A5056296517- Energy Efficient Wireless Sensor Networks
- IoT and Edge/Fog Computing
- Indoor and Outdoor Localization Technologies
- Consumer Perception and Purchasing Behavior
- Reinforcement Learning in Robotics
- Advanced Computational Techniques and Applications
- Technology and Data Analysis
- Human Pose and Action Recognition
- Network Time Synchronization Technologies
- Caching and Content Delivery
- Personal Information Management and User Behavior
- Modular Robots and Swarm Intelligence
- Digital Transformation in Industry
- Scheduling and Optimization Algorithms
- Security in Wireless Sensor Networks
- Robotics and Automated Systems
- Adversarial Robustness in Machine Learning
- Wireless Body Area Networks
- Face recognition and analysis
- Human Mobility and Location-Based Analysis
- Music and Audio Processing
- Advanced Neural Network Applications
- Multi-Agent Systems and Negotiation
- Energy Load and Power Forecasting
- Autonomous Vehicle Technology and Safety
Electronics and Telecommunications Research Institute
2011-2024
Hanwha Solutions (South Korea)
2024
ABSTRACT From ancient Europe, the renaissance and industrialisation eras, to modern times, urban planning paradigms have evolved in many ways, advancing environments where people live in. Nevertheless, recent development of wireless wired communication network technologies low‐power miniature sensors for various application domains provide us with another chance revolutionising cities by making them smarter . Smart , propelled a city‐scale infrastructure, information provided from different...
In this paper, we propose the cluster-based hierarchical time synchronization protocol (CHTS) for wireless sensor networks (WSN). To couple network with real world tightly, provide a network-wide through introduction of abstractions cluster and hierarchy. Generally error has tendency being accumulated as hop count along path increases in multi-hop network. Thus randomly deployed nodes to decrease path. We then apply different optimal mechanisms heads members sake power efficiency an explicit...
The capability to extract human activity context in a room environment can be used as meaningful feedback for various wireless indoor application systems. Being able do so with easily installable resource-limited sensing components even further increase the system's applicability purposes. This paper introduces our efforts design system consisting of heterogeneous low-cost, resource-limited, platforms accurately extracting from an environment. Specifically, we introduce Resource Limited...
The increasing computational demands of Deep Reinforcement Learning (DRL) models, particularly for embedded systems in autonomous vehicles and drones, present significant challenges owing to their extensive neural network complexities. Previous DRL compression strategies predominantly focused on unstructured pruning, effective reducing model size but requiring specialized hardware acceleration. Conversely, models with structured pruning applied can be accelerated standard hardware, though...
A novel adaptive dual‐prediction scheme is introduced for minimising the data communication load wireless sensor networks as a way to maximise lifetime of resource‐limited nodes. Specifically, proposed exposes fact that when sensing context prediction used at both sink node and nodes, amount need be transmitted can minimised. Furthermore, transmission quantity reduced even more by exploiting spatial correlation among different On using this dual scheme, evaluations show transmissions...
Multi-agent RL is a process of training the agents to collaborate with others. We argue that an additional 'reality gap' in system aspects occurs when applying sim2real multi-agent RL, especially performing 'transferred' collaborative task real-world environment. In this paper, we propose ADO framework enabling decentralized participate tasks without suffering from reality gap. Our contribution threefold. First, clearly identify and summarize gaps context RL. Second, new model deal issues...
Infrared images are known to capture the thermal radiation emitted from objects and increasingly essential in Night Vision surveillance. These can be utilized various image processing algorithms, such as object detection tracking. However, infrared is highly complex due sensor degradation status of temperature inversion between background object, which results an inadequate dataset. Data augmentation approaches have been introduced overcome lack datasets by increasing diversity data...
Infrared images are known to capture the thermal radiation emitted from objects and increasingly essential in Night Vision surveillance. These can be utilized various image processing algorithms, such as object detection tracking. However, infrared is highly complex due sensor degradation status of temperature inversion between background object, which results an inadequate dataset. Data augmentation approaches have been introduced overcome lack datasets by increasing diversity data...
Several recent mobile operating systems allow users to configure the smartphone into a "driving mode". This mode suppresses smartphone's incoming SMS/call notifications so that it does not distract driving activities. However, currently available implementations keep all from being delivered, which decreases its practical usability. We identify this as problem and perform survey on preference of drivers when driving. Specifically, we ask 74 participants need for also differentiate call/SMS...
Federated Learning(FL) has emerged as the optimal approach for training machine learning models when dealing with data containing sensitive information, making sharing impractical. Particularly in contexts where privacy is a primary concern, such medical applications, Learning demonstrates its efficacy solution. Motivated by this, we have conducted comprehensive experiments using image dataset. One of key objectives these to evaluate influence Non-IID which frequently encountered Learning,...
For the past decade, wireless sensor networks have focused primarily on data collection. As a result network topology for these systems was usually heavily centralized. However, to form full system, introduction of proper actuation units and decision-making intelligence is inevitable. Such new actuator system enables architectural research issues that not been previously studied. In this work, we introduce DWSAN architecture, which effectively combines both hardware devices single manages so...
With the development of Internet-of-Things and deep learning fields, there is a growing demand for platform that provides intelligent services in various fields. Various are already being provided cloud systems, move to provide at network edge level beginning. This top-down intelligence service level, however we intend even lower device level. In this paper, introduce an low performance devices have implemented, describe its extended design implementation. order more efficiently processing...
The meaning of intelligent systems has changed over time. Traditionally, it refers to an ruled-based context aware system, but nowadays, is more often referred as a deep learning system. Since these two are still insufficient be used separately, this paper try integrate the module on SLICE, which was introduced rule-based IoT platform that easily provides services. First, in paper, pros and cons existing engine learning-based intelligence model examined. After that, how combine advantages...
The deployment of advanced metering infrastructures allows suppliers and consumers to better understand the utility supply usage chain. Data from these systems are typically used analyse in a large scale, but when observed at smaller scales, we can enable number interesting new application. In this work use data collected 300 households over three years perform detailed analysis per-household patterns. We show that introduces high variances low correlations among different even if they...
A sinkhole is occurring in urban area all over the world frequently, fact, it a roadside subsidence geological terminology, which can harm property and safety of government many people. To detect earlier, we monitor aged or damaged underground infrastructures, report sensed values to analyzing system, draw out risk index that area. In this paper, shows interface design order show important features leakage detection unit underlying concepts system.
We consider a problem of retrieving the extreme value among sensed data under deadline constraints in wireless sensor networks with potential applications to alarm systems. The is mapped score which we adopt as unified measure relative urgency data. objective retrieve maximum score. propose fully distributed schemes for contention based medium access and combining. proposed scheme uses randomized back-off controlled on be transmitted. Data combining techniques are further suppress...
Intelligent services are essential for various in a smart city. However, it is still difficult to find perfect example of intelligent our real life. Recently, the term artificial intelligence service has been used more than service. examples applying world due minimal performance requirements run model smoothly. A technology lightweighting models being developed so that can operate even on low-performance devices. If lightweight be freely devices, many suitable IoT era will flourish....