The increasing data produced by IoT devices and the need to harness intelligence in our environments impose the shift of computing and intelligence at the edge, leading to a novel computing paradigm called Edge Intelligence/Edge AI. This paradigm combines Artificial Intelligence and Edge Computing, enables the deployment of machine learning algorithms to the edge, where data is generated, and is able to overcome the drawbacks of a centralized approach based on the cloud (e.g., performance bottleneck, poor scalability, and single point of failure). Edge AI supports the distributed Federated Learning (FL) model that maintains local training data at the end devices and shares only globally learned model parameters in the cloud. This paper proposes a novel, energy-efficient, and dynamic FL-based approach considering a hierarchical edge FL architecture called HED-FL, which supports a sustainable learning paradigm using model parameters aggregation at different layers and considering adaptive learning rounds at the edge to save energy but still preserving the learning model’s accuracy. Performance evaluations of the proposed approach have also been led out considering model accuracy, loss, and energy consumption.

The management of thermal comfort in a building is a challenging and multifaced problem, because the use of objective parameters, for example, the energy consumption, should be combined with subjective requirements, related to human profile and preferences. This article exploits cognitive technologies, based on deep reinforcement learning (DRL), for the automatic control of the heating, ventilation, and air conditioning system in an office. The learning process is driven by a reward that includes multiple components, related to energy consumption, indoor temperature, and user perceptions, which are inferred by the human interactions with the system. This approach is inspired by the human-in-the-loop paradigm, which in our case helps the DRL controller to learn the requirements of users and readily adapt to them. Experimental results show that the appropriate balance of the reward components can be efficiently exploited to give the desired importance to the different objectives.

The Internet of Things (IoT) is the latest example of the System of Systems (SoS), demanding for both innovative and evolutionary approaches to tame its multifaceted aspects. Over the years, different IoT methodologies, frameworks, platforms, and tools have been proposed by industry and academia, but the jumbled abundance of such development products have resulted into a high (and disheartening) entry-barrier to IoT system engineering. In this survey, we steer IoT developers by: 1) providing baseline definitions to identify the most suitable class of development products-methodologies, frameworks, platforms, and tools-for their purposes and 2) reviewing seventy relevant products through a comparative and practical approach, based on general SoS engineering features revised in the light of main IoT systems desiderata (i.e., interoperability, scalability, smartness, and autonomy). Indeed, we aim to lessen the confusion related to IoT methodologies, frameworks, platforms, and tools as well as to freeze their current state, for eventually easing the approach towards IoT system engineering.

In this work, we propose an IoT edge-based energy management system devoted to minimizing the energy cost for the daily-use of in-home appliances. The proposed approach employs a load scheduling based on a load shifting technique, and it is designed to operate in an edge-computing environment naturally. The scheduling considers all together time-variable profiles for energy cost, energy production, and energy consumption for each shiftable appliance. Deadlines for load termination can also be expressed. In order to address these goals, the scheduling problem is formulated as a Markov decision process and then processed through a reinforcement learning technique. The approach is validated by the development of an agent-based real-world test case deployed in an edge context.

This paper presents a novel approach for the management of buildings by leveraging cognitive technologies. The proposed approach exploits the Deep Reinforcement Learning paradigm to learn from both a physical and a simulated environment so as to optimize people comfort and energy consumption.

Smart Environments (SEs) are pervasive systems usually built on top of IoT-based sensing and actuation devices which are spread in an environment. The increase of the on-board computational capacity of the used devices opens to the possibility of naturally exploiting the edge computing paradigm in which the computation is pushed at the edge of the network. Anyway, despite the huge interest towards SEs, there is a lack of approaches for their design. This paper proposes an enhancement of the existing Smart Environment Metamodel (SEM) framework suited for designing SEs. The provided extension aims at taking into account issues related to edge computing, management of timing information and definition of the data types involved in data sources. The effectiveness of the whole proposal is assessed through a case study describing the development of a Smart Office.

Large-scale smart environments (LSEs) are open and dynamic systems typically extending over a wide area and including a huge number of interacting devices with a heterogeneous nature. Thus, during their deployment scalability and interoperability are key requirements to be definitely taken into account. To these, discovery and reputation assessment of services and objects have to be added, given that new devices and functionalities continuously join LSEs. In spite of the increasing interest in this topic, effective approaches to develop LSEs are still missing. This paper proposes an agent-based approach that leverages edge computing and Social Internet of Things paradigms in order to address the above mentioned issues. The effectiveness of such an approach is assessed through a sample case study involving a commercial road environment.

