• International Journal of Control, Automation, and Systems
• /
• v.1 no.3
• /
• pp.339-350
• /
• 2003

In this paper, we propose a simultaneous multithreading (SMT) architecture that improves instruction throughput by exploiting instruction level parallelism (ILP) and thread level parallelism (TLP). The proposed architecture issues and completes instructions belonging to the same thread in exact program order. The issue and completion policy greatly reduces the design complexity and hardware cost of our architecture, compared with others that employ out-of-order issue and completion. On the other hand, when the instructions belong to different threads, the issue and completion orders for those instructions may not necessarily be identical to the fetch order. The processor issues instructions simultaneously from multiple threads to functional units by exploiting ILP and TLP, and by dynamic resource sharing. That parallel execution notably improves performance and resource utilization with minimal additional hardware cost over the conventional superscalar processors. This paper proposes an SMT architecture with grouping as well as one without grouping. Without grouping, all threads dynamically and flexibly share most resources. On the other hand, in the SMT architecture with grouping, in which resources and threads are divided into several groups for design simplification, resources are shared only among threads belonging to the same group as those resources. Simulation results show that our processors with four and eight threads improve performance by three or more times over the conventional superscalar processor with comparable execution resources and policies, and that reasonable grouping reduces the design complexity of SMT processors with little negative effect on performance.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.1859-1864
• /
• 2004

An effective methodology is reported for the optimal design of multisite batch production/transportation and storage networks under uncertain demand forecasting. We assume that any given storage unit can store one material type which can be purchased from suppliers, internally produced, internally consumed, transported to or from other plant sites and/or sold to customers. We further assume that a storage unit is connected to all processing and transportation stages that consume/produce or move the material to which that storage unit is dedicated. Each processing stage transforms a set of feedstock materials or intermediates into a set of products with constant conversion factors. A batch transportation process can transfer one material or multiple materials at once between plant sites. The objective for optimization is to minimize the probability averaged total cost composed of raw material procurement, processing setup, transportation setup and inventory holding costs as well as the capital costs of processing stages and storage units. A novel production and inventory analysis formulation, the PSW(Periodic Square Wave) model, provides useful expressions for the upper/lower bounds and average level of the storage inventory. The expressions for the Kuhn-Tucker conditions of the optimization problem can be reduced to two sub-problems. The first yields analytical solutions for determining lot sizes while the second is a separable concave minimization network flow subproblem whose solution yields the average material flow rates through the networks for the given demand forecast scenario. The result of this study will contribute to the optimal design and operation of large-scale supply chain system.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.2087-2089
• /
• 2003

The presence of narrowband interference (NBI) in Direct-sequence code division multiple access (DS/CDMA) systems is an inevitable problem when the interference is strong enough. The improvement in the system performance employs by adaptive narrowband interference suppression techniques. Basically there have been two types of method for narrowband interference suppression estimator/subtracter approaches and transform domain approaches. In this paper the focus is on the type of estimator/subtracter approaches. However, the binary direct sequence (DS) signal, that acts as noise in the prediction process is highly non-Gaussian. The case of a Gaussian interferer with known in an autoregressive (AR) signal or a digital signal and also in a sinusoidal signal (Tone) that included in is paper. The proposed NBI suppression is presence in an adaptive IIR notch filter for lattice structure and more powerful by using a variable step-size algorithm. The simulation results show that the proposed algorithm can significantly increase the convergence rate and improved system performance when compare with adaptive least mean square algorithm (LMS).

• The Korean Journal of Applied Statistics
• /
• v.24 no.4
• /
• pp.695-708
• /
• 2011

Image fusion is the process of combining multiple images of the same scene into a single fused image with application to many fields, such as remote sensing, computer vision, robotics, medical imaging and military affairs. The widely used image fusion rules that use wavelet transform have been based on a simple comparison with the activity measures of local windows such as mean and standard deviation. In this case, information features from the original images are excluded in the fusion image and distorted fusion images are obtained for noisy images. In this paper, we propose the use of a nonparametric squared ranks test on the quality of variance for two samples in order to overcome the influence of the noise and guarantee the homogeneity of the fused image. We evaluate the method both quantitatively and qualitatively for image fusion as well as compare it to some existing fusion methods. Experimental results indicate that the proposed method is effective and provides satisfactory fusion results.

