Saguna

Complex activity recognition using context-driven activity theory and activity signatures

In pervasive and ubiquitous computing systems, human activity recognition has immense potential in a large number of application domains. Current activity recognition techniques (i) do not handle variations in sequence, concurrency and interleaving of complex activities; (ii) do not incorporate context; and (iii) require large amounts of training data. There is a lack of a unifying theoretical framework which exploits both domain knowledge and data-driven observations to infer complex activities. In this article, we propose, develop and validate a novel Context-Driven Activity Theory (CDAT) for recognizing complex activities. We develop a mechanism using probabilistic and Markov chain analysis to discover complex activity signatures and generate complex activity definitions. We also develop a Complex Activity Recognition (CAR) algorithm. It achieves an overall accuracy of 95.73% using extensive experimentation with real-life test data. CDAT utilizes context and links complex activities to situations, which reduces inference time by 32.5% and also reduces training data by 66%.

M 2 C 2 : A mobility management system for mobile cloud computing

Mobile devices have become an integral part of our daily lives. Applications running on these devices may avail storage and compute resources from the cloud(s). Further, a mobile device may also connect to heterogeneous access networks (HANs) such as WiFi and LTE to provide ubiquitous network connectivity to mobile applications. These devices have limited resources (compute, storage and battery) that may lead to service disruptions. In this context, mobile cloud computing enables offloading of computing and storage to the cloud. However, applications running on mobile devices using clouds and HANs are prone to unpredictable cloud workloads, network congestion and handoffs. To run these applications efficiently the mobile device requires the best possible cloud and network resources while roaming in HANs. This paper proposes, develops and validates a novel system called M2C2 which supports mechanisms for: i.) multihoming, ii.) cloud and network probing, and iii.) cloud and network selection. We built a prototype system and performed extensive experimentation to validate our proposed M2C2. Our results analysis shows that the proposed system supports mobility efficiently in mobile cloud computing.

Forecasting heat load for smart district heating systems: A machine learning approach

The rapid increase in energy demand requires effective measures to plan and optimize resources for efficient energy production within a smart grid environment. This paper presents a data driven approach to forecasting heat load for multi-family apartment buildings in a District Heating System (DHS). The forecasting model is built using six and eleven weeks of data from five building substations. The external factors and internal factors influencing the heat load in substations are parameters used as our models input. Short-term forecast models are generated using four supervised Machine Learning (ML) techniques: Support Vector Regression (SVR), Regression Tree, Feed Forwards Neural Network (FFNN) and Multiple Linear Regression (MLR). Performance comparison among these ML methods was carried out. The effects of combining the internal and external factors influencing heat load at substations was studied. The models are evaluated with varying horizon up to 24-hours ahead. The results show that SVR has the best accuracy of 5.6% MAPE for the best-case scenario.

Remote health care cyber-physical system : quality of service (QoS) challenges and opportunities

There is a growing emphasis to find alternative non-traditional ways to manage patients to ease the burden on health care services largely fuelled by a growing demand from sections of population that is ageing. In-home remote patient monitoring applications harnessing technological advancements in the area of Internet of things (IoT), semantic web, data analytics, and cloud computing have emerged as viable alternatives. However, such applications generate large amounts of real-time data in terms of volume, velocity, and variety thus making it a big data problem. Hence, the challenge is how to combine and analyse such data with historical patient data to obtain meaningful diagnoses suggestions within acceptable time frames (considering quality of service (QoS)). Despite the evolution of big data processing technologies (e.g. Hadoop) and scalable infrastructure (e.g. clouds), there remains a significant gap in the areas of heterogeneous data collection, real-time patient monitoring, and automated decision support (semantic reasoning) based on well-defined QoS constraints. In this study, the authors review the state-of-the-art in enabling QoS for remote health care applications. In particular, they investigate the QoS challenges required to meet the analysis and inferencing needs of such applications and to overcome the limitations of existing big data processing tools.

IReHMo: An efficient IoT-based remote health monitoring system for smart regions

The ageing population worldwide is constantly rising, both in urban and regional areas. There is a need for IoT-based remote health monitoring systems that take care of the health of elderly people without compromising their convenience and preference of staying at home. However, such systems may generate large amounts of data. The key research challenge addressed in this paper is to efficiently transmit healthcare data within the limit of the existing network infrastructure, especially in remote areas. In this paper, we identified the key network requirements of a typical remote health monitoring system in terms of real-time event update, bandwidth requirements and data generation. Furthermore, we studied the network communication protocols such as CoAP, MQTT and HTTP to understand the needs of such a system, in particular the bandwidth requirements and the volume of generated data. Subsequently, we have proposed IReHMo - an IoT-based remote health monitoring architecture that efficiently delivers healthcare data to the servers. The CoAP-based IReHMo implementation helps to reduce up to 90% volume of generated data for a single sensor event and up to 56% required bandwidth for a healthcare scenario. Finally, we conducted a scalability analysis to determine the feasibility of deploying IReHMo in large numbers in regions of north Sweden.