A smart environment is a physical environment enriched with sensing, actuation, communication and computation capabilities aiming at acquiring and exploiting knowledge about the environment so as to adapt it to inhabitants' preferences and requirements. In this domain, there is the need of tools supporting the design and analysis of applications. In this paper, a meta-model framework for smart environments is proposed. This framework allows to model applications by exploiting concepts closer to the smart environment domain. The proposed meta-model framework approaches the modelling from two different points of view, namely the functional and data perspectives. The functional perspective focuses on the services provided by the environment whereas the data perspective is used to characterize data sources of the environment. The effectiveness of the proposal is shown by applying it to the modelling of a smart environment scenario well known in literature.

This paper analyzes and evaluates the strategies and implications related to the execution of parallel algorithms on a distributed Cloud infrastructure, with the focus on an important class of applications for which the execution is performed on spatial data, dislocated on a bidimensional territory. Applications of interest cover a wide spectrum ranging from Internet of Things to social sciences, geology, swarm-inspired computation etc. The territory is partitioned into regions, and regions are assigned to parallel computational nodes to speed up the execution. Parallel nodes are aligned through the exchange of messages in order to ensure a coherent and efficient execution. The paper offers an analysis of the parallelization cost in this context, especially in terms of communication overhead, which is essential to estimate the impact of porting the computation onto a Cloud environment. More in particular, the paper evaluates two different strategies for space partitioning, i.e., linear partitioning and bidimensional partitioning, with a specific focus on scalability analysis, and compares the two strategies when both options are exploitable.

Smart Cyber-Physical Environments are augmented physical environments whose behaviours are enhanced through the use of ICT technologies. The goal is to offer new services and functionalities devoted to meet people's needs and preferences, and to better exploit existing services and infrastructures. The use of IoT technologies, paired with the edge computing, fosters the development of Smart Environment applications having the important features of reliability, scalability and extensibility. This paper proposes an approach for the design and the implementation of Smart Cyber Physical Environment applications having the aforementioned features. The approach relies on the use of isapiens which is an IoT platform enabling edge computing through the exploitation of the agent metaphor. Such platform provides effective abstractions which are able to hide heterogeneity of both the adopted hardware devices and communication protocols. The approach is validated through a case study involving the realization of a Smart Office prototype for profiling and monitoring daily working activities and performing actuations in the environment on the basis of the obtained information.

Recent advancements in embedded systems, computer science and telecommunication fields open up to new application opportunities where many physical entities are disseminated across the world and connected through the Internet. This new scenario, often referred as Internet of Things (IoT), gives raise to different issues and challenges to be coped with. Indeed, each kind of physical device comes with different technology details, low level communication protocols and can exhibit different semantic behaviors. In addition, the interaction between human beings and the physical objects must be properly managed. Two main approaches exist to cope with the mentioned issues: (i) the development of new ad-hoc technologies and solutions to deal with the specific issues raising from the new scenario, or (ii) the exploitation of well-known technologies and solutions also in the new context so as to facilitate the integration of the physical stuff with the preexisting internet services. The latter approach is usually referred as Web of Things (WoT). This paper proposes a possible implementation of the WoT vision. The resulting architecture allows creating complex applications where physical resources, internet resources and human resources can properly interact with each other. The objects of the physical world are virtualized and managed through the Smart Object (SO) concept. The SOs are enclosed in a Smart Gateway which exposes them to the world through an uniform web API based on REST paradigm. The human beings interaction is achieved using the popular micro-blogging platform Twitter. The integration between all the entities is provided by adopting a WS-BPEL workflow technology. To validate the approach an example of a smart room environment controlled through Twitter in a crowd source fashion is detailed.

Swarm intelligence algorithms, based on multi-agent systems, are often used to solve complex problems that are not affordable through classical centralized/deterministic solutions. In many cases, to enhance the performance of such algorithms, the computation can be distributed to parallel/distributed nodes, in accordance with different strategies. Specifically, parallelization can be achieved either by partitioning the space in which agents operate among the nodes, or by assigning the entire space to each node but distributing input data through a sampling approach. Another choice is whether or not the management of conflicts is needed to prevent possible loss of data consistency. This paper discusses such issues, while referring to two well-known types of swarm intelligence algorithms -- ants and flocking -- and compares the mentioned strategies, evaluating the performance results in terms of speedup.

The ICT market is experiencing an important shift from the request/provisioning of products toward a service-oriented view where everything (computing, storage, applications) is provided as a network-enabled service. It often happens that a solution to a problem cannot be offered by a single service, but by composing multiple basic services in a workflow. Service composition is indeed an important research topic that involves issues such as the design and execution of a workflow and the discovery of the component services on the network. This paper deals with the latter issue and presents an ant-inspired framework that facilitates collective discovery requests, issued to search a network for all the basic services that will compose a specific workflow. The idea is to reorganize the services so that the descriptors of services that are often used together are placed in neighbor peers. This helps a single query to find multiple basic services, which decreases the number of necessary queries and, consequently, lowers the search time and the network load.

