Sensor-based human activity recognition (HAR), with the ability to recognise human activities from wearable or embedded sensors, has been playing an important role in many applications including personal health monitoring, smart home, and manufacturing. The real-world, long-term deployment of these HAR systems drives a critical research question: how to evolve the HAR model automatically over time to accommodate changes in an environment or activity patterns. This paper presents an online continual learning (OCL) scenario for HAR, where sensor data arrives in a streaming manner which contains unlabelled samples from already learnt activities or new activities. We propose a technique, OCL-HAR, making a real-time prediction on the streaming sensor data while at the same time discovering and learning new activities. We have empirically evaluated OCL-HAR on four third-party, publicly available HAR datasets. Our results have shown that this OCL scenario is challenging to state-of-the-art continual learning techniques that have significantly underperformed. Our technique OCL-HAR has consistently outperformed them in all experiment setups, leading up to 0.17 and 0.23 improvements in micro and macro F1 scores.

Sensor-driven systems often need to map sensed data into meaningfully labelled activities to classify the phenomena being observed. A motivating and challenging example comes from human activity recognition in which smart home and other datasets are used to classify human activities to support applications such as ambient assisted living, health monitoring, and behavioural intervention. Building a robust and meaningful classifier needs annotated ground truth, labelled with what activities are actually being observed—and acquiring high-quality, detailed, continuous annotations remains a challenging, time-consuming, and error-prone task, despite considerable attention in the literature. In this article, we use knowledge-driven ensemble learning to develop a technique that can combine classifiers built from individually labelled datasets, even when the labels are sparse and heterogeneous. The technique both relieves individual users of the burden of annotation and allows activities to be learned individually and then transferred to a general classifier. We evaluate our approach using four third-party, real-world smart home datasets and show that it enhances activity recognition accuracies even when given only a very small amount of training data.

Human activity recognition plays a significant role in enabling pervasive applications as it abstracts low-level noisy sensor data into high-level human activities, which applications can respond to. With more and more activity-aware applications deployed in real-world environments, a research challenge emerges-discovering and learning new activities that have not been pre-defined or observed in the training phase. This paper tackles this challenge by proposing a hierarchical mixture of directional statistical models. The model supports incrementally, continuously updating the activity model over time with the reduced annotation effort and without the need for storing historical sensor data. We have validated this solution on four publicly available, third-party smart home datasets, and have demonstrated up to 91.5 percent accuracies of detecting and recognising new activities.

We have witnessed an increasing number of activity-aware applications being deployed in real-world environments, including smart home and mobile healthcare. The key enabler to these applications is sensor-based human activity recognition; that is, recognising and analysing human daily activities from wearable and ambient sensors. With the power of machine learning we can recognise complex correlations between various types of sensor data and the activities being observed. However the challenges still remain: (1) they often rely on a large amount of labelled training data to build the model, and (2) they cannot dynamically adapt the model with emerging or changing activity patterns over time. To directly address these challenges, we propose a Bayesian nonparametric model, i.e. Dirichlet process mixture of conditionally independent von Mises Fisher models, to enable both unsupervised and semi-supervised dynamic learning of human activities. The Bayesian nonparametric model can dynamically adapt itself to the evolving activity patterns without human intervention and the learning results can be used to alleviate the annotation effort. We evaluate our approach against real-world, third-party smart home datasets, and demonstrate significant improvements over the state-of-the-art techniques in both unsupervised and supervised settings.

Networked systems form the backbone of modern society, underpinning critical infrastructures such as electricity, water, transport and commerce, and other essential services (e.g., information, entertainment, and social networks). It is almost inconceivable to contemplate a future without even more dependence on them. Indeed, any unavailability of such critical systems is - even for short periods - a rather bleak prospect. However, due to their increasing size and complexity, they also require some means of autonomic formation and self-organization. This paper identifies the design principles and open research issues in the twin fields of self-organization and resilience for networked systems. In combination, they offer the prospect of combating threats and allowing essential services that run on networked systems to continue operating satisfactorily. This will be achieved, on the one hand, through the (self-)adaptation of networked systems and, on the other hand, through structural and operational resilience techniques to ensure that they can detect, defend against, and ultimately withstand challenges.

Experiments in assisted living confirm that such systems can provide context-aware services that enable occupants to remain active and independent. They also demonstrate that abnormal sensor events hamper the correct identification of critical (and potentially life-threatening) situations, and that existing learning, estimation, and time-based approaches are inaccurate and inflexible when applied to multiple people sharing a living space. We propose a technique that integrates the semantics of sensor readings with statistical outlier detection. We evaluate the technique against four real-world datasets that include multiple individuals, and show consistent rates of anomaly detection across different environments.