Complex activity recognition using context driven activity theory in home environments

This paper proposes a context driven activity theory (CDAT) and reasoning approach for recognition of concurrent and interleaved complex activities of daily living (ADL) which involves no training and minimal annotation during the setup phase. We develop and validate our CDAT using the novel complex activity recognition algorithm on two users for three weeks. The algorithm accuracy reaches 88.5% for concurrent and interleaved activities. The inferencing of complex activities is performed online and mapped onto situations in near real-time mode. The developed systems performance is analyzed and its behavior evaluated.

Cloudsimdisk: energy-aware storage simulation in cloudsim

The cloud computing paradigm is continually evolving, and with it, the size and the complexity of its infrastructure. Assessing the performance of a cloud environment is an essential but an arduous task. Further, the energy consumed by data centers is steadily increasing and major components such as the storage systems need to be more energy efficient. Cloud simulation tools have proved quite useful to study these issues. However, these simulation tools lack mechanisms to study energy efficient storage in cloud systems. This paper contributes in the area of cloud computing by extending the widely used cloud simulator CloudSim. In this paper, we propose CloudSimDisk, a scalable module for modeling and simulation of energy-aware storage in cloud systems. We show how CloudSimDisk can be used to simulate energy-aware storage, and can be extended to study new algorithms for energy-awareness in cloud systems. Our simulation results proved to be in accordance with the analytical models that were developed to model energy consumption of hard disk drives in cloud systems. The source code of CloudSimDisk is also made available for the research community for further testing and development.

Opportunistic Data Collection for IoT-Based Indoor Air Quality Monitoring

Opportunistic sensing advance methods of IoT data collection using the mobility of data mules, the proximity of transmitting sensor devices and cost efficiency to decide when, where, how and at what cost collect IoT data and deliver it to a sink. This paper proposes, develops, implements and evaluates the algorithm called CollMule which builds on and extends the 3D kNN approach to discover, negotiate, collect and deliver the sensed data in an energy- and cost-efficient manner. The developed CollMule software prototype uses Android platform to handle indoor air quality data from heterogeneous IoT devices. The CollMule evaluation is based on performing rate, power consumption and CPU usage of single algorithm cycle. The outcomes of these experiments prove the feasibility of CollMule use on mobile smart devices.

BayesForSG: a bayesian model for forecasting thermal load in smart grids

Forecasting the thermal load demand for residential buildings assists in optimizing energy production and developing demand response strategies in a smart grid system. However, the presence of a large number of factors such as outdoor temperature, district heating operational parameters, building characteristics and occupant behaviour, make thermal load forecasting a challenging task. This paper presents an efficient model for thermal load forecast in buildings with different variations of heat load consumption across both winter and spring seasons using a Bayesian Network. The model has been validated by utilizing the realistic district heating data of three residential buildings from the district heating grid of the city of Skellefteå, Sweden over a period of four months. The results from our model show that the current heat load consumption and outdoor temperature forecast have the most influence on the heat load forecast. Further, our model outperforms state-of-the-art methods for heat load forecasting by achieving a higher average accuracy of 77.97% by utilizing only 10% of the training data for a forecast horizon of 1 hour.

ALPINE: A Bayesian System for Cloud Performance Diagnosis and Prediction

Cloud performance diagnosis and prediction is a challenging problem due to the stochastic nature of the cloud systems. Cloud performance is affected by a large set of factors such as virtual machine types, regions, workloads, wide area network delay and bandwidth. Therefore, necessitating the determination of complex relationships between these factors. The current research in this area does not address the challenge of modeling the uncertain and complex relationships between these factors. Further, the challenge of cloud performance prediction under uncertainty has not garnered sufficient attention. This paper proposes, develops and validates ALPINE, a Bayesian system for cloud performance diagnosis and prediction. ALPINE incorporates Bayesian networks to model uncertain and complex relationships between several factors mentioned above. It handles missing, scarce and sparse data to diagnose and predict stochastic cloud performance efficiently. We validate our proposed system using extensive real data and show that it predicts cloud performance with high accuracy of 91.93%.