Accurately detecting passenger traffic flow in public transportation, e.g., for the purpose of identifying congested stations, and detecting anomalies in that flow, are both important for enhancing passenger satisfaction and safety. However, current methods for traffic flow detection and prediction train their models in offline modes using potentially outdated data, which can’t be refreshed with new data. In this paper, we introduce a two-step, continuously updating framework aimed at real-time prediction of anomalies within transportation networks, which we name Adaptive Time-Aware Anomaly Detection Approach (ATADA). Specifically, the first step filters out regular traffic patterns and balances the data set, while the second step employs a sequence-to-sequence attention model, a type of deep learning model, to further detect the traffic anomalies. We propose a dynamic online time-aware learning mechanism which enables our models to continuously train on incoming data and to adapt predictive strategies based on the most recent traffic patterns. The proposed method is validated using real subway AFC data from Suzhou, China and Hangzhou, China. Experimental results demonstrate that our framework significantly improves efficiency while maintaining high accuracy in real-time traffic anomaly detection.

Mobility pattern recognition is a central aspect of transportation and data mining research. Despite the development of various machine learning techniques for this problem, most existing methods face challenges such as reliance on handcrafted features (e.g., user has to specify a feature such as “travel time”) or issues with data imbalance (e.g., fewer older travelers than commuters). In this paper, we introduce a novel Data Balancing Generative Adversarial Network (DBGAN), which is a specifically designed attention mechanism-based GAN model to address these challenges in mobility pattern recognition. DBGAN captures both static (e.g., travel locations) and dynamic (e.g., travel times) features of different passenger groups, and avoids using handcrafted features that may result in information loss, based on a sequence-to-image embedding method. Our model is then applied to overcome the data imbalance issue and perform mobility pattern recognition. We evaluate the proposed method on real-world public transportation smart card data from Suzhou, China, and focus on recognizing two different passenger groups: older people and students. The results of our experiments demonstrate that DBGAN is able to accurately identify the different passenger groups in the data, with the detected mobility patterns being consistent with the ground truth. These results highlight the effectiveness of DBGAN in overcoming data imbalance in mobility pattern recognition, and demonstrate its potential for wider use in transportation and data mining applications.

Network caching in Vehicular Named Data Networks (VNDN) has the potential to support latency-sensitive services, though given the massive amount of content generated by vehicles in a VNDN, it is challenging to cache sufficiently diverse content across the network for an acceptable hit ratio. Redundant interest transmissions can negatively impact network overhead, wasting network resources when retrieving duplicate content. In the literature, a caching decision is typically based on ongoing popularity, computed from a vehicle’s experience with requests received for individual content. This leads to each vehicle caching the most popular content. Instead, we group content into different categories. All content in a popular category of vehicular service (e.g., tourist information category) is cached equally, not just the most popular content within a category (e.g., information relating to a particular museum). This increases content diversity in the cache. Further, we posit that a service requester from a particular location at a specific time would be interested in the same category of service/application as other users from the same location/time. Leveraging this observation, we consider both forthcoming spatial/temporal and ongoing popularity of vehicular services when ranking their popularity. We also address network overhead with a mobility model that captures the communication duration between vehicles. This parameter, which has been overlooked in existing literature, enables the selection of an appropriate forwarder for content requests. We use ndnSIM VNDN with a mobility trace of Luxembourg city for simulation. Compared to an approach where individual content-wise popularity is computed, our scheme shows more than 100% improvement in cache hit ratio where the number of contents in the network varied between 104 and 105. Simultaneously, we also enhanced the diversity of content caching, while considering the trade-off between cache hit ratio and network overhead.

Multi-access Edge Computing (MEC) enables IoT applications to place their services in the edge servers of mobile networks, balancing Quality-of-Service (QoS) and energy-efficiency. Previous works consider compute requirements, while the IoT and latency/bandwidth per-flow communicate requirements are largely ignored. Moreover, the Smart City domain presents unique challenges – modeling the Urban Smart Things (USTs – urban IoT clients), their connectivity with MEC network, diverse resource requirements (compute, communicate, and IoT) of application services, modeling the federation of multiple MEC providers in a city, which we consider in this article. To address these research gaps, we propose: i)UrbanEnQoSMDP– formulation for energy and QoS (latency) optimized service placement for a set of applications in the ‘Urban IoT-Federated MEC-Cloud’ architecture to satisfy applications’ compute, per-flow communicate, and IoT requirements; ii)‘’$\epsilon$ε-greedy with mask”policy for apriori satisfaction of IoT requirements by shortlisting suitable USTs; iii)UrbanEnQoSPlace– proposed multi-action Deep Reinforcement Learning (DRL) model, designed from Dueling Deep-Q Network, that uses the proposed policy to solve the UrbanEnQoSMDP for simultaneously placing all services of an application. Extensive simulation results illustrate efficacy and scalability of proposed model against state-of-the-art DRL algorithms (better convergence, higher rewards, lesser runtime; proposed policy w.r.t fewer violations).

Smart cities generate huge volumes of data to be processed by applications with different criticality and requirements. For example, a healthcare application needs lower latency when requested from an ambulance travelling to a hospital during an emergency compared to applications in less-critical domains. Cities can use Multi-access Edge Computing to reduce latency by placing applications’ services closer to users. A service placement process selects the set of servers to run the services for deployment. Smart cities challenge this selection as a large number of servers and services generate a large number of potential solutions with different QoS properties. Additionally, placement approaches must consider applications’ criticality and users’ mobility to offer an appropriate overall latency. Current approaches have considered servers’ utilisation and users’ location to place services. However, they do not consider applications’ criticality and mobile users’ paths. This paper presents MAACO, a Mobility-Aware, priority-driven, ACO-based service placement model that prioritises applications according to their criticality and minimises critical applications’ latency, while considering predicted paths for mobile users. Evaluation results show that MAACO achieves lower latency and waiting time compared against baselines at the cost of reduced load balance between the network servers.