Summary Rising complexity within multi‐tier computing architectures remains an open problem. As complexity increases, so do the costs associated with operating and maintaining systems within these environments. One approach for addressing these problems is to buildself‐healingsystems (i.e. frameworks) that can autonomously detect and recover from faulty states. Self‐healing systems often combine machine learning techniques with closed control loops to reduce the number of situations requiring human intervention. This is particularly useful in situations where human involvement is both costly to develop, and a source of potential faults. Therefore, a survey of self‐healing frameworks and methodologies in multi‐tier architectures is provided to the reader. Uniquely, this study combines an overview of the state of the art with a comparative analysis of the computing environment, degree of behavioural autonomy, and organisational requirements of these approaches. Highlighting these aspects provides for an understanding of the different situational benefits of these self‐healing systems. We conclude with a discussion of potential and current research directions within this field. Copyright © 2014 John Wiley & Sons, Ltd.

Abstract As ubiquitous computing becomes a reality, its applications are increasingly being used in business-critical, mission-critical and even in safety-critical, areas. Such systems must demonstrate an assured level of correctness. One approach to the exhaustive analysis of the behaviour of systems isformal verification, whereby each important requirement is logically assessed against all possible system behaviours. While formal verification is often used in safety analysis, it has rarely been used in the analysis of deployed pervasive applications. Without such formality it is difficult to establish that the system will exhibit the correct behaviours in response to its inputs and environment. In this paper, we show how model-checking techniques can be applied to analyse the probabilistic behaviour of pervasive systems. As a case study we apply this technique to an existing pervasive message-forwarding system,Scatterbox. Scatterbox incorporates many typical characteristics of pervasive systems, such as dependence on sensor reliability and dependence on context. We assess the dynamic temporal behaviour of the system, including the analysis of probabilistic elements, allowing us to verify formal requirements even in the presence of uncertainty in sensors. We also draw some tentative conclusions concerning the use of formal verification for pervasive computing in general.

Recognising high-level human activities from low-level sensor data is a crucial driver for pervasive systems that wish to provide seamless and distraction-free support for users engaged in normal activities. Research in this area has grown alongside advances in sensing and communications, and experiments have yielded sensor traces coupled with ground truth annotations about the underlying environmental conditions and user actions. Traditional machine learning has had some success in recognising human activities; but the need for large volumes of annotated data and the danger of overfitting to specific conditions represent challenges in connection with the building of models applicable to a wide range of users, activities, and environments. We present USMART, a novel unsupervised technique that combines data- and knowledge-driven techniques. USMART uses a general ontology model to represent domain knowledge that can be reused across different environments and users, and we augment a range of learning techniques with ontological semantics to facilitate the unsupervised discovery of patterns in how each user performs daily activities. We evaluate our approach against four real-world third-party datasets featuring different user populations and sensor configurations, and we find that USMART achieves up to 97.5% accuracy in recognising daily activities.

Wireless Sensor Network (WSN) technology has recently moved out of controlled laboratory settings to real-world deployments. Many of these deployments experience high rates of failure. Common types of failure include node failure, link failure, and node reboot. Due to the resource constraints of sensor nodes, existing techniques for fault detection in enterprise networks are not applicable. Previously proposed WSN fault detection algorithms either rely on periodic transmission of node status data or inferring node status based on passive information collection. The former approach significantly reduces network lifetime, while the latter achieves poor accuracy in dynamic or large networks. Herein, we propose Sequence-Based Fault Detection (SBFD), a novel framework for network fault detection in WSNs. The framework exploits in-network packet tagging using the Fletcher checksum and server-side network path analysis to efficiently deduce the path of all packets sent to the sink. The sink monitors the extracted packet paths to detect persistent path changes which are indicative of network failures. When a failure is suspected, the sink uses control messages to check the status of the affected nodes. SBFD was implemented in TinyOS on TelosB motes and its performance was assessed in a testbed network and in TOSSIM simulation. The method was found to achieve a fault detection accuracy of 90.7% to 95.0% for networks of 25 to 400 nodes at the cost of 0.164% to 0.239% additional control packets and a 0.5% reduction in node lifetime due to in-network packet tagging. Finally, a comparative study was conducted with existing solutions.

Wireless sensor networks (WSN) show enormous potential for collection and analysis of physical data in real-time. Many papers have proposed methods for improving the network reliability of WSNs. However, real WSN deployments show that sensor data-faults are very common. Several server-side data reliability techniques have been proposed to detect these faults and impute missing or erroneous data. Typically, these techniques reduce the lifetime of the network due to redundant data transmission, increase latency, and are computation and storage intensive. Herein, we propose Packet-Level Attestation (PLA), a novel framework for sensor data reliability assessment. It exploits the spatial correlation of data sensed at nearby sensors. The method does not incur additional transmission of control message between source and sink; instead, a verifier node sends a validation certificate as part of the regular data packet. PLA was implemented in TinyOS on TelosB motes and its performances was assessed. Simulations were performed to determine its scalability. It incurs only an overhead of 1.45% in terms of packets transmitted. Fault detection precision of the framework varied from 100% to 99.48%. Comparisons with existing methods for data reliability analysis showed a significant reduction in data transmission, prolonging the network lifetime.