• Korean Journal of Applied Biomechanics
• /
• v.27 no.2
• /
• pp.141-147
• /
• 2017

Objective: We obtained force-displacement curves for countermovement jumps of multiple heights and examined the effect of an arm swing on changes in vertical jumping strategy. Countermovement jumps with hands on hips (Condition 1) and with an arm swing (Condition 2) were evaluated to investigate the mechanical effect of the arm movement on standing vertical jumps. We hypothesized that the ground reaction force (GRF) and/or center of mass (CoM) motion resulting from the countermovement action would significantly change depending on the use of an arm swing. Method: Eight healthy young subjects jumped straight up to five different levels ranging from approximately 10% (~25 cm) to 35% (~55 cm) of their body heights. Each subject performed five sets of jumps to five randomly ordered vertical elevations in each condition. For comparison of the two jumping strategies, the characteristics of the boundary point on the force-displacement curve, corresponding to the vertical GRF and the CoM displacement at the end of the countermovement action, were investigated to understand the role of arm movement. Results: Based on the comparison between the two conditions (with and without an arm swing), the subjects were grouped into type A and type B depending on the change observed in the boundary point across the five different jump heights. For both types (type A and type B) of vertical jumps, the initial vertical force at the start of push-off significantly changed when the subjects employed arm movement. Conclusion: The findings may imply that the jumping strategy does change with the inclusion of an arm swing, predominantly to modulate the vertical force advantage (i.e., the difference between the vertical force at the start of push-off and the body weight).

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.6
• /
• pp.842-849
• /
• 2022

With the development of deep learning technology, there are many cases of using DNNs in embedded systems such as unmanned vehicles, drones, and robotics. Typically, in the case of an autonomous driving system, it is crucial to run several DNNs which have high accuracy results and large computation amount at the same time. However, running multiple DNNs simultaneously in an embedded system with relatively low performance increases the time required for the inference. This phenomenon may cause a problem of performing an abnormal function because the operation according to the inference result is not performed in time. To solve this problem, the solution proposed in this paper first reduces the computation by applying the Tucker decomposition to DNN models with big computation amount, and then, make DNN models run in parallel as much as possible in the unit of hidden layer inside the GPU. The experimental result shows that the DNN inference time decreases by up to 75.6% compared to the case before applying the proposed technique.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.1
• /
• pp.177-186
• /
• 2017

Recently, Autonomous vehicles are getting hot attention. Amazon, the biggest online shopping service provider is developing a delivery system that uses drones. This kinds of platforms are need accurate attitude information for navigation. In this paper, a structure design of AHRS using low-cost inertia sensor is proposed. To estimate attitudes a Kalman filter which uses a quaternion based dynamic model, bias-removed measurements from MEMS Gyro, raw measurements from MEMS accelerometer and magnetometer, is designed. To remove bias from MEMS Gyro, an additional Kalman filter which uses raw Gyro measurements and attitude estimates, is designed. The performance of implemented AHRS is compared with high price off-the-shelf 3DM-GX3-25 AHRS from Microstrain. The Gyro bias was estimated within 0.0001[deg/s]. And from the estimated attitude, roll and pitch angle error is smaller than 0.2 and 0.3 degree. Yaw angle error is smaller than 6 degree.

• Journal of Sensor Science and Technology
• /
• v.31 no.3
• /
• pp.156-162
• /
• 2022

This study estimates the relative position between body segments using segment orientation and segment-to-joint center (S2J) vectors. In many wearable motion tracking technologies, the S2J vector is treated as a constant based on the assumption that rigid body segments are connected by a mechanical ball joint. However, human body segments are deformable non-rigid bodies, and they are connected via ligaments and tendons; therefore, the S2J vector should be determined as a time-varying vector, instead of a constant. In this regard, our previous study (2021) proposed a method for determining the time-varying S2J vector from the learning dataset using a regression method. Because that method uses a deformation-related variable to consider the deformation of S2J vectors, the optimal variable must be determined in terms of estimation accuracy by motion and segment. In this study, we investigated the effects of deformation-related variables on the estimation accuracy of the relative position. The experimental results showed that the estimation accuracy was the highest when the flexion and adduction angles of the shoulder and the flexion angles of the shoulder and elbow were selected as deformation-related variables for the sternum-to-upper arm and upper arm-to-forearm, respectively. Furthermore, the case with multiple deformation-related variables was superior by an average of 2.19 mm compared to the case with a single variable.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.138-142
• /
• 2004