City services are frequently supported by software services that are managed by service-oriented architectures. However, a large number of software services is likely to cause performance issues when discovering software services. The distributed organisation of services information improves discovery performance. Existing research proposes to organise services information according to service location, domains, or city context, keeping that organisation constant under an assumption that cities do not change. However, cities are dynamic environments where entities interact, causing events that in turn, effect changes in the city. The organisation of services information must evolve or it will become outdated, negatively impacting discovery performance. We propose a self-adaptive service model for smart cities to support service discovery. This model adapts the organisation of services information according to city events. We introduce a self-adaptive architecture that keeps track of the discovery metrics and moves information about services between registries to maintain the discovery efficiency. We evaluate the proposed model in simulated environments and a real IoT testbed. Results show that our model outperforms competitors when reactive adaptation is triggered by a specific event. However, proactive adaptation needs further research. Results from the real IoT testbed present the costs of the proposed model.

The Internet of Things envisions billions of physical devices connecting over the Internet to provide a near real-time view of the state of the world. These devices' capabilities can be abstracted as IoT services and provided on demand. To enable quality-aware service provision, Service Level Agreements (SLA) are widely used as legally binding contracts to obligate service providers to comply with a pre-negotiated Quality of Service (QoS). With a possible ever-increasing number of service providers in an IoT environment, multi-bilateral SLA negotiation is likely to be prohibitively time-consuming without an a-priori process to select trusted candidate providers with whom to negotiate. In this article, a trust model is proposed to identify trusted service providers in a dynamic IoT environment before attempting to negotiate an SLA. A trust credit that indicates both the SLA’s fulfillment and the possible negotiation success rate is derived based on historical information relating to a service’s previous negotiations and its monitored run-time performance. Indiscernibility analysis in Rough Set theory is used to predict the negotiation success rate, while Bayesian inference is applied to deduce the possibility of SLA violation according to the monitored data. The simulation results demonstrate the feasibility and efficiency of the proposed trust model.

Automatic planning, with dynamic binding and adaptive composition recovery, has been used to tackle complex service provisioning in mobile environments, but given frequent network topology changes, and services with time-dependent QoS, finding composites that can functionally and non-functionally satisfy a user's request remains difficult. Many service composition mechanisms either require a centralised perspective of the environment, or use optimisation mechanisms that trade off computational efficiency for optimality. Stigmergy-based approaches have been used to model decentralised service interactions between service providers, using a community of mobile software agents that share the same goal to approximate the set of QoS-optimal service compositions. Inspired by this model, this article addresses computational efficiency concerns using a collaborative approach to engage multiple communities of agents for provisioning QoS-optimal service compositions in mobile environments. New compositions can emerge from local decisions and interactions with agents from diverse communities. We assess whether having multiple communities improves the diversity and optimality of solutions. We also measure the proposed approach’ efficiency in dealing with incomplete information. The results show that the proposed approach trades optimality for a more diverse set of solutions, at a cost of higher overhead.

Accurate short-term forecasts allow dynamic systems to adapt their behaviour when degradation is forecast e.g., transportation forecasting allows for alternative routing of traffic before gridlock. This rationale can be applied to service-oriented computing when creating and managing service applications. Recent approaches to improve reliability in service applications have focused on reducing the time to recovery of application using collaborative filtering-based approaches to make QoS predictions for similar users. In this article, we focus on reducing the time to detection of a failure by forecasting when a service is about to degrade in quality. Previous approaches that have focused on QoS forecasting have used traditional time-series methods that are not designed for sudden peaks caused by network congestion or battery-powered IoT devices that can reduce processing capabilities to extend battery life. More modern recurrent neural network-based approaches such as GRUs and LSTMs have long training times, which are unsuitable for dynamic environments. We propose a noisy echo state network-based approach that has been designed to reduce training time allowing the model to incorporate recent QoS values on devices at the edge. Our results show increased response time forecasting accuracy compared to state of the art approaches when tested on IoT and web services datasets.

Pervasive applications in smart cities rely on a large number of IoT devices, which are deployed in large geographic areas. Smart cities can manage these devices using Service-Oriented Architectures (e.g., micro-services) by encapsulating devices capabilities as IoT services. Distributed service discovery architectures reduce search spaces and perform discovery processes closer to consumers on edge devices. However, request management, a key task in distributed service discovery, is still challenging because requests must be forwarded through large networks where nodes have partial knowledge about other participants. Previous research has shown that social-based and bio-inspired methods can be used to manage requests in small-scale environments, but such approaches do not scale to large environments. This paper adds urban context to a social-based and bio-inspired mechanism to forward requests where they are most likely to be solved. Results show that our model has the best rate of solved requests, and intermediate latency.

This document introduces the artifacts that implement the request manager proposed in the paper "An Urban-driven Service Request Management Model". The artifacts can be found in the public TCD GitLab project percom2020-srmm 1 .

