China
Africa
Explore chapters and articles related to this topic
A Geo-Referenced Data Collection Microservice Based on IoT Protocols for Smart HazMat Transportation
Published in Bharat Bhushan, Sudhir Kumar Sharma, Bhuvan Unhelkar, Muhammad Fazal Ijaz, Lamia Karim, Internet of Things, 2022
Ghyzlane Cherradi, Azedine Boulmakoul, Lamia Karim, Meriem Mandar
The collection system was designed to be part of a larger IoT infrastructure, which means it should serve as a data provider to a larger IoT platform. In this way, the data collected can be used in other contexts, for example related to smart parking or fleet management. This was inspired by the fact that sensors, platforms, and services operate in different locations and are provided by different entities. With the influence of distributed architectures and the fog computing paradigm, the proposed system is designed as event-driven microservices (see Figure 5.2). This design gives the system great flexibility in scalability and adaptability to various data applications [2, 25]. The system needs to integrate dynamic data streams to create predefined tasks in near real time. Several streams of spatiotemporal data are sent regularly from the onboard system via the MQTT protocol. To enable the integration and management of these streams, we have created a cluster of scalable local brokers to which the onboard systems send their data.
Online conflict resolution strategies for human activity recognition in smart homes
Published in Journal of Control and Decision, 2023
Amina Jarraya, Amel Bouzeghoub, Amel Borgi
Human Activity Recognition (HAR) research in the smart home environment is an active area due to its numerous applications (Ranasinghe et al., 2016). These applications have completely increased such as healthcare, fitness and sports (Ermes et al., 2008), child and elderly care (Khan & Sohn, 2011), home monitoring (Das & Cook, 2004, January). etc. In general, smart homes employ different types of sensors: photocell, presence sensor, motion sensor, proximity sensor, temperature sensor, etc. In such environments, data driven applications have to exploit the large amount of data captured from distributed, heterogeneous, and dynamic (i.e. whose characteristics are varying over time) sensor data to identify current human activities. However, using a high volume of dynamic data also presents important challenges when dealing with distributed HAR in smart homes: C1: sensor data management: how to handle distributed data coming from deployed sensors in different locations of the smart home? C2: freshness: how to keep data freshness ? C3: heterogeneity: how to deal with the nature of sensor data? C4: uncertainty: how to trust data arrival from other sensors ? C5: conflict: how to treat contradictory identified activities coming from different agents? C6: online learning: how to handle the time-varying characteristics of the underlying data, i.e. adequately deal with the concept-drift (Zliobaite, 2010). The collected data are streams where observations come in one by one and need to be analysed online.
Goal-Oriented Mesh Adaptivity of Three-Dimensional Neutron Transport Calculation Using Weighted Difference Scheme and Dual-Weighted Residual Error Indicators
Published in Nuclear Science and Engineering, 2023
Cong Liu, Junxia Wei, Bin Zhang, Jinhong Li, Zhiqiang Sheng, Shuang Tan
To meet the requirements of dynamic data management, multilayer structure variables are used to store flux and source moments. A grid can be defined by a level number and the grid index at its level. The grid state identifier indicates whether the given grid is in the active or sleep state, and data pointers and physical quantities are assigned to the active grid. During mesh refinement and coarsening, memory allocation and recovery of individual meshes do not affect other meshes. Unfortunately, because of its low cache hit rate, the memory access to the variable using multilayer structures is much slower than that using one-layer arrays. The transport sweep process requires repeatedly reading and updating physical quantities that depend on the iterative solutions. The low access speed greatly affects the execution efficiency of the adaptive algorithm and even offsets the benefits of the reduced computational amount. During source iterations, we use allocatable arrays instead of structure variables to solve the problem of slow memory access. Knowing the total number of active meshes in a given AMR round, we allocate the fixed-size array variables before the iterative computation. We record the correspondence between the level-grid numbers in the multilevel structure and the positions in the array. The physical quantities in the multilevel structure are assigned to these arrays as initial guesses for the source iterations. After the stopping criterion is met, the iterative solutions are stored in the multilevel structure, and these arrays are de-allocated after the SN calculation of this round is completed. This process is repeated when the adaptive algorithm adjusts computational meshes.