In this paper, we present a new method for monitoring of ECU's sensor signals of vehicle. In order to measure the ECU's sensor signals, the interfaced circuit is designed to communicate ECU and the Embedded Linux is used to monitor communication result through Web the Embedded Linux system and this system is said "ECU Interface Part". In ECU Interface Part the interface circuit is designed to match voltage level between ECU and SA-1110 micro controller and interface circuit to communicate ECU according to the ISO, SAE communication protocol standard. Because Embedded Linux does not allow to access hardware directly in application level, anyone who wants to modify any low level hardware must develop device driver. To monitor ECU's sensor signals the most important thing is to match serial level between ECU and ECU Interface Part. It means to communicate correctly between two hardware we need to match voltage and signal level, and need to match baudrate. The voltage of SA-1110 is 0 ${\sim}$ +3.3V and ECU is 0 ${\sim}$ +12V and, ECU's communication Line K does multiple operation so, the interface circuit is used to match voltage and signal level. In Addition to ECU's baudrate is 10400bps, it's not standard baudrate in computer environment. So, we need to develop a device driver to control the interface circuit, and change baudrate. To monitor ECU's sensor signals through web there's a network socket program is working in Embedded Linux. It works as server program and manages user's connections and commands. Anyone who wants to monitor ECU's sensor signals he just only connect to Embedded Linux system with web browser then, Embedded Linux webserver will return the ActiveX webbased measurement software. It works in web browser and inits ECU, as a result it returns sensor signals through web. All the programs are developed with GCC(GNU C Compiler) and, webbased measurement software is developed with Borland C++ Builder.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.620-626
• /
• 2004

For general nonlinear processes, it is difficult to control with a linear model-based control method and nonlinear controls are considered. Among the numerous approaches suggested, the most rigorous approach is to use dynamic optimization. Many general engineering problems like control, scheduling, planning etc. are expressed by functional optimization problem and most of them can be changed into dynamic programming (DP) problems. However the DP problems are used in just few cases because as the size of the problem grows, the dynamic programming approach is suffered from the burden of calculation which is called as 'curse of dimensionality'. In order to avoid this problem, the Neuro-Dynamic Programming (NDP) approach is proposed by Bertsekas and Tsitsiklis (1996). To get the solution of seriously nonlinear process control, the interest in NDP approach is enlarged and NDP algorithm is applied to diverse areas such as retailing, finance, inventory management, communication networks, etc. and it has been extended to chemical engineering parts. In the NDP approach, we select the optimal control input policy to minimize the value of cost which is calculated by the sum of current stage cost and future stages cost starting from the next state. The cost value is related with a weight square sum of error and input movement. During the calculation of optimal input policy, if the approximate cost function by using simulation data is utilized with Bellman iteration, the burden of calculation can be relieved and the curse of dimensionality problem of DP can be overcome. It is very important issue how to construct the cost-to-go function which has a good approximate performance. The neural network is one of the eager learning methods and it works as a global approximator to cost-to-go function. In this algorithm, the training of neural network is important and difficult part, and it gives significant effect on the performance of control. To avoid the difficulty in neural network training, the lazy learning method like k-nearest neighbor method can be exploited. The training is unnecessary for this method but requires more computation time and greater data storage. The pH neutralization process has long been taken as a representative benchmark problem of nonlin ar chemical process control due to its nonlinearity and time-varying nature. In this study, the NDP algorithm was applied to pH neutralization process. At first, the pH neutralization process control to use NDP algorithm was performed through simulations with various approximators. The global and local approximators are used for NDP calculation. After that, the verification of NDP in real system was made by pH neutralization experiment. The control results by NDP algorithm was compared with those by the PI controller which is traditionally used, in both simulations and experiments. From the comparison of results, the control by NDP algorithm showed faster and better control performance than PI controller. In addition to that, the control by NDP algorithm showed the good results when it applied to the cases with disturbances and multiple set point changes.