Multiaccess edge computing (MEC) will strengthen forthcoming 5G networks by improving the Quality of Service (QoS), in particular, reducing latency, increasing data processing rates, and providing real-time information to develop high-value Internet-of-Things (IoT) services. To enable data-intensive network services and support advanced analytics, many network operators have proposed to integrate MEC systems with network function virtualization (NFV) consolidating virtual network functions (VNFs) and edge capabilities on a shared infrastructure. As of yet, this integration is not fully established, with various architectural issues currently open, even at standardization level. For instance, any update to VNFs deployed in a MEC system requires a time-consuming manual effort, which affects the overall infrastructure operations. To address these pitfalls, VNFs can be decomposed into microservices, which maintain their own states and exhibit different resource consumption requirements. This article presents an approach to integration that leverages serverless computing to merge MEC and NFV at the system level and to deploy VNFs on demand, by combining MEC functional blocks with an NFV orchestrator using a Kubernetes cluster. We further investigate whether the resource utilization of a MEC system can be improved by leveraging networked FPGA-enabled MEC servers, through an extension of the edge layer that takes advantage of available programmable hardware. We quantitatively evaluate and demonstrate the improvement of 75% end-to-end latency, 99.96% VNF execution time, 26.9% resource utilization, and 15.8% energy consumption in comparison with traditional baselines of cloud, edge, and serverless-edge test cases for a high-definition real-time video streaming application.

A variety of self-tracking applications and devices have been developed in recent years to support users in tracking their weight, calories eaten, physical activities, sleep and productivity. The availability of all this data from multiple streams provides a rich environment for experimentation that allows users to improve certain aspects of their lives such as losing weight, getting better sleep or being more productive. In this paper we propose a framework that guides users to define, track, analyse, improve and control goals for better personal productivity. We present the outcome of a single-subject case study that was implemented over one year based on the proposed framework for academic productivity. This pilot study demonstrates how longitudinal multistream self-tracking data can be leveraged to gain actionable insights into personal productivity.

Multi-agent Reinforcement Learning (RL) is frequently used in large-scale autonomous systems to learn the behaviours that best suit the system's operating environment. Learning can take a significant amount of time during which an RL system's performance is necessarily suboptimal. Transfer learning (TL), a method of reusing knowledge which has been gained in one task to improve the performance in another, has been used to speed up learning in single RL agent systems. TL requires learning on a source task to complete before transferring it to a target task, i.e., transfer is done offline. Parallel Transfer Learning (PTL), a technique which enables the source and target tasks to run concurrently, has been proposed to enable online transfers. However, the online selection of knowledge to be transferred, as well as online ways of integration of that knowledge on the receiving agents remain open issues. This paper proposes methods for selecting the knowledge to be transferred in PTL, frequency and size of transfers, and methods for knowledge integration into the target task. We evaluate the proposed approaches in two canonical RL examples: Mountain Car and Co-operative Predator Prey Pursuit. We show that PTL, similarly to RL, is highly sensitive to parameter selection and that suitable parameters differ per scenario.

In service-oriented architectures, collaborative filtering is a key technique for service recommendation based on QoS prediction. Matrix factorisation has emerged as one of the main approaches for collaborative filtering as it can handle sparse matrices and produces good prediction accuracy. However, this process is resource-intensive and training must take place in the cloud, which can lead to a number of issues for user privacy and being able to update the model with new QoS information. Due to the time-varying nature of QoS it is essential to update the QoS prediction model to ensure that it is using the most recent values to maintain prediction accuracy. The request time, which is the time for a middleware to submit a user's information and receive QoS metrics for a candidate services is also important due to the limited time during dynamic service adaptations to choose suitable replacement services. In this paper we propose a stacked autoencoder with dropout on a deep edge architecture and show how this can be used to reduce training and request time compared to traditional matrix factorisation algorithms, while maintaining predictive accuracy. To evaluate the accuracy of the algorithms we compare the actual and predicted QoS values using standard error metrics such as MAE and RMSE. In addition, we propose an alternative evaluation technique using the predictions as part of a service composition and measuring the impact that the predictions have on the response time and throughput of the final composition. This more clearly shows the direct impact that these algorithms will have in practice.

The proliferation of Internet of Things (IoT) and the success of resource-rich cloud services have pushed the data processing horizon towards the edge of the network. This has the potential to address bandwidth costs, and latency, availability and data privacy concerns. Serverless computing, a cloud computing model for stateless and event-driven applications, promises to further improve Quality of Service (QoS) by eliminating the burden of always-on infrastructure through ephemeral containers. Open source serverless frameworks have been introduced to avoid the vendor lock-in and computation restrictions of public cloud platforms and to bring the power of serverless computing to on-premises deployments. In an IoT environment, these frameworks can leverage the computational capabilities of devices in the local network to further improve QoS of applications delivered to the user. However, these frameworks have not been evaluated in a resource-constrained, edge computing environment. In this work we evaluate four open source serverless frameworks, namely, Kubeless, Apache OpenWhisk, OpenFaaS, Knative. Each framework is installed on a bare-metal, single master, Kubernetes cluster. We use the JMeter framework to evaluate the response time, throughput and success rate of functions deployed using these frameworks under different workloads. The evaluation results are presented and open research opportunities are discussed.

Agents frequently collaborate to achieve a shared goal or to accomplish a task that they cannot do alone. However, collaboration is difficult in open multi-agent systems where agents share constrained resources to achieve both individual and shared goals. In current approaches to collaboration, agents are organised into disjoint groups and social reasoning is used to capture their capabilities when selecting a qualified set of collaborators. These approaches are not useful when agents are in multiple, overlapping groups; depend on each other when using shared resources; have multiple goals to achieve simultaneously; and have to share the overall costs and benefits. In this article, agents use social reasoning to enhance their understanding of other agents’ goals and their dependencies, and self-adaptive techniques to adapt their level of self-interest in a collaborative process, with a view to contributing to lowering shared costs or increasing shared benefits. This model aims at improving the extent to which agents’ goals are met while improving shared resource usage efficiency. For example, in a public transport system where each mode of transport has limited capacity, commuters will be enabled to make choices that avoid over-capacity in different modes, or in a smart energy grid with limited capacity, users can make choices as to when they increase their demand. The model simultaneously helps avoid overloading a shared resource while allowing users to achieve their own goals. The proposed model is evaluated in an open multi-agent system with 100 agents operating in multiple overlapping groups and sharing multiple constrained resources. The impact of agents’ varying levels of social dependencies, mobility, and their groups’ density on their individual and shared goal achievement is analysed.