Wireless sensor networks (WSNs) are highly resource constrained in terms of power supply, memory capacity, communication bandwidth, and processor performance. Compression of sampling, sensor data, and communications can significantly improve the efficiency of utilization of three of these resources, namely, power supply, memory and bandwidth. Recently, there have been a large number of proposals describing compression algorithms for WSNs. These proposals are diverse and involve different compression approaches. It is high time that these individual efforts are put into perspective and a more holistic view taken. In this article, we take a step in that direction by presenting a survey of the literature in the area of compression and compression frameworks in WSNs. A comparative study of the various approaches is also provided. In addition, open research issues, challenges and future research directions are highlighted.

Resource pools are collections of computational resources (e.g., servers) which can be used by different applications in a shared way. A crucial aspect in these pools is to allocate resources so as to ensure their proper usage, taking into account workload and specific requirements of each application. An interesting approach, in this context, is to allocate the resources in the best possible way, aiming at optimal resource usage. Workload, however, varies over time, and in turn, resource demands will vary too. To ensure that optimal resource usage is always in place, resource shares should be defined dynamically and over time. It has been claimed that utility functions are the main tool for enabling such self-optimizing behavior. Whereas many solutions with this characteristic have been proposed to date, none of them presents true decentralization within the context of shared pools. In this article, we then propose a decentralized model for optimal resource usage in shared resource pools, providing practical and theoretical evidence of its feasibility.

Both modularity and loose-coupling properties inherent to the self-adaptive systems offer the opportunity for ad-hoc service compositions, dynamic change and adaptation. To provide such a dynamic and self-adapting behavior, developers emphasize special self-management policies. ASSL (Autonomic System Specification Language) is a formal tool where such policies might be formally specified, validated and implemented. Intrinsically, the ASSL-developed policies are very strict and may impose quite restrictive behavior, which sometimes is undesirable. To solve the problem, we are currently developing special mechanisms for ASSL that help to specify policies that might evolve in order to satisfy system goals changing in the course of system adaptation. This paper presents our work on a mechanism imposing special loose self-management policies introducing flexibility into the self-adapting behavior.

Shared resource pools are facilities featuring a certain amount of resources which can be used by different applications. For managing resources in such pools, the demand of each application can be used. Such a demand, however, is driven by the workload, which varies over time. For that reason, adaptive approaches have been proposed for the management of shared resources pools. Whereas a number of solutions exist in this context, they are either not truly decentralized or do not apply to the problem we are dealing with. In this paper, we then present Darma, an approach for managing shared resource pools in a truly decentralized, adaptive, and optimal way.

Highly sensorised systems present two parallel challenges: how to design a sensor suite that can efficiently and cost-effectively support the needs of given services; and to extract the semantically relevant interpretations, or ldquosituationsrdquo, from the flood of context data collected by the sensors. We describe mathematical structures called situation lattices that can be used to address these two problems simultaneously, allowing designers to both design and refine situation identification whilst offering insights into the design of sensor suites. We validate the accuracy and efficiency of our technique against a third-party data set and demonstrate how it can be used to evaluate sensor suite designs.

Most research in the area of smart environments focuses on improving the accuracy with which human activities can be recognised. Relatively little research has been done into how designers can gain insights into the behaviours their systems are observing, and feed these insights back into improving systems design. We describe a mathematical structure, the situation lattice, and show how it can be used to discover knowledge about activities and the way in which they can be sensed. We show how this knowledge can be used to improve activity recognition, using the example of a real-world smart home data set.

We present a solid study on the performance of a homological sensor network in partial sensing coverage, which means the network has at least one sensing coverage hole and we demonstrate that when sacrificing a little coverage the system lifetime can be prolonged significantly. In particular, we showed that when there is one sensing coverage hole (with a coverage rate of 97%) the system lifetime can be extended to 3-7 times compared with a full coverage strategy which gives a system lifetime increase with 1.2-3 times only. An algebraic topology tool, homology group, is used in our work to calculate sensing coverage of a sensor network. Unlike other approaches, our method does not need any node location or orientation information and it does not have any assumption about the code deployment control and domain geometry either. The only thing needed to calculate sensing coverage is a node to node communication graph.

