Joint Optimization of Time-Slot Allocation and Traffic Steering for Large-Scale Deterministic Networks

Wu, Wenhao,Zhang, Xiaoning,Pan, Jiaming,Zhou, Yihui 한국통신학회 2023 Journal of communications and networks Vol.25 No.6

Recently, time-sensitive services have expanded from traditional industrial control systems to more scenarios. Some time-sensitive applications, such as remote surgery, autonomous driving, Augmented Reality (AR), etc., require deterministic end-to-end delay and jitter in data transmission. Deterministic Network (DetNet) is proposed as a promising technology for providing deterministic service in Wide Area Networks (WAN). DetNet guarantees deterministic end-to-end delay and jitter by specifying a certain routing path and transmission time-slots for time-sensitive flows. However, how to efficiently steer time-sensitive flows while jointly allocating transmission time-slots is still an open problem. Existing flow scheduling algorithms are limited in the scenarios of Local Area Networks (LAN), and do not consider the impact of propagation delay in large-scale networks. To this end, we study the joint optimization of time-slot allocation and traffic steering, while considering the propagation delay of WAN links. Our objective is to maximize the number of successfully deployed time-sensitive flows under the constraints of required end-to-end delay. Accordingly, we formulate the studied problem as an Integer Linear Programming (ILP) model. Since it is proved to be an NP-hard problem, we design a heuristic algorithm named Genetic-based Deterministic Network Traffic Scheduling (GDNTS). The solution with the largest number of deployed time-sensitive flows can be obtained from the evolution of chromosomes in GDNTS. Compared with the benchmark algorithms, extensive simulation results show that GDNTS improves the deployed time sensitive flows number by 22.85% in average.

신경망의 결정론적 이완에 의한 자기공명영상 분류

전준철,민경필,권수일 대한자기공명의과학회 2002 Investigative Magnetic Resonance Imaging Vol.6 No.2

목적: 본 논문에서는 신경망을 이용한 자기공명영상의 분류에 있어 결정론적 이완 방법(deterministic relaxation)과 응집 군집화(agglomerative clustering) 방법에 의한 개선된 영상 분류방법을 제시한다. 제안된 방법은 신경망을 이용한 영상의 분류시 지역적 최소치로의 수렴문제와 입력 패턴의 증대로 인하여 수렴 속가 늦어지는 문제를 해결한다. 대상 및 방법: 신경망을 이용한 영상의 분류는 지역적 계산과 병렬 계산이 가능한 특성을 갖고 있어 기존의 통계적 방법을 대신하는 방법으로 주목을 받고 있다. 그러나 일반적으로 신경망에 의한 분류알고리즘이 지닌 문제점의 하나는 에너지함수가 항상 전역적 최소치로 수렴하지 않고 지역적 최소치로도 수렴할 수 있다는 점이고, 또 다른 문제점은 반복수렴을 수행하는 에너지함수의 수렴속도가 너무 늦다는 점이다. 따라서 지역적 최소치로의 수렴을 방지하고 전역적 최소치로의 수렴속도를 가속화시키기 위하여 본 논문에서는 결정적 이완 알고리즘의 하나인 MFA(Mean Field Annealing) 방법을 적용하여 지역적 최소치로의 수렴문제를 해결하는 방법을 제시한다. MFA는 모의 애닐링의 통계적 성질을 변수의 평균값에 적용하는 결정론적인 수정 법칙들로 대신하고, 이러한 평균값을 최소화함으로서 수렴속도를 개선한 방법이다 아울러 신경망이 갖고 있는 문제점인 과다한 클래스 패턴의 생성에 따른 처리속도 지연의 문제점을 해결하기 위하여 응집 군집화 알고리즘을 이용하여 영상을 구성하는 군집을 결정하여 신경망에 입력되는 값을 초기화하여 영상패턴이 증가되는 것을 제한하였다. 결과: 본 논문에서 제시된 응집 군집화 방법 및 결정론적 이완 방법은 신경망에 의한 자기공명영상의 분류 시 발생할 수 있는 지역적 최적 치로의 수렴 문제를 해결하여 전역적 최적화로 신속히 수렴함을 알 수 있었다. 결론: 본 논문에서는 클러스터의 분석과 결정론적 이완 방법에 의하여 신경망에 의한 자기공명영상의 분류결과를 향상시키기 위한 새로운 방법을 소개하였으며 실험결과를 통하여 그러한 사실을 확인할 수 있었다. Purpose : This paper introduces an improved classification approach which adopts a deterministic relaxation method and an agglomerative clustering technique for the classification of MRI using neural network. The proposed approach can solve the problems of convergency to local optima and computational burden caused by a large number of input patterns when a neural network is used for image classification. Materials and methods : Application of Hopfield neural network has been solving various optimization problems. However, major problem of mapping an image classification problem into a neural network is that network is opt to converge to local optima and its convergency toward the global solution with a standard stochastic relaxation spends much time. Therefore, to avoid local solutions and to achieve fast convergency toward a global optimization, we adopt MFA to a Hopfield network during the classification. MFA replaces the stochastic nature of simulated annealing method with a set of deterministic update rules that act on the average value of the variable. By minimizing averages, it is possible to converge to an equilibrium state considerably faster than standard simulated annealing method. Moreover, the proposed agglomerative clustering algorithm which determines the underlying clusters of the image provides initial input values of Hopfield neural network. Results : The proposed approach which uses agglomerative clustering and deterministic relaxation approach resolves the problem of local optimization and achieves fast convergency toward a global optimization when a neural network is used for MRI classification. Conclusion : In this paper, we introduce a new paradigm to classify MRI using clustering analysis and deterministic relaxation for neural network to improve the classification results.

