Optimization of silicon solar cell fabrication based on neural network and genetic programming modeling

This study describes techniques for the cascade modeling and the optimization that are required to conduct the simulator-based process optimization of solar cell fabrication. Two modeling approaches, neural networks and genetic programming, are employed to model the crucial relation for the consecutively connected two processes in solar cell fabrication. One model (Model 1) is used to map the five inputs (time, amount of nitrogen and DI water in surface texturing and temperature and time in emitter diffusion) to the two outputs (reflectance and sheet resistance) of the first process. The other model (Model 2) is used to connect the two inputs (reflectance and sheet resistance) to the one output (efficiency) of the second process. After modeling of the two processes, genetic algorithms and particle swarm optimization were applied to search for the optimal recipe. In the first optimization stage, we searched for the optimal reflectance and sheet resistance that can provide the best efficiency in the fabrication process. The optimized reflectance and sheet resistance found by the particle swarm optimization were better than those found by the genetic algorithm. In the second optimization stage, the five input parameters were searched by using the reflectance and sheet resistance values obtained in the first stage. The found five variables such as the texturing time, amount of nitrogen, DI water, diffusion time, and temperature are used as a recipe for the solar cell fabrication. The amount of nitrogen, DI water, and diffusion time in the optimized recipes showed considerable differences according to the modeling approaches. More importantly, repeated applications of particle swarm optimization yielded process conditions with smaller variations, implying greater consistency in recipe generation.     

Motion teaching method for complex robot links using motor current

Conventional robot motion teaching methods use a teaching pendant or a motion capture device and are not the most convenient or intuitive ways to teach a robot sophisticated and fluid movements such as martial arts motions. Ideally, a robot could be set up as if it were a clothing mannequin that has light limbs and flexible yet frictional joints which can be positioned at desirable shape and hold all the positions. To do the same with a robot, an operator could pull or push the links with minor forces until the desired robot posture is attained. For this, a robot should measure the applied external force by using torque sensors at the robot joints. However, torque sensors are bulky and expensive to install in every DOF joints while keeping a compact design, which is essential to humanoid robots. In this paper, we use only motor current readings to acquire joint torques. The equations used to compensate for the effect of gravity on the joint torques and the self-calibration method to earn link parameters are presented. Additionally, kinematic restrictions can be imposed on the robot’s arms to simplify the motion teaching. Here, we teach the Kendo training robot with this method and the robot’s learnt martial art motions are demonstrated.     

Trajectory generation and control for a biped robot walking upstairs

This paper presents an effective and systematic trajectory generation method, together with a control method for enabling a biped robot to walk upstairs. The COG (center of gravity) trajectory is generated by the VHIPM (virtual height inverted pendulum mode) for the horizontal motion and by a 6th order polynomial for the vertical motion; an ankle compliance control (ACC) is also added into the robot control. The proposed methods are evaluated by simulations as well as being implemented in a robot for the performance verification. The results show that the proposed methods can generate stable motions when walking upstairs, and these can significantly reduce the zero moment point (ZMP) errors compared with other methods, enabling the robot to walk up steeper stairs.     

Performance comparison of tracking filters in terms of predicted hitting point

This article describes the performance comparison of several tracking filters from the viewpoint of gun fire control. It is common that filter performance can be represented by the RMSE of the current state. However, for the design of a gun fire control system, the performance of the tracking filter in terms of prediction is more important since the current state proceeds with a designed dynamic model to a certain future which is used for the final gun order. It is shown that the predicted results amplify the current state RMSE and can be a good measure.     

Vehicle license plate tilt correction based on the straight line fitting method and minimizing variance of coordinates of projection points

Tilt correction is a very crucial and inevitable task in the automatic recognition of the vehicle license plate (VLP). In this paper, according to the least square fitting with perpendicular offsets (LSFPO), the VLP region is fitted to a straight line. After the line slope is obtained, rotation angle of the VLP is estimated. Then the whole image is rotated for tilt correction in horizontal direction by this angle. Tilt correction in vertical direction by minimizing the variance of coordinates of the projection points is proposed. Character segmentation is performed after horizontal correction and character points are projected along the vertical direction after shear transform. Despite the success of VLP detection approaches in the past decades, a few of them can effectively locate license plate (LP), even when vehicle bodies and LPs have similar color. A common drawback of color-based VLP detection is the failure to detect the boundaries or border of LPs. In this paper, we propose a modified recursive labeling algorithm for solving this problem and detecting candidate regions. According to different colored LP, these candidate regions may include LP regions. Geometrical properties of the LP such as area, bounding box and aspect-ratio are then used for classification. Various LP images were used with a variety of conditions to test the proposed method and results are presented to prove its effectiveness.     