Existing density-based data stream clustering algorithms use a two-phase scheme approach consisting of an online phase, in which raw data is processed to gather summary statistics, and an offline phase that generates the clusters by using the summary data. In this paper we propose a data stream clustering method based on a multi-agent system that uses a decentralized bottom-up self-organizing strategy to group similar data points. Data points are associated with agents and deployed onto a 2D space, to work simultaneously by applying a heuristic strategy based on a bio-inspired model, known as flocking model. Agents move onto the space for a fixed time and, when they encounter other agents into a predefined visibility range, they can decide to form a flock if they are similar. Flocks can join to form swarms of similar groups. This strategy allows to merge the two phases of density-based approaches and thus to avoid the offline cluster computation, since a swarm represents a cluster. Experimental results show the capability of the bio-inspired approach to obtain very good results on real and synthetic data sets.

This article presents So-Grid, a set of bio-inspired algorithms tailored to the decentralized construction of a Grid information system that features adaptive and self-organization characteristics. Such algorithms exploit the properties of swarm systems, in which a number of entities/agents perform simple operations at the local level, but together engender an advanced form of swarm intelligence at the global level. In particular, So-Grid provides two main functionalities: logical reorganization of resources, inspired by the behavior of some species of ants and termites that move and collect items within their environment, and resource discovery, inspired by the mechanisms through which ants searching for food sources are able to follow the pheromone traces left by other ants. These functionalities are correlated, since an intelligent dissemination can facilitate discovery. In the Grid environment, a number of ant-like agents autonomously travel the Grid through P2P interconnections and use biased probability functions to: (i) replicate resource descriptors in order to favor resource discovery; (ii) collect resource descriptors with similar characteristics in nearby Grid hosts; (iii) foster the dissemination of descriptors corresponding to fresh (recently updated) resources and to resources having high quality of service (QoS) characteristics. Simulation analysis shows that the So-Grid replication algorithm is capable of reducing the entropy of the system and efficiently disseminating content. Moreover, as descriptors are progressively reorganized and replicated, the So-Grid discovery algorithm allows users to reach Grid hosts that store information about a larger number of useful resources in a shorter amount of time. The proposed approach features characteristics, including self-organization, scalability and adaptivity, which make it useful for a dynamic and partially unreliable distributed system.

A boosting algorithm based on cellular genetic programming (GP) to build an ensemble of predictors is proposed. The method evolves a population of trees for a fixed number of rounds and, after each round, it chooses the predictors to include in the ensemble by applying a clustering algorithm to the population of classifiers. Clustering the population allows the selection of the most diverse and fittest trees that best contribute to improve classification accuracy. The method proposed runs on a distributed hybrid environment that combines the island and cellular models of parallel GP. The combination of the two models provides an efficient implementation of distributed GP, and, at the same time, the generation of low sized and accurate decision trees. The large amount of memory required to store the ensemble affects the performance of the method. This paper shows that, by applying suitable pruning strategies, it is possible to select a subset of the classifiers without increasing misclassification errors; indeed for some data sets, for up to 30% of pruning, ensemble accuracy increases. Experimental results show that the combination of clustering and pruning enhances classification accuracy of the ensemble approach.

The paper presents an adaptive GP boosting ensemble method forthe classification of distributed homogeneous streaming data that comes from multiple locations. The approach is able to handle concept drift via change detection by employing a change detection strategy, based on self-similarity of the ensemble behavior, and measured by its fractal dimension. It is efficient since each nodeof the network works with its local streaming data, and communicate only the local model computed with the otherpeer-nodes. Furthermore, once the ensemble has been built, it isused to predict the class membership of new streams of data until concept drift is detected. Only in such a case the algorithm is executed to generate a new set of classifiers to update the current ensemble. Experimental results on a synthetic and reallife data set showed the validity of the approach in maintaining an accurate and up-to-date GP ensemble.

An adaptive boosting ensemble algorithm for classifying homogeneous distributed data streams is presented. The method builds an ensemble of classifiers by using Genetic Programming (GP) to inductively generate decision trees, each trained on different parts of the distributed training set. The approach adopts a co-evolutionary platform to support a cooperative model of GP. A change detection strategy, based on self-similarity of the ensemble behavior, and measured by its fractal dimension, permits to capture time- evolving trends and patterns in the stream, and to reveal changes in evolving data streams. The approach tracks online ensemble accuracy deviation over time and decides to recompute the ensemble if the deviation has exceeded a pre- specified threshold. This allows the maintenance of an accurate and up-to-date ensemble of classifiers for continuous flows of data with concept drifts. Experimental results on a real life data set show the validity of the approach.