Mobile ad hoc networks (MANETs) generally adopt a peer-to-peer architecture in which the nodes themselves provide routing and services to the network. Disconnectivity with peer nodes, induced by mobility, power drains and damage makes route maintenance difficult and degrade the network's ability to offer services reliably to its peer nodes. In this paper, we present a routing scheme for proactive management of such disconnections, by fusing and leveraging information derived from multiple levels of the network protocol stack using cross-layering. In addition to the disconnectivity information, this routing scheme utilises node's service level information and data/service replication to provide service from an alternate source (if there is one) even in the absence of the targeted source. Simulation results demonstrate significant improvements in route maintenance and service availability over other similar schemes.

BACKGROUND: Various measures of semantic similarity of terms in bio-ontologies such as the Gene Ontology (GO) have been used to compare gene products. Such measures of similarity have been used to annotate uncharacterized gene products and group gene products into functional groups. There are various ways to measure semantic similarity, either using the topological structure of the ontology, the instances (gene products) associated with terms or a mixture of both. We focus on an instance level definition of semantic similarity while using the information contained in the ontology, both in the graphical structure of the ontology and the semantics of relations between terms, to provide constraints on our instance level description.Semantic similarity of terms is extended to annotations by various approaches, either though aggregation operations such as min, max and average or through an extrapolative method. These approaches introduce assumptions about how semantic similarity of terms relates to the semantic similarity of annotations that do not necessarily reflect how terms relate to each other. RESULTS: We exploit the semantics of relations in the GO to construct an algorithm called SSA that provides the basis of a framework that naturally extends instance based methods of semantic similarity of terms, such as Resnik's measure, to describing annotations and not just terms. Our measure attempts to correctly interpret how terms combine via their relationships in the ontological hierarchy. SSA uses these relationships to identify the most specific common ancestors between terms. We outline the set of cases in which terms can combine and associate partial order constraints with each case that order the specificity of terms. These cases form the basis for the SSA algorithm. The set of associated constraints also provide a set of principles that any improvement on our method should seek to satisfy. CONCLUSION: We derive a measure of semantic similarity between annotations that exploits all available information without introducing assumptions about the nature of the ontology or data. We preserve the principles underlying instance based methods of semantic similarity of terms at the annotation level. As a result our measure better describes the information contained in annotations associated with gene products and as a result is better suited to characterizing and classifying gene products through their annotations.

Construct differs from other pervasive systems platforms in a number of key respects. It is completely standards-based, using RDF as its data exchange model and ZeroConf for resource discovery. It supports a knowledge-centric model of interaction where clients' actions are driven by queries and triggers about the context of the system. It uses gossiping to maintain a consistent state across a distributed data structure, which maximises robustness and scalability and avoids many problems with hot-spots and hot-paths in communications. Finally, it treats all information sources uniformly as sensors acting as inputs to uncertain reasoning algorithms.

Layered architectures are not sufficiently flexible to cope with the dynamics of wireless-dominated next generation communications. Cross-layer approaches may provide a better solution: allowing interactions between two or more non-adjacent layers in the protocol stack. Cross-layer architectures based on purely local information will not be able to support system-wide cross-layer performance optimization, context-awareness, etc. A new cross-layer architecture which provides a hybrid local and global view, using gossiping to maintain consistency has been proposed in [1]. This paper studies the possibilities of context-awareness in communications through this architecture by two examples. The first example uses user-centric context to control the available link-bandwidth and satisfy user accordingly. The second uses contextual information to control the transmission power of a mobile node.

Control systems must increasingly be designed to involve collections of hardware and software components, both of which may evolve over the lifetime of the system, and which are expected to provide self-managing, adaptive, autonomic behaviour. Understanding the behaviour such a system will exhibit under any specific conditions is a significant design challenge. We present a model derived from approaches to modelling dynamical systems in which the adaptive behaviour of an autonomic system may be described and analysed as a whole. We explain our ideas with reference to a hybrid hardware/software system, and argue that it generalises to other classes of autonomic systems.

Autonomic communications seek to improve the ability of network and services to cope with unpredicted change, including changes in topology, load, task, the physical and logical characteristics of the networks that can be accessed, and so forth. Broad-ranging autonomic solutions require designers to account for a range of end-to-end issues affecting programming models, network and contextual modeling and reasoning, decentralised algorithms, trust acquisition and maintenance---issues whose solutions may draw on approaches and results from a surprisingly broad range of disciplines. We survey the current state of autonomic communications research and identify significant emerging trends and techniques.

An executing thread, in an object oriented programming language, is spawned, directly or indirectly, by a main process. This in turn gets its instructions from a primary class. In Java there is no close coupling of a thread and the objects from which they were created. The use of a container abstraction allows us to group threads and their respective objects into a single structure. A container that holds threads whose variables are all housed within the container is a perfect candidate for strong migration. To achieve this we propose a combination of three techniques to allow the containers to migrate in a manner that approaches strong mobility yet does not resort to retaining bindings to resources across distant and unreliable networks.