Non-symmetric unscented transformation with application to in-flight alignment

To overcome the extended Kalman Filter (EKF) limitation, the unscented transformation (UT) was developed as a method to propagate mean and covariance information through nonlinear transformation. However, the ordinary UT has both local and global sampling problems. In this paper, non-symmetric unscented transformation (NSUT) is proposed to solve these problems. The sigma point sets of the NSUT are non-symmetrically located with respect to the zeroth sigma point set which is equal to the prior mean. Then, the weights of the sigma points are derived by using some conditions for mean and covariance. The non-symmetric unscented Kalman filter (NSUKF) based on the sigma points and weights of the NSUT are applied to the in-flight alignment (IFA) with a large heading error. With a low cost MEMS IMU, the simulations are carried out at an S-type trajectory to analyze the performance of the proposed method.     

Fire detection based on hidden Markov models

In this paper, a novel method of real-time fire detection based on HMMs is presented. First, we present an analysis of fire characteristics that provides evidence supporting the use of HMMs to detect fire; second, we propose an algorithm for detecting candidate fire pixels that entails the detection of moving pixels, fire-color inspection, and pixels clustering. The main contribution of this paper is the establishment and application of a hidden Markov fire model by combining the state transition between fire and non-fire with fire motion information to reduce data redundancy. The final decision is based on this model which includes training and application; the training provides parameters for the HMM application. The experimental results show that the method provides both a high detection rate and a low false alarm rate. Furthermore, real-time detection has been effectively realized via the learned parameters of the HMM, since the most time-consuming components such as HMM training are performed off-line.     

Modeling and simulation of powered hip orthosis by pneumatic actuators

In this study, a simulation model for a powered hip orthosis (PHO) with air muscles to predict the gait of paraplegics is presented which can be used as a design tool for hip orthoses. Before simulation, mathematical models for a human dummy with an orthosis and a pneumatic muscle actuator were generated. For the air muscle, coefficients required were obtained by static and dynamic experiments of the air muscle and experiments for the valve controlling the air pressure. The computation was conducted on the ADAMS package together with MATLAB. Computer simulation of the flexion of hip joints by the pneumatic muscle results in similar values to those from gait analysis. With the development of a simulation model for a PHO, the gait simulation model using pneumatic muscles can be used to analyze and evaluate the characteristics and efficiency of a PHO by setting the input and boundary conditions.     

Adaptive product tracking in RFID-enabled large-scale supply chain

The Radio Frequency Identification (RFID) technology is gradually being adopted and deployed for product flow management in the supply chain. In order to track RFID-tagged products efficiently in the RFID-enabled, large-scale supply chain, this paper first presents the design of a product tracking system that can collaborate with the EPC Network, a suite of network services for RFID data management in the supply chain. Next, we explain a product monitoring procedure that is performed by comparing the actual path of a product with its planned path. Finally, we develop an adaptive product search algorithm based on a reinforcement learning technique to efficiently locate a product deviated from its planned path. Experiment results are provided to show the performance of the search algorithm.     

Self-adaptive and dynamic clustering for online anomaly detection

As recent Internet threats are evolving more rapidly than ever before, one of the major challenges in designing an intrusion detection system is to provide early and accurate detection of emerging threats. In this study, a novel framework is developed for fully unsupervised training and online anomaly detection. The framework is designed so that an initial model is constructed and then it gradually evolves according to the current state of online data without any human intervention. In the framework, a self-organizing map (SOM) that is seamlessly combined with K-means clustering is transformed into an adaptive and dynamic algorithm suitable for real-time processing. The performance of the proposed approach is evaluated through experiments using the well-known KDD Cup 1999 data set and further experiments using the honeypot data recently collected from Kyoto University. It is shown that the proposed approach can significantly increase the detection rate while the false alarm rate remains low. In particular, it is capable of detecting new types of attacks at the earliest possible time.