A boosting algorithm based on cellular genetic programming to build an ensemble of predictors is proposed. The method evolves a population of trees for a fixed number of rounds and, after each round, it chooses the predictors to include into the ensemble by applying a clustering algorithm to the population of classifiers. The method proposed runs on a distributed hybrid multi-island environment that combines the island and cellular models of parallel genetic programming. The large amount of memory required to store the ensemble makes the method heavy to deploy. The paper shows that by applying suitable pruning strategies it is possible to select a subset of the classifiers without increasing misclassification errors; indeed, up to 20 of pruning, ensemble accuracy increases. Experiments on several data sets show that combining clustering and pruning enhances classification accuracy of the ensemble approach.

An extension of cellular genetic programming for data classification (CGPC) to induce an ensemble of predictors is presented. Two algorithms implementing the bagging and boosting techniques are described and compared with CGPC. The approach is able to deal with large data sets that do not fit in main memory since each classifier is trained on a subset of the overall training data. The predictors are then combined to classify new tuples. Experiments on several data sets show that, by using a training set of reduced size, better classification accuracy can be obtained, but at a much lower computational cost

This paper presents a study that evaluates the influence of the parallel genetic programming (GP) models in maintaining diversity in a population. The parallel models used are the cellular and the multipopulation one. Several measures of diversity are considered to gain a deeper understanding of the conditions under which the evolution of both models is successful. Three standard test problems are used to illustrate the different diversity measures and analyze their correlation with performance. Results show that diversity is not necessarily synonym of good convergence.

The purpose of this work is to investigate the use of new swarm intelligence based techniques for data mining. According to this approach the data mining task is constructed as a set of biologically inspired agents. Each agent represents a simple task and the success of the method depends on the cooperative work of the agents. In this paper, we present a novel algorithm that uses techniques adapted from models originating from biological collective organisms to discover clusters of arbitrary shape, size and density in spatial data. The algorithm combines a smart exploratory strategy based on the movements of a flock of birds with a shared nearest-neighbor clustering algorithm to discover clusters in parallel. In the algorithm, birds are used as agents with an exploring behavior foraging for clusters. Moreover, this strategy can be used as a data reduction technique to perform approximate clustering efficiently. We have applied this algorithm on synthetic and real world data sets and we have measured, through computer simulation, the impact of the flocking search strategy on performance.

A new parallel implementation of genetic programming (GP) based on the cellular model is presented and compared with both canonical GP and the island model approach. The method adopts a load-balancing policy that avoids the unequal utilization of the processors. Experimental results on benchmark problems of different complexity show the superiority of the cellular approach with respect to the canonical sequential implementation and the island model. A theoretical performance analysis reveals the high scalability of the implementation realized and allows to predict the size of the population when the number of processors and their efficiency are fixed.

A parallel genetic programming approach to induce decision trees in large data sets is presented. A population of trees is evolved by employing the genetic operators and every individual is evaluated by using a fitness function based on the J-measure. The method is able to deal with large data sets since it uses a parallel implementation of genetic programming through the grid model and an out of core technique for those data sets that do not fit in main memory. Preliminary experiments on data sets from the UCI machine learning repository give good classification outcomes and assess the scalability of the method.

A parallel hybrid method for solving the satisfiability (SAT) problem that combines cellular genetic algorithms (GAs) and the random walk SAT (WSAT) strategy of greedy SAT (GSAT) is presented. The method, called cellular genetic WSAT (CGWSAT), uses a cellular GA to perform a global search from a random initial population of candidate solutions and a local selective generation of new strings. The global search is then specialized in local search by adopting the WSAT strategy. A main characteristic of the method is that it indirectly provides a parallel implementation of WSAT when the probability of crossover is set to zero. CGWSAT has been implemented on a Meiko CS-2 parallel machine using a 2D cellular automaton as a parallel computation model. The algorithm has been tested on randomly generated problems and some classes of problems from the DIMACS and SATLIB test set.

A new parallel hybrid method for solving the satisfiability problem that combines cellular genetic algorithms and the random walk (WSAT) strategy of GSAT is presented. The method, called CGWSAT, uses a cellular genetic algorithm to perform a global search on a random initial population of candidate solutions and a local selective generation of new strings. Global search is specialized in local search by adopting the WSAT strategy. CGWSAT has been implemented on a Meiko CS-2 parallel machine using a two-dimensional cellular automaton as a parallel computation model. The algorithm has been tested on randomly generated problems and some classes of problems from the DIMACS test set.

Abstract This paper describes the design, the implementation, and the performance results of a routing algorithm which provides deadlock‐free communication in a tightly coupled message‐passing concurrent computer. The algorithm is adaptive, isolated and uses the store‐and‐forward technique. It allows message communication between two processes regardless of where they are physically located on the network. The routing algorithm has many positive characteristics including provable deadlock freedom, guaranteed message arrival, and automatic local congestion reduction. It can be used as a basis for the design of high‐level communication primitives. An Occam implementation on a network of inmos Transputers is discussed. The experimental results show that the routing algorithm is effective to support process to process communication on a concurrent computer.