In industrial environments, most production-related activities performed by human operators are often complex. Accurate detections of these activities are pivotal as it can greatly help to assess productivity that can lead to improvement in worker training, as well as in other scenarios ensure a safe work environment and reducing injuries. Existing datasets on wearable Internet of Things (IoT) for human activity recognition primarily focuses on general activities, such as walking, running, etc., and therefore, related machine learning models and datasets are not suitable for application to industrial environments. In this paper, we present a novel dataset for classifying human operator activities in a meat processing plant where production line operators use knives to cut, process and produce meat products. Our dataset contains human operator activity data captured using wearable IoT sensors collected from a meat processing production facility. Through extensive experiments using machine and deep learning, we demonstrate that our dataset is effective and useful for detecting different activities of a human operator working in an industrial environment. To the best of our knowledge, this is the only real-world IoT dataset that will be made publicly available to support further research into industrial activities recognition. Our dataset and related experiments are available at https://digitalinnovationlab.github.io/mppdataset.

Producing high-quality product consistently is crucial in manufacturing, as discarding or reprocessing low-quality products increases waste and energy consumption and reduces overall production efficiency. Ensuring high-quality manufactured products is challenging due to relying on human activities for product quality and related consistency assessment, which is often performed postproduction instead of assessing these during each production run and making real-time production adjustments that can mitigate product quality and related consistency issues. In this article, we proposes a novel machine-state learner algorithm that captures the dependencies between product quality and related consistency and machine data (specifically the machine settings and corresponding sensor data). In addition, this article shows how this novel machine-state learner algorithm can be used to predict product quality during the production runs and how such prediction are used to make machine setting recommendations that mitigate product quality and related consistency issues before or during the production runs. These advances in machine state-based data modeling, predictive data analysis and recommendation are incorporated into an inline prediction and decision support system that achieves significant improvement in producing high-quality products consistently by guiding decision-making via recommendations during production in a digital manufacturing environment. In this article, we present an evaluation of the above contributions in a real-world manufacturing plant and yield double digit first pass and nearly perfect second pass product improvements in terms of product quality and related consistency and production efficiency.

Abstract The transformation of cities into smarter and more efficient environments relies on proactive and timely detection and maintenance of city-wide infrastructure, including roadside infrastructure such as road signs and the cleaning of illegally dumped rubbish. Currently, these maintenance tasks rely predominantly on citizen reports or on-site checks by council staff. However, this approach has been shown to be time-consuming and highly costly, resulting in significant delays that negatively impact communities. This paper presents AIoT-CitySense, an AI and IoT-driven city-scale sensing framework, developed and piloted in collaboration with a local government in Australia. AIoT-CitySense has been designed to address the unique requirements of roadside infrastructure maintenance within the local government municipality. A tailored solution of AIoT-CitySense has been deployed on existing waste service trucks that cover a road network of approximately 100 kms in the municipality. Our analysis shows that proactive detection for roadside infrastructure maintenance using our solution reached an impressive 85%, surpassing the timeframes associated with manual reporting processes. AIoT-CitySense can potentially transform various domains, such as efficient detection of potholes and precise line marking for pedestrians. This paper exemplifies the power of leveraging city-wide data using AI and IoT technologies to drive tangible changes and improve the quality of city life.

Low-cost transient resources such as Amazon's Elastic Compute Cloud (EC2) Spot instances can be opportunistically leveraged to reduce the ongoing costs of cloud applications. However, they are susceptible to unilateral revocations by the vendor making them a risky proposition for long-running applications with strict performance requirements. It is challenging to effectively balance the cost savings that transient resources provide with the associated revocation risk which, if realised, can impact application performance. To address this challenge, we propose an approach for risk-aware cloud cost optimization that is inspired by the concept ofportfolio diversification.Contract diversificationmitigates the revocation risk by procuring the required compute capacity as a mixed portfolio of transient and non-transient resources.Resource diversificationfurther diversifies the risk by using multiple transient resource types. Using our approach, consumers can leverage contract and resource diversification to proactively (re-)configure their application's resource portfolio to handle workload and resource price fluctuations while minimizing ongoing cost and keeping ongoing revocation risk within tolerable limits. Simulative evaluation using three real-world workload traces and Amazon's EC2 offerings demonstrate that our proposed approach can achieve meaningful cost savings compared to the baseline costs, while significantly reducing the portfolio's exposure to revocation risk.

