• Journal of the Korea Institute of Military Science and Technology
• /
• v.27 no.4
• /
• pp.474-484
• /
• 2024

Reinforcement learning, which are also studied in the field of defense, face the problem of sample efficiency, which requires a large amount of data to train. Transfer learning has been introduced to address this problem, but its effectiveness is sometimes marginal because the model does not effectively leverage prior knowledge. In this study, we propose a stochastic initial state randomization(SISR) method to enable robust knowledge transfer that promote generalized and sufficient knowledge transfer. We developed a simulation environment involving a cooperative robot transportation task. Experimental results show that successful tasks are achieved when SISR is applied, while tasks fail when SISR is not applied. We also analyzed how the amount of state information collected by the agents changes with the application of SISR.

• 제어로봇시스템학회:학술대회논문집
• /
• 1995.10a
• /
• pp.360-363
• /
• 1995

In this paper, we describe the composition of frequency response bands based on experimental data of plants (controlled systems) with uncertainty and nonlinearity, and the robust stability evaluation of feedback control systems. Analysis and design of control systems using the upper and lower bounds of such experimental data would be effective as a practicable method which is not heavily dependent upon mathematical models such as the transfer function. First, we present a method to composite gain characteristic bands of frequency response of cascade connected plants with uncertainty and a recurrent inequality for the composition. Next, evaluation methods of the robust stability of multi-loop control systems obtained through feedback from the output terminals and multi-loop control systems obtained through feedback into the input terminals are described. In actual control systems, experimental data of frequency responses often depends on the amplitude of input. Therefore, we present the evaluation method of the nominal value and the width of the frequency response band in such a case, and finally give numerical examples based on virtual experimental data.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.9
• /
• pp.3870-3884
• /
• 2020

The proposed framework of Four Layer Robust Storage in Cloud (FLRSC) architecture involves host server, local host and edge devices in addition to Virtual Machine Monitoring (VMM). The goal is to protect the privacy of stored data at edge devices. The computational intelligence (CI) part of our algorithm distributes blocks of data to three different layers by partially encoded and forwarded for decoding to the next layer using hash and greed Solomon algorithms. VMM monitoring uses snapshot algorithm to detect intrusion. The proposed system is compared with Tiang Wang method to validate efficiency of data transfer with security. Hence, security is proven against the indexed efficiency. It is an important study to integrate communication between local host software and nearer edge devices through different channels by verifying snapshot using lamport mechanism to ensure integrity and security at software level thereby reducing the latency. It also provides thorough knowledge and understanding about data communication at software level with VMM. The performance evaluation and feasibility study of security in FLRSC against three-layered approach is proven over 2³² blocks of data with 98% accuracy. Practical implications and contributions to the growing knowledge base are highlighted along with directions for further research.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.5
• /
• pp.17-28
• /
• 2023

Potholes that occur on paved roads can have fatal consequences for vehicles traveling at high speeds and may even lead to fatalities. While manual detection of potholes using human labor is commonly used to prevent pothole-related accidents, it is economically and temporally inefficient due to the exposure of workers on the road and the difficulty in predicting potholes in certain categories. Therefore, completely preventing potholes is nearly impossible, and even preventing their formation is limited due to the influence of ground conditions closely related to road environments. Additionally, labeling work guided by experts is required for dataset construction. Thus, in this paper, we utilized the Mean Teacher technique, one of the semi-supervised learning-based knowledge distillation methods, to achieve robust performance in pothole image classification even with limited labeled data. We demonstrated this using performance metrics and GradCAM, showing that when using semi-supervised learning, 15 pre-trained CNN models achieved an average accuracy of 90.41%, with a minimum of 2% and a maximum of 9% performance difference compared to supervised learning.

• Transactions on Control, Automation and Systems Engineering
• /
• v.2 no.3
• /
• pp.149-156
• /
• 2000

Time delay observer (TDO), thanks to the time delay control (TDC) concept, requires little knowledge of a plant model, and hence is easy to design, robust to parameter variation and computationally efficient, yet can reconstruct states rather reliable for nonlinear plant. In this paper, we propose an improved version of TDO that solves two problems inherent in TDO as follows: TDO displays large reconstruction errors due to low-frequency uncertainty and has some restrictions on selecting its gains. By introducing a low pass filter and a state associated with it, we obtain an enhanced time delay observer (ETDO). This observer turns out to have smaller reconstruction errors than those of TDO and not to have any restriction on selecting its gains, thereby solving the problems. Through performance comparison by transfer function and simulation, we validate the analysis results of two observers (TDO and ETDO) and evaluate the performances. Finally, through experiments on BLDC motor system, the analysis results are clearly conformed.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.8 no.2
• /
• pp.483-503
• /
• 2014

In this paper, human action recognition using pyramid histograms of oriented gradients and collaborative multi-task learning is proposed. First, we accumulate global activities and construct motion history image (MHI) for both RGB and depth channels respectively to encode the dynamics of one action in different modalities, and then different action descriptors are extracted from depth and RGB MHI to represent global textual and structural characteristics of these actions. Specially, average value in hierarchical block, GIST and pyramid histograms of oriented gradients descriptors are employed to represent human motion. To demonstrate the superiority of the proposed method, we evaluate them by KNN, SVM with linear and RBF kernels, SRC and CRC models on DHA dataset, the well-known dataset for human action recognition. Large scale experimental results show our descriptors are robust, stable and efficient, and outperform the state-of-the-art methods. In addition, we investigate the performance of our descriptors further by combining these descriptors on DHA dataset, and observe that the performances of combined descriptors are much better than just using only sole descriptor. With multimodal features, we also propose a collaborative multi-task learning method for model learning and inference based on transfer learning theory. The main contributions lie in four aspects: 1) the proposed encoding the scheme can filter the stationary part of human body and reduce noise interference; 2) different kind of features and models are assessed, and the neighbor gradients information and pyramid layers are very helpful for representing these actions; 3) The proposed model can fuse the features from different modalities regardless of the sensor types, the ranges of the value, and the dimensions of different features; 4) The latent common knowledge among different modalities can be discovered by transfer learning to boost the performance.

• Journal of Microbiology and Biotechnology
• /
• v.25 no.8
• /
• pp.1339-1348
• /
• 2015

Until now, considerable effort has been made to engineer novel nitrogen-fixing organisms through the transfer of nif genes from various diazotrophs to non-nitrogen fixers; however, regulatory coupling of the heterologous nif genes with the regulatory system of the new host is still not well understood. In this work, a 49 kb nitrogen fixation island from P. stutzeri A1501 was transferred into E. coli using a novel and efficient transformation strategy, and a series of recombinant nitrogen-fixing E. coli strains were obtained. We found that the nitrogenase activity of the recombinant E. coli strain EN-01, similar to the parent strain P. stutzeri A1501, was dependent on external ammonia concentration, oxygen tension, and temperature. We further found that there existed a regulatory coupling between the E. coli general nitrogen regulatory system and the heterologous P. stutzeri nif island in the recombinant E. coli strain. We also provided evidence that the E. coli general nitrogen regulator GlnG protein was involved in the activation of the nif-specific regulator NifA via a direct interaction with the NifA promoter. To the best of our knowledge, this work plays a groundbreaking role in increasing understanding of the regulatory coupling of the heterologous nitrogen fixation system with the regulatory system of the recipient host. Furthermore, it will shed light on the structure and functional integrity of the nif island and will be useful for the construction of novel and more robust nitrogen-fixing organisms through biosynthetic engineering.