유틸리티 네트워크와 수소 공급망 통합 네트워크 설계를 위한결정론적 최적화 모델 개발

황보순호 ( Soon Ho Hwang Bo ),한지훈 ( Jee Hon Han ),이인범 ( In Beum Lee ) 한국화학공학회 2014 Korean Chemical Engineering Research(HWAHAK KONGHA Vol.52 No.5

Lots of networks are constructed in a large scale industrial complex. Each network meet their demandsthrough production or transportation of materials which are needed to companies in a network. Network directly producesmaterials for satisfying demands in a company or purchase form outside due to demand uncertainty, financial factor,and so on. Especially utility network and hydrogen network are typical and major networks in a large scale industrialcomplex. Many studies have been done mainly with focusing on minimizing the total cost or optimizing the networkstructure. But, few research tries to make an integrated network model by connecting utility network and hydrogen networkIn this study, deterministic mixed integer linear programming model is developed for integrating utility networkand hydrogen network. Steam Methane Reforming process is necessary for combining two networks. After producing hydrogen from Steam-Methane Reforming process whose raw material is steam vents from utility network, producedhydrogen go into hydrogen network and fulfill own needs. Proposed model can suggest optimized case in integrated networkmodel, optimized blueprint, and calculate optimal total cost. The capability of the proposed model is tested byapplying it to Yeosu industrial complex in Korea. Yeosu industrial complex has the one of the biggest petrochemicalcomplex and various papers are based in data of Yeosu industrial complex. From a case study, the integrated networkmodel suggests more optimal conclusions compared with previous results obtained by individually researching utilitynetwork and hydrogen network.

A visiting scheme of mobile sink system in distributed sensor networks

Sang-Joon Park(박상준),Jong-Chan Lee(이종찬) 한국컴퓨터정보학회 2021 韓國컴퓨터情報學會論文誌 Vol.26 No.11

센서 네트워크는 네트워크 응용 목적에 따라 적합하게 설계되어야 하며, 이에 따라서 유효한 응용 기능을 지원할 수 있다. 특정 네트워크 환경을 고려하지 않은 일반적인 전략을 사용하는 것 보다 적합한 네트워크 모델의 설계를 기반으로 네트워크 수명시간을 극대화 시킬 수 있다. 본 논문에서는 분산 무선 센서 네트워크에서 이동 싱크에 대한 비결정형 에이전트 방식을 제안한다. 센서 네트워크 지역은 여러 분산 구역으로 나누어질 수 있다. 그러므로 이러한 네트워크에 대해 만족스러운 네트워크 관리를 구현하기 위하여 특정 네트워크 모델에 따른 적합한 방식이 요구된다. 본 논문에서는 제안한 방식에 대한 분석과 시뮬레이션 결과의 평가를 제공한다. The sensor networks should be appropriately designed by applied network purpose, so that they can support proper application functions. Based on the design of suitable network model, the network lifetime can be maximized than using other general strategies which have not the consideration of specific network environments. In this paper, we propose a non-deterministic agent scheme to the mobile sink in distributed wireless sensor networks. The sensor network area can be divided into several sensor regions. Hence, to these such networks, the specified suitable scheme is requested by the applied network model to implement satisfactory network management. In this paper, we theoretically represent the proposed scheme, and provide the evaluation with the simulation results.