Taxi-sharing is an emergent transport mode, which has shown promising results economically, by splitting the travel cost between passengers and environmentally, by serving more people in each trip. Intelligent taxi-dispatch approaches can also manage demand by distributing taxis according to population density in a city. Current approaches to taxi-sharing recommend passengers share a taxi by matching their origin and destination, and taxi-dispatch approaches simply send more taxis to populated areas. However, each passenger may have multiple preferences (e.g., level of convenience, time, cost, and environmental factors), and require a mechanism that offers options considering these preferences. Similarly, taxi drivers may have multiple preferences (e.g., number of hours to work, minimum revenue per day) that need to be considered during a taxi-dispatch planning process. This paper presents a multi-agent collaborative passenger matching and taxi-dispatch model. Passengers and drivers are modeled as autonomous agents having multiple often-conflicting preferences. Passenger agents collaboratively take actions to form a group for a taxi-share, and taxi agents collaborate to achieve a dispatch plan.

Many modern software systems and applications are built using heterogeneous services provided by a range of devices, from high-power devices located in the Cloud to potentially resource-constrained and/or mobile services from IoT devices at the edge of the network. The large growth in the number of these services has led to some functionally similar services. When selecting services, a critical criterion is Quality of Service (QoS), which includes factors such as response time, location and cost. As the value of dynamic QoS attributes vary with time, there is a need to accurately forecast future QoS values to identify if a service may be about to fail. In this paper, we propose using an LSTM-based neural network to forecast future QoS values. We evaluate the use of an LSTM network against the existing state of the art in experiments using an established web service dataset and a new dataset collected by deploying services on low power IoT devices, which we publicly release. This mixture of datasets covers the heterogeneity that would be expected in a typical IoT environment.

Smart cities provide software services to citizens that are likely to be deployed in large, dynamic, heterogeneous, and distributed environments. The discovery of these services needs to be efficient and pervasive, based on the specific context of the city, and the integration of diverse providers. We identify a trade-off between accuracy and performance in the discovery of services in this scenario. Existing research has proposed solutions that focus either on semantic methods to improve accuracy with performance negatively affected, or vice versa. Additionally, the composition of services from different sources has not been explored in smart cities and large scenarios. We propose to address the trade-off by extending both how service information is organised, and the service discovery process. Service organisation uses urban context to spread service descriptions to the right urban-places; the service discovery process uses this model to forward requests where they are more likely to be solved. We simulate our model as a network of gateways that covers Dublin city center and manages services information. Results show that our model solves more requests than previous work in a smart city environment. In addition, response time keeps acceptable even when there are 100 thousand services.

Internet of Things (IoT) applications can be built from a number of heterogeneous services provided by a range of devices, which are potentially resource constrained and/or mobile. As these services and applications continue to be more widespread, a key research question is how to predict user-side quality of service (QoS), to ensure the optimal selection, composition and adaptation of IoT services. The exponential growth in the number of these services means that it is not practical to invoke all candidate services to test their QoS, especially during runtime service adaptation. QoS can vary by time and location, which makes it difficult for service providers to give accurate estimates of how the service will perform for users located in changing network topologies. We propose IoTPredict, a novel neighbourhood-based prediction approach for the IoT, which uses an alternative similarity computation mechanism. Our collaborative approach requires no additional invocation of services, which is a key requirement for resource constrained devices in the IoT. We evaluate our algorithm on a QoS dataset and show that it achieves higher QoS prediction accuracy than other state of the art approaches.

Mobile users can form a service-sharing community within a geographic area by using their mobile devices. Finding Quality of Service (QoS) optimal service compositions in such mobile environments is challenging because of the inherent dynamism in services deployed on mobile devices. Existing service composition proposals for mobile environments either use template-matching composition or require a-priori knowledge about the QoS objectives' weights, which limits the composition flexibility in such environments. This paper introduces a QoS optimisation mechanism for planning-based service composition in mobile environments, where mobile software agents use stigmergic coordination to iteratively explore parts of the distributed service composition space to approximate a set of QoS optimal configurations. We present a mechanism that minimises the exploration of previously identified non-optimal solutions to encourage exploration of different parts of the service space. We evaluate the performance of the proposed approach and compare the results with a baseline variant, a Dijkstra-based, a Greedy and a Random approach. The results show that the proposed approach can achieve higher utility compared to the evaluated proposals at the cost of increased overhead.

Abstract Human drivers naturally adapt their behaviour depending on the traffic conditions, such as the current weather and road type. Autonomous vehicles need to do the same, in a way that is both safe and efficient in traffic composed of both conventional and autonomous vehicles. In this paper, we demonstrate the applicability of a reconfigurable vehicle controller agent for autonomous vehicles that adapts the parameters of a used car‐following model at runtime, so as to maintain a high degree of traffic quality (efficiency and safety) under different weather conditions. We follow a dynamic software product line approach to model the variability of the car‐following model parameters, context changes and traffic quality, and generate specific configurations for each particular context. Under realistic conditions, autonomous vehicles have only a very local knowledge of other vehicles' variables. We investigate a distributed model predictive controller agent for autonomous vehicles to estimate their behavioural parameters at runtime, based on their available knowledge of the system. We show that autonomous vehicles with the proposed reconfigurable controller agent lead to behaviour similar to that achieved by human drivers, depending on the context.

