Transmitting updates of high-dimensional models between client IoT devices and the central aggregating server has always been a bottleneck in collaborative learning - especially in uncertain real-world IoT networks where congestion, latency, bandwidth issues are common. In this scenario, gradient quantization is an effective way to reduce bits count when transmitting each model update, but with a trade-off of having an elevated error floor due to higher variance of the stochastic gradients. In this paper, we propose ElastiCL, an elastic quantization strategy that achieves transmission efficiency plus a low error floor by dynamically altering the number of quantization levels during training on distributed IoT devices. Experiments on training ResNet-18, Vanilla CNN shows that ElastiCL can converge in much fewer transmitted bits than fixed quantization level, with little or no compromise on training and test accuracy.

Industry 4.0 is a recent trend of automation for manufacturing technologies and represents the fourth industrial revolution which transforms current industrial processes with the use of technologies such as automation, data analytics, cyber-physical systems, IoT, artificial intelligence, etc. The vision of Industry 4.0 is to build an end-to-end industrial transformation with the support of digitization. Data analytics plays a key role to get a better understanding of business processes and to design intelligent decision support systems. However, a key challenge faced by industry is to integrate multiple autonomous processes, machines and businesses to get an integrated view for data analytics activities. Another challenge is to develop methods and mechanisms for real-time data acquisition and analytics on-the-fly. In this paper, we propose a semantically interoperable framework for historical data analysis combined with real-time data acquisition, event detection, and real-time data analytics for very precise production forecasting within a manufacturing unit. Besides historical data analysis techniques, our middleware is capable of collecting data from diverse autonomous applications and operations in real time using various IoT devices, analyzing the collected data on the fly, and evaluating the impact of any detected unexpected events. Using semantic technologies we integrate multiple autonomous systems (e.g. production system, supply chain management and open data). The outcome of real-time data analytics is used in combination with machine learning models trained over historical data in order to precisely forecast production in a manufacturing unit in real time. We also present our key findings and challenges faced while deploying our solution in real industrial settings for a large manufacturing unit.

Summary With an ever‐increasing variety and complexity of Internet of Things (IoT) applications delivered by increasing numbers of service providers, there is a growing demand for an automated mechanism that can monitor and regulate the interaction between the parties involved in IoT service provision and delivery. This mechanism needs to take the form of a contract, which, in this context, is referred to as a service level agreement (SLA). As a first step toward SLA monitoring and management, an SLA specification is essential. We believe that current SLA specification formats are unable to accommodate the unique characteristics of the IoT domain, such as its multilayered nature. Therefore, we propose a grammar for a syntactical structure of an SLA specification for IoT. The grammar is built based on a proposed conceptual model that considers the main concepts that can be used to express the requirements for hardware and software components of an IoT application on an end‐to‐end basis. We followed the goal question metric approach to evaluate the generality and expressiveness of the proposed grammar by reviewing its concepts and their predefined lists of vocabularies against two use cases with a considerable number of participants whose research interests are mainly related to IoT. The results of the analysis show that the proposed grammar achieved 91.70% of its generality goal and 93.43% of its expressiveness goal.

While the challenge of connecting Internet of Things (IoT) devices at the lowest layer has been widely studied, integrating and interoperating huge amounts of sensed data of heterogeneous IoT devices is becoming increasingly important because of the possibility of consuming such data in supporting many potential novel IoT applications. A common approach to processing and consuming IoT data is a centralized paradigm: sensor data is sent over the network to a comparatively powerful central server or a cloud service, where all processing takes place. However, this approach has some limitations as it requires devices to interact directly with a cloud which is not cost effective. First, it has high demands on the device's storage and computational capabilities. Second, as devices grow rapidly in a deployment area, sending all the data to a centralized cloud server requires high network bandwidth. Moreover, this often creates data privacy concerns as all raw data will be sent to a centralized place. To address the above limitations for building future Internet of Things applications, we present an early design of a novel framework that combines Internet of Things, Semantic Web, and Big Data concepts. We not only present the core components to build an IoT system, but also list existing alternatives with their merits. This framework aims to incorporate open standards to address the potential challenges in building future IoT applications. Therefore, our discussion revolves around open standards to build the framework, rather than proprietary standards.

Internet of Things (IoT) application development is technically challenging and complex due to several factors such as the presence of different IoT protocols, interoperability issues, lack of industry standards and dealing with the heterogeneity that exists both in Physical and Internet worlds. Furthermore, developers involved in the IoT application development have to address issues pertaining to different life-cycles ranging from design, implementation to deployment (deployment complexities due to the possibilities of wide range of devices). Manual effort involved in all above three phases for heterogeneous devices is a time-consuming and error-prone process. We demonstrate IoTSuite which is an application development platform consisting of a suite of tools motivated by the need to simplify IoT application development. The IoTSuite supports several features such as automatic code generation and integration with a set of modeling languages. It takes high-level specification as input that abstract heterogeneity related complexity. It integrates compiler and deployment module to provide automation at different phases of application development process.

The last few years have seen two key trends maturing in the industry: IoT and Big Data. These trends build on more than a decade of research in both academia and industry. As the cost of instrumentation and microprocessor chips has declined, it is now possible to monitor the environment on a widening scale. The cost decline is matched by cloud computing (exposed as web services) that provides infrastructure for storage and processing. Furthermore, on top of cloud, advances in big data tools/techniques provide a platform to analyze and understand the massive amount of data. But data generated from IoT is expected not only to be Big (volume) but also Fast (velocity). Data analysis and machine learning will play a key role in unlocking the value generated by IoT data. The Internet of Things (IoT) can be thought of in terms of connecting and combining the above mentioned elements. Combination of all these elements at various levels (e.g., physical objects, cloud service, mobile with rich user interfaces, analytics) will allow access and analysis of an enormous amount of fast data, which could be used to improve efficiency and performance of the whole enterprise. Moreover, it opens the possibilities of developing IoT applications in novel scenarios such as smart metering, smart electric car recharge stations, retail & logistics, and so on. An important challenge that needs to be addressed is to enable the rapid development of IoT applications. Similar challenges have already been addressed in the closely related fields of Wireless Sensor and Actuator Networks (WSANs) and Pervasive/Ubiquitous computing. While the main challenge in the former is largely limited to similar nodes, the primary concern in the latter largely has been the heterogeneity of physical objects. The upcoming field of IoT will include both WSANs as well as heterogeneous physical objects, in addition to this it brings heterogeneity at various levels (e.g., physical objects, cloud services, smart phones with rich user interfaces, analytics). Therefore, Software Engineering (SE) support for IoT applications is needed to develop methodologies, abstractions, and techniques. Nevertheless, so far this topic has received very little attention by SE community.