A Study on Reinforcement Learning Techniques for Deterministic Ultra-Low Latency in 5G and Beyond

Lilian C. Mutalemwa,Seokjoo Shin 한국차세대컴퓨팅학회 2021 한국차세대컴퓨팅학회 학술대회 Vol.2021 No.05

The IEEE 802.1 time sensitive networking (TSN) and IETF deterministic networking (DetNet) standards guarantee ultra-low latency (ULL) communications in 5G networks and beyond. The DetNet standard can warrant deterministic ULL through the use of reinforcement learning (RL)-based data forwarding algorithms. Therefore, this study presents an overview of the DetNet mechanisms and explores the RL data forwarding techniques. It is shown that RL algorithms are capable of adjusting effectively the data transmission for deterministic applications, according to the resource usage of the networks.

Energy Optimization for Deterministic Cluster Routing in Wireless Sensor Networks

강상혁 한국통신학회 2024 韓國通信學會論文誌 Vol.49 No.2

We investigate the optimization of the number of clusters for load-balanced deterministic cluster routing schemes in wireless sensor networks. Taking the advancement of technologies into account, we assume sensor networks with high-frequency transmission signals and low-power consumption circuitry. By analyzing the operations of cluster head nodes and non-cluster head nodes in each round, we calculate energy consumption as a function of the size of sensor field, the number of nodes, circuit energy consumption, and transmission power factors. The energy consumption function leads us to determine the optimum number of clusters in a deterministic cluster routing algorithm, which prolongs the network lifespan with a minimum level of energy consumption. Given the optimum number of clusters, we also investigate the effect of cluster formations on energy optimization by relating the formation to the problem of densest packing of congruent circles in a circle. Our optimization method is verified by computer simulations.

확정적 네트워크에서의 동적 처리순위를 활용한 강화학습 기반 스케줄러

류지혜,박규동,권주혁,정진우 한국통신학회 2023 韓國通信學會論文誌 Vol.48 No.4

Smart industry, metaverse, digital-twin, and military applications require deterministic data delivery in large scale networks. This paper proposes reinforcement learning-based scheduling that assigns dynamically different precedences to the flows, in addition to the flow's class or priority, and determines the scheduling algorithm according to the flow's precedence. In the proposed reinforcement learning-based scheduling algorithm with two precedence queues, the reinforcement learning agent takes two actions that assigns the precedence of flows according to a specified criterion and selects a scheduling algorithm. Depending on the purpose of the network, any factor with high importance could be a criterion for determining the precedence. In this study, the deadline required by the flow is designated as the major factor for precedence decision. By utilizing DDQN (Double Deep Q-Network), a deep learning-based reinforcement learning model, the precedence and the scheduling algorithm are determined by observing the state of the network and selecting an action at each decision period with a fixed length. In the network simulator developed for the study, it was confirmed that the DDQN agent showed better performance than various heuristic algorithms. 스마트 인더스트리, 메타버스, 디지털 트윈, 군사용 어플리케이션 등에서 확정적 데이터 전달을 요구하고 있다. 본 논문은 일반적으로 통용되는 플로우들의 클래스 혹은 우선순위와는 별도로, 네트워크 상황과 중요도에 따라 플로우 별로 동적으로 처리순위(precedence)를 할당하고, 이에 따라 스케줄링 알고리즘을 결정하는 강화학습 기반의스케줄링 프레임워크를 제안한다. 이를 실증하기 위해서 두 개의 처리순위 큐가 존재하는 환경을 상정하여, 강화학습 에이전트가 지정된 기준에 따라 플로우들의 처리순위를 지정하며 스케줄링 알고리즘을 선택하는 두 가지의행동(action)을 취한다. 네트워크 특성에 따라 다양한 기준으로 처리순위를 결정할 수 있다. 본 연구에서는 플로우가 요구하는 마감기한(deadline)을 처리순위 결정의 중요한 기준으로 사용하였다. 딥러닝 기반의 강화학습 모델인DDQN(Double Deep Q-Network)을 활용하여, 고정된 길이의 결정 주기마다 네트워크의 상태(state)를 관측하고행동을 선택함으로써 처리순위를 결정한다. 본 연구의 환경에 맞게 개발한 네트워크 시뮬레이터를 통해 DDQN 에이전트가 여러 휴리스틱 알고리즘과 비교하여 높은 성능을 보이는 것을 확인하였다.