Mobile, pervasive computing environments respond to users’ requirements by providing access to and composition of various services over networked devices. In such an environment, service composition needs to satisfy a request’s goal, and be mobile-aware even throughout service discovery and service execution. A composite service also needs to be adaptable to cope with the environment’s dynamic network topology. Existing composition solutions employ goal-oriented planning to provide flexible composition, and assign service providers at runtime, to avoid composition failure. However, these solutions have limited support for complex service flows and composite service adaptation. This paper proposes a self-organizing, goal-driven service model for task resolution and execution in mobile pervasive environments. In particular, it proposes a decentralized heuristic planning algorithm based on backward-chaining to support flexible service discovery. Further, we introduce an adaptation architecture that allows execution paths to dynamically adapt, which reduces failures, and lessens re-execution effort for failure recovery. Simulation results show the suitability of the proposed mechanism in pervasive computing environments where providers are mobile, and it is uncertain what services are available. Our evaluation additionally reveals the model’s limits with regard to network dynamism and resource constraints.

This paper discusses the challenges in developing an IoT platform for registering, discovering and composing heterogeneous services from multiple provider types, viz., Wireless Sensor Networks (WSNs), Web Service Providers (WSPs), and Autonomous Service Providers (ASPs), without human intervention. The platform executes a service composition in a decentralised fashion, with a mechanism to detect service provider failure and fallback to previously discovered services to complete a service composition flow. We comment on technical and scientific challenges involved in managing these heterogeneous, autonomous, and resilient IoT services.

Recently, there have been a large number of proposals for IoT middleware solutions. In addition, a few recent studies have surveyed and qualitatively evaluated these IoT middleware proposals against functional and non-functional features. A quantitative evaluation is also needed to complement these existing qualitative studies and provide a more in-depth perspective of the state of the art. This paper presents a quantitative evaluation of 4 representative proposals: OpenIoT, CHOReOS, LinkSmart and UBIWARE. The evaluation results, based on a small real-life scenario, show that research is needed in the area of autonomous and scalable service registration, discovery and composition, heterogeneity, and interoperability of IoT middlewares.

Multi-agent reinforcement learning (MARL) is a widely researched technique for decentralised control in complex large-scale autonomous systems. Such systems often operate in environments that are continuously evolving and where agents’ actions are non-deterministic, so called inherently non-stationary environments. When there are inconsistent results for agents acting on such an environment, learning and adapting is challenging. In this article, we propose P-MARL, an approach that integrates prediction and pattern change detection abilities into MARL and thus minimises the effect of non-stationarity in the environment. The environment is modelled as a time-series, with future estimates provided using prediction techniques. Learning is based on the predicted environment behaviour, with agents employing this knowledge to improve their performance in realtime. We illustrate P-MARL’s performance in a real-world smart grid scenario, where the environment is heavily influenced by non-stationary power demand patterns from residential consumers. We evaluate P-MARL in three different situations, where agents’ action decisions are independent, simultaneous, and sequential. Results show that all methods outperform traditional MARL, with sequential P-MARL achieving best results.

This article describes Decentralised Emergence Detection (DETect), a novel distributed algorithm that enables agents to collaboratively detect emergent events in Complex Adaptive Systems (CAS). Non-deterministic interactions between agents in CAS can give rise to emergent behaviour or properties at the system level. The nature, timing, and consequence of emergence is unpredictable and may be harmful to the system or individual agents. DETect relies on the feedback that occurs from the system level (macro) to the agent level (micro) when emergence occurs. This feedback constrains agents at the micro level and results in changes occurring in the relationship between an agent and its environment. DETect uses statistical methods to automatically select the properties of the agent and environment to monitor and tracks the relationship between these properties over time. When a significant change is detected, the algorithm uses distributed consensus to determine if a sufficient number of agents have simultaneously experienced a similar change. On agreement of emergence, DETect raises an event, which its agent or other interested observers can use to act appropriately. The approach is evaluated using a multi-agent case study.

Smartphones with various embedded sensors and wirelessly connected external sensors will enable new applications across a wide variety of domains. Continuous or long-term sensing, processing, and communication of sensor data using smartphones will consume a significant amount of energy of the resource-constrained smartphones. Compression techniques, including predictive coding (PC) and compressed sensing (CS), are promising ways of minimizing energy consumption. A number of compression-based proposals are available to improve the energy efficiency in smartphones. Most of these proposals are sensor and application-specific, and their sampling rates may not be adaptive. This article proposes an energy efficient data gathering (DG) framework based on compression techniques, in particular as a component of a middleware for the smartphones. The framework adaptively selects the best possible compression technique (CT) for a sensor DG from a list of CTs, using the application's requirements and contexts. This work also presents a CS-based PC (CPC) to minimize the learning cost in PC. The framework includes the CPC along with other CTs and exploits context-aware adaptive sampling rate to improve energy efficiency. An initial evaluation of the framework using two real datasets highlights the potential of it and the CPC.

The Internet of Things (IoT) envisages a future in which digital and physical things or objects (e.g., smartphones, TVs, cars) can be connected by means of suitable information and communication technologies, to enable a range of applications and services. The IoT's characteristics, including an ultra-large-scale network of things, device and network level heterogeneity, and large numbers of events generated spontaneously by these things, will make development of the diverse applications and services a very challenging task. In general, middleware can ease a development process by integrating heterogeneous computing and communications devices, and supporting interoperability within the diverse applications and services. Recently, there have been a number of proposals for IoT middleware. These proposals mostly addressed wireless sensor networks (WSNs), a key component of IoT, but do not consider RF identification (RFID), machine-to-machine (M2M) communications, and supervisory control and data acquisition (SCADA), other three core elements in the IoT vision. In this paper, we outline a set of requirements for IoT middleware, and present a comprehensive review of the existing middleware solutions against those requirements. In addition, open research issues, challenges, and future research directions are highlighted.