Timely detection of roadside assets that require maintenance is essential for improving citizen satisfaction. Currently, the process of identifying such maintenance issues is typically performed manually, which is time consuming, expensive, and slow to respond. In this paper, we present Mobile IoT-RoadBot, a mobile 5G-based Internet of Things (IoT) solution, powered by Artificial Intelligence (AI) techniques to enable opportunistic real-time identification and detection of maintenance issues with roadside assets. The Mobile IoT-RoadBot solution has been deployed on 11 bin service (waste collection) trucks in the western suburbs of Melbourne, Australia, performing real-time assessments of road-side assets as they service areas within the local government. We present the architecture of Mobile IoT-RoadBot and demonstrate its capability via an online 'points of maintenance' (PoMs) map.

Cloud consumers have access to an increasingly diverse range of resource and contract options, but lack appropriate resource scaling solutions that can exploit this to minimize the cost of their cloud-hosted applications. Traditional approaches tend to use homogeneous resources and horizontal scaling to handle workload fluctuations and do not leverage resource and contract heterogeneity to optimize cloud costs. In this paper, we propose a novel opportunistic resource scaling approach that exploits both resource and contract heterogeneity to achieve cost-effective resource allocations. We model resource allocation as anunbounded knapsack problem, and resource scaling as anone-step ahead resource allocation problem. Based on these models, we propose two scaling strategies: (a)delta capacity optimization, which focuses on optimizing costs for the difference between existing resource allocation and the required capacity based on the forecast workload, and (b)full capacity optimization, which focuses on optimizing costs for resource capacity corresponding to the forecast workload. We evaluate both strategies using two real world workload datasets, and compare them against three different scaling strategies. The results show that our proposed approach, particularly full capacity optimization, outperforms all of them and offers in excess of 70 percent cost savings compared to the traditional scaling approach.

A chatbot is a software that interacts with humans by conducting conversations via textual or auditory methods. Chatbots have recently been used for plethora of applications including travel, medical, education, retail etc. Several cloud-based platforms (e.g. IBM, Amazon, Google, Microsoft) are available for developing and deploying chatbots. However, there is a lack of an evaluation methodology and a tool for evaluating chatbots comprehensively. Current approaches for comparing cloud-based chatbots are manual and rely on expert's judgement. In this short paper, we propose, devise, implement and demonstrate a tool namely ECHO for empirical evaluation of cloud-based chatbots. ECHO is capable of automatically evaluating multiple cloud-based chatbots and report the outcomes of the comparative evaluation. We validate the efficacy of ECHO by conducting comparative evaluation of 3 popular cloud-based chatbots in 2 different question-answering application scenarios with 3 levels of complexities.

Smart buildings equipped with various building management systems and digital control systems produce enormous amounts of sensor data that can be used to investigate and diagnose operational issues such as unsatisfactory thermal comfort outcomes, excessive energy consumption and/or predicting failures before they occur. However, current building management systems often face the issues with incomplete or unstructured metadata associated with sensor data which prevent such pro-active, predictive and prescriptive analysis. Currently, building service engineers manually map the sensor data streams to aid their diagnostic process. This process is expensive, ineffective and is also prone to human errors. This paper proposes a novel semi-automated approach that annotates incoming sensor data streams. We also propose extensions to Project Haystack, a well-known ontology used for naming conventions and taxonomies for building equipment and operational data. We have developed a tool that is currently used by our industry partner and incorporates the proposed automatic annotation approach and maps the data streams to our Haystack-extended ontology. The tool includes an easy to use interface for engineers to easily diagnose issues in mechanical building services. The proposed approach has been validated via both usability and technical evaluation.

Smart mobile devices coupled with the Internet of Things (IoT) and Artificial Intelligence (AI) have emerged as a key enabler of modern digital health applications. While cloud computing is now a well established paradigm for analysing IoT captured data in mobile health applications, on-board analysis of data using AI approaches such as Deep Learning (DL) is gaining significant momentum. This is driven primarily by advances in on-board resources enabling modern mobile devices to execute complex DL models, while also offering improved response time and accuracy for rapid decision-making, and enhanced user privacy. While the number of mobile digital health applications that use IoT and DL is increasing, progress is currently impeded by a lack of framework for profiling and evaluating the performance of DL models on mobile devices. To this end, we propose MobDL, a framework for profiling and evaluating DL models running on smart mobile devices. We present the architecture of this framework and devise a novel evaluation methodology for conducting quantitative comparisons of various DL models running on mobile devices. Three diverse digital health applications using heterogeneous data (e.g. image, time series) are introduced. We conduct extensive experimental evaluations using several DL models that have been developed using the data sets obtained for the three digital health applications to validate the effectiveness of the proposed MobDL framework.