Statistical Investigation of Connected Structures of Stock Networks in a Financial Time Series

오갑진,엄철준,김승환 한국물리학회 2008 THE JOURNAL OF THE KOREAN PHYSICAL SOCIETY Vol.53 No.6

In the study, we have evaluated the deterministic factors that can in°uence the connected struc-ture of a stock network. The representative index for the topological properties of a stock network is the number of links it has with other stocks. We utilize the coe±cient of determination from the multi-factor model, which attempts to explain stock returns with several explanatory variables. In this study, factors refer to common factors that commonly a®ect stocks, on the basis of the arbitrage pricing theory acknowledged in the relevant ¯nancial literature. We utilize individual stocks traded on the market indices of Korea, Japan, Canada, Italy, and the UK. We ¯nd that stocks with a large number of links in a stock network can be better explained by factors in the multi-factor model than stocks with a small number of links with other stocks. These results indicate that common factors in the ¯eld of ¯nance are signi¯cant deterministic factors that should be taken into consideration when constructing a stock network.

이기종 네트워크를 구성하는 무선접속 기술간 주파수 이격 분석

이경재 ( Kyoung-jae Lee ),조한신 ( Han-shin Jo ) 한국인터넷정보학회 2015 인터넷정보학회논문지 Vol.16 No.6

본 논문에서는 이기종 네트워크를 구성하는 대표 무선접속 기술인 WiMAX, CDMA, 무선랜 간의 주파수 공유 분석을 수행하였다. 3가지 접속기술은 각기 다른 주파수 대역을 사용하지만 대역외 방사특성으로 인해 타 무선접속 기술에 간섭을 일으킬 수 있다. 따라서 간섭 허용기준을 만족하기 위한 최소 인접채널 간섭률을 산출하고 이를 바탕으로 간섭/피간섭 시스템간의 최소 주파수 이격(보호 대역)을 산출하였다. 분석 결과 WiMAX와 무선랜 간에는 적어도 20 MHz의 보호대역, WiMAX와 CDMA시스템 간에는 적어도 15 MHz의 보호대역이 필요함을 확인하였다. In this paper, we study the coexistence of typical radio access technologies (WiMAX, CDMA, WLAN) in heterogeneous networks. Although the radio accesses employ different frequency band, they can interfere with each other due to out-of- band emission. We compute the minimum adjacent channel interference ratio (ACIR) to satisfy the allowable level of interference, and the resulting minimum frequency separation (guard band) between interfering and victim system. We observe that WiMAX-WLAN coexistence and WiMAX-CDMA coexistence are feasible with at least 20 MHz and 15 MHz guard bands, respectively.

Development of an integrated network for utility supply and carbon dioxide mitigation systems: applicability of biodiesel production

Ahn, Yuchan,Han, Jeehoon ELSEVIER 2019 JOURNAL OF CLEANER PRODUCTION Vol.232 No.-

<P><B>Abstract</B></P> <P>This study develops a deterministic model for optimal design of an integrated network to determine the utility supply (US) and CO<SUB>2</SUB> mitigation (CM) systems in consideration of technology to utilize CO<SUB>2</SUB> as a raw material. The objective of the model is to minimize the expected total cost of an integrated network that satisfies US and CM demands of multi-site companies in an industrial complex during a multi-period planning horizon. This model determines the optimal locations and amounts of: (1) the utility (steam) transferred among companies, and (2) CO<SUB>2</SUB> storage (CS) and (3) CO<SUB>2</SUB> utilization (CU) considering CO<SUB>2</SUB> capture (CC) systems. The proposed model is tested by applying it to Yeosu Industrial Complex in Korea. The total cost for the Alternative Model that considers CS and CU systems (US$ 411.62 × 10<SUP>6</SUP>/y) is 24.9% (US$ 136.49 × 10<SUP>6</SUP>/y difference) lower than the Base Model that considers only the CS system (US$ 548.11 × 10<SUP>6</SUP>/y). The most prominent difference between two models was the variation in the CM system. This study examined whether the variation of the integrated network would be affected by the variety of CM systems, and confirmed that CO<SUB>2</SUB> was supplied to produce maximum biodiesel yield when the CU system was selected. This result would be interested in many researchers to study the supply chain network problem over multi-site and multi-period planning horizon while considering the capability of biodiesel production based on CO<SUB>2</SUB> as a raw material.</P> <P><B>Highlights</B></P> <P> <UL> <LI> Utility supply and CO<SUB>2</SUB> mitigation systems are integrated to optimize the network model. </LI> <LI> Microalgae-based biodiesel production is considered as a CO<SUB>2</SUB> mitigation system. </LI> <LI> This model considers multi-site and multi-period planning problems. </LI> <LI> Integrated network model is formulated using mixed-integer linear programming. </LI> <LI> Optimal design of the integrated model is determined under two deterministic demands. </LI> </UL> </P>