We introduce a novel concept of ``architecture diversity'' for adaptive systems and posit that increased diversity has an inverse correlation with adaptation costs. We propose an index to quantify diversity and a static method to estimate the adaptation cost, and conduct an initial experiment on an exemplar cloud-based system which reveals the posited correlation.

Small scale electrical demand forecasting is an emerging field motivated by the penetration of renewable energy sources and the growth of microgrids and virtual power plants. These advances pose more complex forecasting challenges compared to the already established large scale forecasting approaches. Current short term load forecasting methods deal with two types of day, normal and anomalous, which are predicted separately. Anomalous days are classified as such ahead of time, based on key calendar events such as public holidays. However, there are some anomalous days which are not always predictable on a day ahead basis. Due to unforeseen events, a seemingly normal day can progress towards an anomalous case causing high errors in prediction. We propose a new dynamic forecasting mechanism that actively monitors residential electrical demand along a forecasted day, and detects anomalous pattern changes from a previously predicted demand of the day. A self-organising map is employed to detect anomalous days as they progress. Once an anomaly is detected, a neural network based prediction system changes its input neurons according to a previously detected and recorded match found in a database of anomalous days, in order to accommodate the anomalous day prediction. Results are based on measured power demands recorded in Ireland from domestic smart-meters between 2009-2011, and focus on small scale residential electrical demands of up to 350 kWh. During anomalous days our dynamic prediction approach achieves forecasting results within 3.63% of the real load, down from the 7.37% obtained by the initial prediction algorithm and the 5.41% achieved by standalone re-prediction, without pattern matching.

Large-scale, multi-agent systems are too complex for optimal control strategies to be known at design time and as a result good strategies must be learned at runtime. Learning in such systems, particularly those with multiple objectives, takes a considerable amount of time because of the size of the environment and dependencies between goals. Transfer Learning (TL) has been shown to reduce learning time in single-agent, single-objective applications. It is the process of sharing knowledge between two learning tasks called the source and target. The source is required to have been completed prior to the target task. This work proposes extending TL to multi-agent, multi-objective applications. To achieve this, an on-line version of TL called Parallel Transfer Learning (PTL) is presented. The issues involved in extending this algorithm to a multi-objective form are discussed. The effectiveness of this approach is evaluated in a smart grid scenario. When using PTL in this scenario learning is significantly accelerated. PTL achieves comparable performance to the base line in one third of the time.

Mobile devices capture, process, and exchange sensory data about their operating environment making them attractive service providers for pervasive computing. Composing services from different devices supports context-aware applications for smart spaces. However, carrier mobility and participation autonomy cause frequent changes to the network and service topology and impose a high failure probability on composites. Decentralised composition solutions assign service providers at runtime to provide for flexibility and to avoid a single point of failure. However, existing solutions rely on a pre-established service overlay network and employ a conservative allocation strategy which limits their applicability for highly dynamic environments. This paper presents a novel service composition protocol that allocates and invokes service providers opportunistically to minimise the impact of topology changes and to reduce failure. The protocol supports service sequences and parallel service flows. Automated model checking verifies that the protocol does not deadlock and that it terminates in a valid end state after having allocated the correct number of service providers for all required sub-services. The results of the simulation-based evaluation demonstrate that the opportunistic approach generally reduces composition failure. At the same time, the evaluation reveals the protocol's limits with regard to composite complexity, network density, and service demand.

Monitoring and adaptation of multilayer systems are challenging, because the mismatches and adaptations are interrelated across the layers.This interrelation introduces two important but difficult questions. 1) When a system change causes mismatches in one layer, how to identify all the cascaded mismatches on the other layers?2) When an adaptation is performed at one layer, how to find out all the complementary adaptations required in other layers.This paper presents a model-driven engineering approach towards cross-layer monitoring and adaption of multilayer systems.We provide standard meta-modeling languages for system experts to specify the concepts and constraints separately for each layer, as well as the relations among the concepts from different layers.An automated engine uses these meta-level specifications to 1) represent the system states on each layer as a runtime model, 2) evaluate the constraints to detect mismatches and assist adaptations within a layer, and 3) synchronize the models to identify cascaded mismatches and complementary adaptations across the layers.We illustrate the approach on a simulated crisis management system, and are using it on a number of ongoing projects.

Advances in pervasive technology have made it possible to consider large-scale application types that potentially span heterogeneous organizations, technologies, and device types. This class of application will have a multilayer architecture, where each layer is likely to use languages and technologies appropriate to its own concerns. An example application is a geographically large-scale crisis management system. Typically, such applications are required to dynamically adapt their behavior based on current circumstances, with adaptations potentially affecting all layers of the application. The complexities involved in dynamically adapting multilayer applications will significantly benefit from formal approaches to its specification. This article presents a new methodology for flexible, multilayer application adaptation, with layer-specific adaptation solution templates bound to application mismatches that are organized into hierarchical taxonomies. Templates can be linked either through direct invocations or through adaptation events, supporting flexible cross-layer adaptation. The methodology illustrates the use of different formalisms for different elements of its specification. In particular, we combine semiformal metamodeling techniques for the system model specification with formal Petri nets, which are used to capture template matchmaking using reachability analysis. This work demonstrates how existing formalisms can be used for the specification of a generic adaptation model for pervasive applications.