Obstructive sleep apnea (OSA) is a serious disorder in which people repeatedly stop breathing during sleep. This paper presents the Technology Integrated Apnea Respiration Analyser (TIARA), an innovative wearable device as an alternative to polysomnography (PSG) that has the potential to enable cost-effective and repeated overnight tests of OSA even in a home environment. PSG normally collects bio-signals such as brain waves, oxygen saturation, heart rate in a controlled environment (e.g. sleep lab) to diagnose sleep disorders. Here we present the design and development of the TIARA device and demonstrate its potential to perform as well as PSG at differentiating sleep and wake states and different phases of sleep, based on machine learning models.

The Industrial Internet of Things (IIoT) is a key pillar of the Fourth Industrial Evolution or Industry 4.0. It aims to achieve direct information exchange between industrial machines, people, and processes. By tapping and analysing such data, IIoT can more importantly provide for significant improvements in productivity, product quality, and safety via proactive detection of problems in the performance and reliability of production machines, workers, and industrial processes. While the majority of existing IIoT research is currently focusing on the predictive maintenance of industrial machines (unplanned production stoppages lead to significant increases in costs and lost plant productivity), this paper focuses on monitoring and assessing worker productivity. This IIoT research is particularly important for large manufacturing plants where most production activities are performed by workers using tools and operating machines. With this aim, this paper introduces a novel industrial IoT solution for monitoring, evaluating, and improving worker and related plant productivity based on workers activity recognition using a distributed platform and wearable sensors. More specifically, this IIoT solution captures acceleration and gyroscopic data from wearable sensors in edge computers and analyses them in powerful processing servers in the cloud to provide a timely evaluation of the performance and productivity of each individual worker in the production line. These are achieved by classifying worker production activities and computing Key Performance Indicators (KPIs) from the captured sensor data. We present a real-world case study that utilises our IIoT solution in a large meat processing plant (MPP). We illustrate the design of the IIoT solution, describe the in-plant data collection during normal operation, and present the sensor data analysis and related KPI computation, as well as the outcomes and lessons learnt.

Air quality is an important factor in planning activities in our everyday life. The information presented though captured data using Internet of Things (IoT) in smart cities is mostly single-dimensional where citizens do not have much opportunities to directly interact with the system to get personalised insights. Recent years have seen dire reports of extreme air pollution in mega cities around the world, which has led to government authorities grappling with solutions. Taking into account the existing IoT sensor setup in smart cities, it is now very convenient to visually explore the level of pollution of any places in real-time. In this context, this paper presents AqVision, a flexible visualisation tool for future citizens in smart cities that combines personalised awareness with generalised needs and leverages to envisage air pollution hotspots using more interactive manners considering individualised health and safety concerns.

Context-aware monitoring is an emerging technology that provides real-time personalised health-care services and a rich area of big data application. In this paper, we propose a knowledge discovery-based approach that allows the context-aware system to adapt its behaviour in runtime by analysing large amounts of data generated in ambient assisted living (AAL) systems and stored in cloud repositories. The proposed BDCaM model facilitates analysis of big data inside a cloud environment. It first mines the trends and patterns in the data of an individual patient with associated probabilities and utilizes that knowledge to learn proper abnormal conditions. The outcomes of this learning method are then applied in context-aware decision-making processes for the patient. A use case is implemented to illustrate the applicability of the framework that discovers the knowledge of classification to identify the true abnormal conditions of patients having variations in blood pressure (BP) and heart rate (HR). The evaluation shows a much better estimate of detecting proper anomalous situations for different types of patients. The accuracy and efficiency obtained for the implemented case study demonstrate the effectiveness of the proposed model.

Chronic diseases are major causes of deaths in Australia and throughout the world. This necessitates the need for a self-care, preventive, predictive and protective assisted living system where a patient can be monitored continuously using wearable and wireless sensors. In real-time home monitoring system, various biological signals of a patient are obtained continuously using a mobile device (smart phone or tablet) and sent to the cloud to discover patient-specific abnormalities. The objective of this work is to develop a probabilistic model that identifies the future clinical abnormalities of a patient using recent and past values of multiple vital signs (e.g. heart rate, blood pressure, respiratory rate). Chronic patients living alone in home die of various diseases for the lack of an efficient automated system having prior prediction ability in the irregularities of vital signs. In this paper, Hidden Markov Model (HMM) is adopted to predict different clinical onsets using the temporal behaviours of six biosignals. The HMM models are trained and evaluated using continuous monitoring data of more than 1000 patients collected from the MIMIC-II database of MIT physiobank archive. The best models are selected using expectation maximisation (EM) algorithm and used in personalized remote monitoring system to forecast the most probable forthcoming clinical states of a continuously monitored patient. The scalable power of cloud computing is utilized for fast learning of various clinical events from large samples. The results obtained from the innovative home-based monitoring application show a new approach of detecting clinical anomalies using multi-parameter trends.