^*$ by combining with the sampling method to improve the path nearby robot. With comparison among basic $RRT^*$, informed $RRT^*$ and the proposed $RRT^*$ in online motion planning, the proposed $RRT^*$ showed the best result by representing the closest solution to optimum.

A brain-computer interface or BCI provides an alternative method for acting on the world. Brain signals can be recorded from the electrical activity along the scalp using an electrode cap. By analyzing the EEG, it is possible to determine whether a person is thinking about his/her hand or foot movement and this information can be transferred to a machine and then translated into commands. However, we do not know which information relates to motor imagery and which channel is good for extracting features. A general approach is to use all electronic channels to analyze the EEG signals, but this causes many problems, such as overfitting and problems removing noisy and artificial signals. To overcome these problems, in this paper we used a new optimization method called the Fruit Fly optimization algorithm (FOA) to select the best channels and then combine them with CSP method to extract features to improve the classification accuracy by linear discriminant analysis. We also used particle swarm optimization (PSO) and a genetic algorithm (GA) to select the optimal EEG channel and compared the performance with that of the FOA algorithm. The results show that for some subjects, the FOA algorithm is a better method for selecting the optimal EEG channel in a short time.

In this paper, an unmanned delivery service using drone was proposed and verified the feasibility. The multicopter has GPS for autopilot and a camera for remote control by human operator. The gripper for manipulation of delivery object was designed and evaluated. The multicopter flies to a given position automatically based on GPS, and approaches to the prepared delivery desk by remote control of human operator using the received image from the multicopter. GPS sensor verification and experimental PID tuning were performed to ensure the flight stability. The flight tests were carried out to verify the feasibility of delivery service.

This paper presents a novel collision avoidance technique for mobile robots based on omni-directional vision simultaneous localization and mapping (SLAM). This method estimates the avoidance path and speed of a robot from the location of an obstacle, which can be detected using the Lucas-Kanade Optical Flow in images obtained through fish-eye cameras mounted on the robots. The conventional methods suggest avoidance paths by constructing an arbitrary force field around the obstacle found in the complete map obtained through the SLAM. Robots can also avoid obstacles by using the speed command based on the robot modeling and curved movement path of the robot. The recent research has been improved by optimizing the algorithm for the actual robot. However, research related to a robot using omni-directional vision SLAM to acquire around information at once has been comparatively less studied. The robot with the proposed algorithm avoids obstacles according to the estimated avoidance path based on the map obtained through an omni-directional vision SLAM using a fisheye image, and returns to the original path. In particular, it avoids the obstacles with various speed and direction using acceleration components based on motion information obtained by analyzing around the obstacles. The experimental results confirm the reliability of an avoidance algorithm through comparison between position obtained by the proposed algorithm and the real position collected while avoiding the obstacles.

In this paper, we propose an improved path planning method and obstacle avoidance algorithm for two-wheel mobile robots, which can be effectively applied in an environment where obstacles can be represented by circles. Firstly, we briefly introduce the rapidly exploring random tree (RRT) and single polar polynomial (SPP) algorithm. Secondly, we present additional two methods for applying our proposed method. Thirdly, we propose a global path planning, smoothing and obstacle avoidance method that combines the RRT and SPP algorithms. Finally, we present a simulation using our proposed method and check the feasibility. This shows that proposed method is better than existing methods in terms of the optimality of the trajectory and the satisfaction of the kinematic constraints.

This paper describes the design of a smart three-axis force sensor for measuring forces Fx, Fy and Fz. The smart three-axis force sensor is composed of a three-axis force sensor, a force-measuring device, housing and a cover, where the three-axis force sensor and the force-measuring device are inside the housing and the cover. The measuring device measures forces Fx, Fy and Fz from the three-axis force sensor, and calculates the resultant force using the measured forces, and then sends the resultant force and forces to a PC or other controller using RS-485 communication. The repeatability error and the non-linearity error of the smart three-axis force sensor are less than 0.03%, and the interference error of the sensor is less than 0.87%. It is thought that the sensor can be used for measuring forces in a robot, automatic systems and so on.

This paper describes a three-dimensional reference trajectory generation method for giving commands to an unmanned air vehicle (UAV). The trajectory is a set of consecutive curves with constant acceleration during each interval and passing through via-points at specified times or speeds. The functional inputs are three-dimensional positions and times (or speeds) at via-points, and velocities at both boundaries. Its output is the time series of position values satisfying the piecewise constant acceleration condition. To be specific, the shape of the trajectory, known as the path, is first represented by splines using third degree polynomials. A numeric algorithm is then suggested, which can overcome the demerits of cubic spline method and promptly generate a piecewise constant acceleration trajectory from the given path. To show the effectiveness of the present scheme, trajectory generation cases were treated, and their speed calculation errors were evaluated.

Nowadays many people have an interest in facial expression and the behavior of people. These are human-robot interaction (HRI) researchers utilize digital image processing, pattern recognition and machine learning for their studies. Facial feature point detector algorithms are very important for face recognition, gaze tracking, expression, and emotion recognition. In this paper, a cascade facial feature point detector is used for finding facial feature points such as the eyes, nose and mouth. However, the detector has difficulty extracting the feature points from several images, because images have different conditions such as size, color, brightness, etc. Therefore, in this paper, we propose an algorithm using a modified cascade facial feature point detector using a convolutional neural network. The structure of the convolution neural network is based on LeNet-5 of Yann LeCun. For input data of the convolutional neural network, outputs from a cascade facial feature point detector that have color and gray images were used. The images were resized to $32{\times}32$. In addition, the gray images were made into the YUV format. The gray and color images are the basis for the convolution neural network. Then, we classified about 1,200 testing images that show subjects. This research found that the proposed method is more accurate than a cascade facial feature point detector, because the algorithm provides modified results from the cascade facial feature point detector.