Developing testbenches for dynamic functional verification of hardware designs is a software-intensive process that lies on the critical path of electronic system design. The increasing capabilities of electronic components is contributing to the construction of complex verification environments that are increasingly difficult to understand, maintain, extend, and reuse across projects. Model-driven software engineering addresses issues of complexity, productivity, and code quality through the use of high-level system models and subsequent automatic transformations. Reasoning about verification testbench decomposition becomes simpler at higher levels of abstraction. In particular, the aspect-oriented paradigm, when applied at the model level, can minimize the overlap in functionality between modules, improving maintainability and reusability. This article presents an aspect-oriented model-driven engineering process and toolset for the development of hardware verification testbenches. We illustrate how this process and toolset supports modularized design and automatic transformation to verification environment-specific models and source code through an industry case study.

Dynamic service composition has emerged as a promising approach to build complex runtime-adaptable applications as it allows for binding service providers only shortly before service execution. However, the dynamic and ad hoc nature of mobile computing environments poses a significant challenge for dynamic service composition. In particular, the lack of central control and the potential volatility of service providers increase the complexity and failure probability of the composition process. Although, current research has led to decentralised composition algorithms and failure recovery strategies, the key question of how to reduce the failure probability of a composition still remains. We address this question and propose opportunistic service composition, an optimised execution model for complex service requests. The model merges the execution phase into the dynamic binding phase and supports the immediate fulfilment of partially composed service requests. We evaluated our model in mobile ad hoc network simulations. The results show an improvement over a baseline approach regarding composition success, response time, and communication effort.

Wireless sensor networks necessitate a programming model different from those used to develop desktop applications. Typically, resources in terms of power and memory are constrained. C is the most common programming language used to develop applications on very small embedded sensor devices. We claim that C does not provide efficient mechanisms to address the implicit asynchronous nature of sensor sampling. C applications for these devices suffer from a disruption in their control flow. In this paper, we present FlowTalk, a new object-oriented programming language aimed at making software development for wireless embedded sensor devices easier. FlowTalk is an object-oriented programming language in which dynamicity (e.g., object creation) has been traded for a reduction in memory consumption. The event model that traditionally comes from using sensors is adapted in FlowTalk with controlled disruption, a light-weight continuation mechanism. The essence of our model is to turn asynchronous long-latency operations into synchronous and blocking method calls. FlowTalk is built for TinyOS and can be used to develop applications that can fit in 4 KB of memory for a large number of wireless sensor devices.

The service-oriented computing paradigm is in widespread use for adaptive systems that face changing conditions in their operational environment as well as the integration of new services. In many domains, adaptations may occur dynamically and in real-time, using services from heterogeneous, possibly unknown sources. This motivates a need to ensure the correct behaviour of the adapted system, and its continuing compliance to time bounds. The complexity of dynamic adaptation (DA) is significant, but unfortunately currently not well understood or formally specified. Formal methods are an attractive option for solving this problem as they provide a means to precisely model a software system. There are many formal languages targeted to different domains, and in this paper, we present the results of our analysis of three languages as potential candidates for modelling our time-constrained DA problem. In particular, we selected JOLIE, PiDuce and COWS for analysis, as they are targeted towards service-based systems and each provide means to model at least some of our requirements. Our results illustrate the strengths and limitations of each, and justify our selection of COWS as the best-fit, though limited, language for our purposes.

Model-driven engineering addresses issues of platform heterogeneity and code quality through the use of high-level system models and subsequent automatic transformations. Adoption of the model-driven software engineering paradigm for embedded systems necessitates specification of appropriate models of often complex systems. Modern embedded systems are typically composed of multiple functional and nonfunctional concerns, with the nonfunctional concerns (e.g., timing and performance) typically affecting the design and implementation of the functional concerns. The presence of crosscutting concerns makes specification of adequate platform-independent models a significant challenge. Aspect-oriented software development is a separation of concerns technique that decomposes systems into distinct features with minimal overlap. In this article, we illustrate how Theme/UML, an aspect-oriented modeling approach, can be used to separate embedded systems concerns and reduce complexity in design. We also present Model-Driven Theme/UML, a toolset for model-driven engineering of embedded systems that supports modularised design with Theme/UML and automatic transformations to composed models and source code.

Service-oriented enterprise systems, which tend to be heterogeneous, loosely coupled, long-lived, and continuously running, have to cope with frequent changes to their requirements and the environment. In order to address such changes, applications need to be inherently flexible and adaptive, supported by appropriate infrastructures. In this paper, we propose a model-driven approach for the dynamic adaptation of Web services based on ontology-aware service templates. Model-driven engineering raises the level of abstraction from concrete Web service implementations to high-level service models, which leads to more flexible and automated adaptations through template designs and transformations. The ontological semantics enhances the service matching capabilities required by the dynamic adaptation process. Service templates are based on OWL-S descriptions and provide the necessary means to capture and parameterize specific behavior patterns of service models. In this paper, we apply our approach in the context of the EU-funded ALIVE project and illustrate, as an example, how the proposed framework supports the adaptation of the authentication mechanism used by an interactive tourist recommendation system.

Given the heterogeneity of healthcare software systems, data from each system is often incompatible inhibiting interoperability. To enable the sharing and exchange of healthcare information interoperability standards must be adhered to. Health Level Seven (HL7) is the international standards organisation that promotes and enforces the standardisation of electronic healthcare information to facilitate its exchange and management. Incorporating HL7 functionality into existing applications requires significant modification and intrusive extensions. Using aspect-oriented programming (AOP), we can introduce HL7 functionality into existing applications without the requirement for refactoring or modification. HL7 data formatting affects multiple parts of an application and hence is a ldquocrosscut-ting concernrdquo. These concerns which entwine with base functionality introduce complexity and reduce modularity. A second benefit of AOP is its advanced modularisation capabilities which are capable of modularising ldquocrosscut-ting concernsrdquo. We illustrate the benefits of using AOP in HL7 by example and measure the effects of the approach on healthcare applications.

Mobile ad hoc networks rely on the opportunistic interaction of autonomous nodes to form networks without the use of infrastructure. Given the radically decentralized nature of such networks, their potential for autonomous communication is significantly improved when the need for a priori consensus among the nodes is kept to a minimum. This paper addresses an issue within the domain of semantic content discovery, namely, its current reliance on the preexisting agreement between the schema of content providers and consumers. We present OntoMobil, a semantic discovery model for ad hoc networks that removes the assumption of a globally known schema and allows nodes to publish information autonomously. The model relies on the randomized dissemination and replication of metadata through a gossip protocol. Given schemas with partial similarities, the randomized metadata dissemination mechanism facilitates eventual semantic agreement and provides a substrate for the scalable discovery of content. A discovery protocol can then utilize the replicated metadata to identify content within a predictable number of hops using semantic queries. A stochastic analysis of the gossip protocol presents the different trade-offs between discoverability and replication. We evaluate the proposed model by comparing OntoMobil against a broadcast-based protocol and demonstrate that semantic discovery with proactive replication provides good scalability properties, resulting in a high discovery ratio with less overhead than a reactive nonreplicated discovery approach.

Software is increasingly deployed in vehicles as demand for new functionality increases and cheaper and more powerful hardware becomes available. Likewise, emerging wireless communication protocols allow the integration of new software into vehicles, thereby enabling time-bounded adaptive response to changes that occur in mobile environments. Examples of time-bounded adaptation include adaptive cruise control and the dynamic integration of location-aware services within fixed time bounds.

Handling context is required for applications to dynamically and appropriately adapt to their changing environment. Incorporating context into applications involves the consideration of a set of concerns related to the handling of various context types and the adaptation of the application behaviour relative to the current context. These concerns are usually heavily tangled with the base code of the applications, resulting in code that is badly modularised and therefore is hard to understand, manage and modify. We propose a modularised design for the handling of different kinds of context using aspect-oriented programming techniques. We demonstrate that a context-aware application built in this manner exhibits improved modularity, with corresponding improvements in comprehensibility, manageability and maintainability. The proposed aspect-oriented modularisation is evaluated against traditional object-oriented techniques, and also against a popular context framework, using metrics indicating coupling, cohesion and complexity. The results show the positive effect of modular code on context-aware applications by quantitatively illustrating the improvements in modularisation quality factors

Aspects are behaviours that are tangled and scattered across a system. In requirements documentation, aspects manifest themselves as descriptions of behaviours that are intertwined, and woven throughout. Some aspects may be obvious, as specifications of typical crosscutting behaviour. Others may be more subtle, making them hard to identify. In either case, it is difficult to analyse requirements to locate all points in the system where aspects should be applied. These issues lead to problems achieving traceability of aspects throughout the development lifecycle. To identify aspects early in the software lifecycle, and establish sufficient traceability, developers need support for aspect identification and analysis in requirements documentation. To address this, we have devised the Theme approach for viewing the relationships between behaviours in a requirements document, identifying and isolating aspects in the requirements, and modelling those aspects using a design language. This paper describes the approach, and illustrates it with a case study and analysis.

Some concerns, such as debugging or logging functionality, cannot be captured cleanly, and are often tangled and scattered throughout the code base. These concerns are called crosscutting concerns. Aspect-oriented programming (AOP) is a paradigm that enables developers to capture crosscutting concerns in separate aspect modules. The use of aspects has been shown to improve understandability and maintainability of systems. It has been shown that real-time concerns, such as memory management and thread scheduling, are crosscutting concerns [A. Corsaro et al., (2002), M.Deters et al., (2001), A. Gal et al., (2002)]. However it is unclear whether encapsulating these concerns provides benefits. We were interested in determining whether using AOP to encapsulate real-time crosscutting concerns afforded benefits in system properties such as understandability and maintainability. This paper presents research comparing the system properties of two systems: a real-time sentient traffic simulator and its aspect-oriented equivalent. An evaluation of AOP is presented indicating both benefits and drawbacks with this approach.

article Free Access Share on Subject-oriented design: towards improved alignment of requirements, design, and code Authors: Siobhán Clarke School of Computer Applications, Dublin City University, Dublin 9, Republic of Ireland School of Computer Applications, Dublin City University, Dublin 9, Republic of IrelandView Profile , William Harrison IBM T.J. Watson Research Center, P.O. Box 704, Yorktown Heights, NY IBM T.J. Watson Research Center, P.O. Box 704, Yorktown Heights, NYView Profile , Harold Ossher IBM T.J. Watson Research Center, P.O. Box 704, Yorktown Heights, NY IBM T.J. Watson Research Center, P.O. Box 704, Yorktown Heights, NYView Profile , Peri Tarr IBM T.J. Watson Research Center, P.O. Box 704, Yorktown Heights, NY IBM T.J. Watson Research Center, P.O. Box 704, Yorktown Heights, NYView Profile Authors Info & Claims ACM SIGPLAN NoticesVolume 34Issue 10Oct. 1999 pp 325–339https://doi.org/10.1145/320385.320420Online:01 October 1999Publication History 94citation1,710DownloadsMetricsTotal Citations94Total Downloads1,710Last 12 Months24Last 6 weeks3 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF