Search

To enable autonomous vessels to operate on inland waterways, they need to detect, track and localize objects at close range to safely navigate. We deployed current deep learning techniques to detect and track these objects. As there are no large labeled datasets of European inland waterways, we used transfer learning to overcome the lack of data. By using preexisting similar datasets, we were able to significantly decrease the required amount of labeled data from the target distribution. Furthermore, we improved the mean Average Precision from 0.461 to 0.814 by using a limited number of labeled target data samples. We estimated the relative distance of the objects based on the generated bounding boxes. The information from the camera is then combined with LiDar data to generate a top-view map of the environment which is used as input for an object-avoidance control agent. All these methods can run in real-time on the vessel with an fps of 1.83 on a 2.7GHz vCPU.

The numerous applications of industrial automation have always posed many challenges for wireless connectivity. In the last decade, IEEE 802.15.4e Time-Slotted Channel Hopping (TSCH) networks have provided high reliability and low-power operation in such challenging industrial environments. Typically, TSCH networks employ one modulation at the physical layer and are thus limited by the characteristics of the chosen modulation in terms of, among others, data rate, reliability and energy efficiency. To tackle these limitations and to improve network performance and flexibility in those challenging industrial environments, this work explores the simultaneous use of multiple modulations in a TSCH network. Traditionally, TSCH relies on fixed-duration slots, large enough to send a packet of any size given the fixed data rate. In order to avoid wasting airtime when simultaneously using modulations with different data rates, we propose the concept of slot bonding. This allows the creation of different-sized bonded slots with a duration adapted to the data rate of each chosen modulation. To analyse the proposed slot bonding technique, we formally describe the TSCH slot bonding problem in terms of optimizing the packet delivery ratio while minimizing radio on time, with the inclusion of parent selection and interference avoidance. Afterwards, we propose a genetic algorithm that allows us to implement the problem and find solutions heuristically. Finally, we provide insights into preferred parent selection and modulation configurations by using this heuristic approach during extensive simulation experimentation in which the scalability advantage of slot bonding over longer fixed-duration slots is also shown.

The service placement problem considers the placement of multiple connected services across a heterogeneous device network and is one of the core problems of fog computing. We discuss the complexity of this service placement problem, and propose a model for solving it using Multi-Objective Reinforcement Learning (MORL) methodologies. Using a trained neural network greatly reduces the resource consumption of the placement algorithm, making it viable for resource-constrained scenarios. Starting from state-of-the-art techniques, we develop a generic max reward formulation model and apply several MORL methodologies, which solve the placement problem in scenarios where the preference weights change. We compare the results to a baseline methodology and showcase the value of MORL on the placement problem.

Internet services grow increasingly demanding and diverse. Privacy and security concerns, coupled with requirements of ultra-low latency, close to zero downtime, and high throughput, are leading to a shift in network design. This combination of non-technical and technical service requirements is pushing cloud infrastructure to the edge of the network. While cloud services initially interconnected resources within a single data center, services are also being orchestrated across edge clouds. Traditional and new cloud architectures rely on network functions virtualization for the management and orchestration of network services. However, the orchestration of highly diverse and geographically distributed resources brings its own challenges, as networking becomes a bottleneck, failure rates of nodes are higher, and orchestration objectives increasingly conflict with one another. Hence, specific orchestration algorithms are needed that can adapt the placement configuration to the local cloud infrastructure and balance orchestration goals through SLAs. This article investigates the state of the art in service orchestration for NFV-based cloud environments along with the research challenges related to orchestrating resources within and across edge locations, to finally identify and discuss opportunities for future research.

In this letter, we propose an airtime-based Resource Allocation (RA) model for network slicing in IEEE 802.11 Radio Access Networks (RANs). We formulate this problem as a Quadratically Constrained Quadratic Program (QCQP), where the overall queueing delay of the system is minimized while strict Ultra-Reliable Low Latency Communication (URLLC) constraints are respected. We evaluated our model using three different solvers where the optimal and feasible sets of airtime configurations were computed. We also validated our model with experimentation in real hardware. Our results show that the solution time for computing optimal and feasible configurations vary according to the slice's demand distribution and the number of slices to be allocated. Our findings support the need for precise RA over IEEE 802.11 RANs and present the limitations of performing such optimizations at runtime.

Today’s and tomorrow’s networks are becoming increasingly complex and heterogeneous with a large diversity of devices and technologies. To meet growing demand, and support client mobility there is need for intelligent mechanisms like multi-technology load balancing and handovers. Current solutions, like Multipath Transmission Control Protocol (MPTCP), fail to provide a fine-grained, coordinated, and transparent answer to this heterogeneity, while the lower layers of the Open Systems Interconnection stack simply ignore it by providing full separation of layers. Therefore, we introduce ORCHESTRA, a data link layer framework for the management of multi-technology networks and devices, enabling packet-level dynamic handovers, load balancing, and duplication across network technologies. The framework is the first of its kind in providing fine-grained packet-level control across different technologies in a network-wide manner. Moreover, it works on top of existing standards without the need for hardware changes. This is achieved through a fully transparent virtual Medium Access Control layer and a Software-Defined Networking controller with global intelligence. The framework is implemented in a prototype running on off-the-shelf hardware and we demonstrate its features across different IEEE 802.11 technologies and 4G (Long Term Evolution). We demonstrate that ORCHESTRA outperforms MPTCP and allows for real-time inter-technology handovers and that overall throughput and reliability are improved in wireless networks.

Recent work in multi-agent reinforcement learning has investigated inter agent communication which is learned simultaneously with the action policy in order to improve the team reward. In this paper, we investigate independent Q-learning (IQL) without communication and differentiable inter-agent learning (DIAL) with learned communication on an adaptive traffic control system (ATCS). In real world ATCS, it is impossible to present the full state of the environment to every agent so in our simulation, the individual agents will only have a limited observation of the full state of the environment. The ATCS will be simulated using the Simulation of Urban MObility (SUMO) traffic simulator in which two connected intersections are simulated. Every intersection is controlled by an agent which has the ability to change the direction of the traffic flow. Our results show that a DIAL agent outperforms an independent Q-learner on both training time and on maximum achieved reward as it is able to share relevant information with the other agents.

Traffic Classification (TC) systems allow inferring the application that is generating the traffic being analyzed. State-of-the-art TC algorithms are based on Deep Learning (DL) and have outperformed traditional methods in complex and modern scenarios, even if traffic is encrypted. Most of the works on TC assume the traffic flows on a wired network under the same network management domain. This assumption limits the capabilities of TC systems in wireless networks since users’ traffic on one network domain can be negatively impacted by undetected traffic transmissions from users in other network domains or detected ones but with no traffic context in a shared spectrum. To solve this problem, we introduce a novel framework to achieve TC at any layer on the radio network stack. We propose a spectrum-based procedure that uses a DL-based classifier to realize this framework. We design two DL-based classifiers, a novel Convolutional Neural Network (CNN) spectrum-based TC and a Recurrent Neural Networks (RNN) as baseline architecture, and benchmark their performance on three TC tasks at different radio stack layers. The datasets were generated by combining packet traces from real transmissions with a standard-compliant waveform generator for 802.11 radio technologies. Performance evaluations show that the best model can achieve an accuracy above 92% in the most demanding TC task, a drop of only 4.37% in accuracy compared to a byte-based DL approach, with microsecond per-packet prediction time, which is very promising for delivering real-time spectrum-based traffic analyzers. This preprint is complemented with the dataset (https://doi.org/10.5281/zenodo.5208201) and the code used to generate the results (https://www.doi.org/10.5281/zenodo.5208627 and https://github.com/miguelhdo/tc spectrum).&nbsp

Artificial Intelligence (AI) powered building control allows deriving policies that are more flexible and energy efficient than standard control. However, there are challenges: environment interaction is used to train Reinforcement Learning (RL) agents but for building control it is often not possible to use a physical environment, and creating high fidelity simulators is a difficult task. With offline RL an agent can be trained without environment interaction, it is a data-driven approach to RL. In this paper, Conservative Q-Learning (CQL), an offline RL algorithm, is used to control the temperature setpoint in a room of a university campus building. The agent is trained using only the historical data available for this room. The results show that there is potential for offline RL in the field of building control, but also that there is room for improvement and need for further research in this area.

Situational awareness is getting traction in the field of autonomous inland vessels. Large amounts of data needs to be shared in order to set up this awareness. This ranges from relatively small positional updates, to consistent streams of sensory data. Point clouds, captured by LiDAR sensors, are heavily used by inland vessels as they give a detailed sense of range. This sharing is not a major complexity when vessels are in close proximity with each other; dedicated networks could handle this consistent stream of data. However, when vessels are farther away from each other, long range networks are needed, at the cost of high bandwidth capabilities. Therefore, the sensor message size should be reduced while retaining a reasonable quality. In this paper, we investigate the trade-off between lossless and lossy point cloud compression with Google Draco and its resulting quality. The results show that a considerable size reduction can be applied while the point cloud maintains acceptable quality.

The Sim2Real gap is a topic that has been receiving a great deal of attention lately. Many Artificial Intelligence techniques, for example Reinforcement Learning, require millions of iterations to achieve satisfactory performance. This requirement often forces these techniques to solely train in simulation. If the gap between the simulated environment and the target environment is too broad, however, the trained agents will lose out on performance when deployed. Bridging this gap lowers the performance loss during deployment, in turn improving the effectiveness of these agents. This paper proposes a new technique to tackle this issue. The technique focuses on the use of demonstration samples gathered in the target environment and is based on two transfer learning fundamentals. By combining the advantages of Domain Randomization and Domain Adaptation, agents are able to transfer training performance to the target environment more successfully. Experimental results show a strong decrease in performance loss during deployment when the agent is exposed to the demonstration samples during training. The proposed technique describes a methodology that we believe can be applied in fields other than autonomous driving in order to improve transfer learning performance.

Fog and edge computing paradigms were adopted to enable Internet of Things (IoT) applications, improving response time and reducing network load. Task allocation algorithms are used on IoT-enabled networks to determine the optimal software placement. However, managing such a network is considerably more complex than allocating the tasks. To simplify management, we propose a general Monitor - Analyze - Plan - Execute over a Knowledge base (MAPE-K) framework in which all requirements for task allocation are fulfilled, and where components can easily be adapted to the use case at hand. This research identifies several pitfalls and proposes solutions. Additionally, we apply this approach to a distributed testbed, comparing it to traditional cloud approaches.

Convolutional neural networks (CNNs) have accomplished magnificent performance on object classification tasks. This work introduces a novel image classification approach based on feature vector fusion of two CNN architectures using graph attention networks (GAT). In the proposed method we extract feature maps from shallow and deep layers of two CNN architectures. These extracted feature vectors are represented as nodes in a graph, and edges between nodes are constructed based on similarities between the feature vectors. The GAT is used to aggregate and fuse connected nodes based on their importance and relevance to the classification task. We believe that this approach compensates for convolution defects during feature processing when using a single CNN. This paper attempts to show that graph-based deep learning can be used to fuse two CNN architectures and not to push the state-of-the-art of image classification accuracy. Our experimental results prove that GAT can be used to fuse feature vectors resulting from CNN architectures.

Professional road cycling is a very competitive sport, and many factors influence the outcome of the race. These factors can be internal (e.g., psychological preparedness, physiological profile of the rider, and the preparedness or fitness of the rider) or external (e.g., the weather or strategy of the team) to the rider, or even completely unpredictable (e.g., crashes or mechanical failure). This variety makes perfectly predicting the outcome of a certain race an impossible task and the sport even more interesting. Nonetheless, before each race, journalists, ex-pro cyclists, websites and cycling fans try to predict the possible top 3, 5, or 10 riders. In this article, we use easily accessible data on road cycling from the past 20 years and the Machine Learning technique Learn-to-Rank (LtR) to predict the top 10 contenders for 1-day road cycling races. We accomplish this by mapping a relevancy weight to the finishing place in the first 10 positions. We assess the performance of this approach on 2018, 2019, and 2021 editions of six spring classic 1-day races. In the end, we compare the output of the framework with a mass fan prediction on the Normalized Discounted Cumulative Gain (NDCG) metric and the number of correct top 10 guesses. We found that our model, on average, has slightly higher performance on both metrics than the mass fan prediction. We also analyze which variables of our model have the most influence on the prediction of each race. This approach can give interesting insights to fans before a race but can also be helpful to sports coaches to predict how a rider might perform compared to other riders outside of the team. ispartof: FRONTIERS IN SPORTS AND ACTIVE LIVING vol:3 ispartof: location:Switzerland status: published

Topological data analysis is a recent and fast growing field that approaches the analysis of datasets using techniques from (algebraic) topology. Its main tool, persistent homology (PH), has seen a notable increase in applications in the last decade. Often cited as the most favourable property of PH and the main reason for practical success are the stability theorems that give theoretical results about noise robustness, since real data is typically contaminated with noise or measurement errors. However, little attention has been paid to what these stability theorems mean in practice. To gain some insight into this question, we evaluate the noise robustness of PH on the MNIST dataset of greyscale images. More precisely, we investigate to what extent PH changes under typical forms of image noise, and quantify the loss of performance in classifying the MNIST handwritten digits when noise is added to the data. The results show that the sensitivity to noise of PH is influenced by the choice of filtrations and persistence signatures (respectively the input and output of PH), and in particular, that PH features are often not robust to noise in a classification task., 24 pages, 7 figures, 4 tables

While IEEE 802.15.4e Time-Slotted Channel Hopping (TSCH) networks should be equipped to deal with the hard wireless challenges of industrial environments, the sensor networks are often still limited by the characteristics of the used physical (PHY) layer. Therefore, the TSCH community has recently started shifting research efforts to the support of multiple PHY layers, to overcome this limitation. On the one hand, integrating such multi-PHY support implies dealing with the PHY characteristics to fit the resource allocation in the TSCH schedule, and on the other hand, defining policies on how to select the appropriate PHY for each network link. As such, first a heuristic is proposed that is a step towards a distributed PHY and parent selection mechanism for slot bonding multi-PHY TSCH sensor networks. Additionally, a proposal on how this heuristic can be implemented in the IPv6 over the TSCH mode of IEEE 802.15.4e (6TiSCH) protocol stack and its Routing Protocol for Low-power and Lossy network (RPL) layer is also presented. Slot bonding allows the creation of different-sized bonded slots with a duration adapted to the data rate of each chosen PHY. Afterwards, a TSCH slot bonding implementation is proposed in the latest version of the Contiki-NG Industrial Internet of Things (IIoT) operating system. Subsequently, via extensive simulation results, and by deploying the slot bonding implementation on a real sensor node testbed, it is shown that the computationally efficient parent and PHY selection mechanism approximates the packet delivery ratio (PDR) results of a near-optimal, but computationally complex, centralized scheduler.

Network function (NF) virtualization is a promising way for service providers to improve configurability of the offered network services. First, NFs and their corresponding flows can be embedded in the substrate network in an automated way. Second, the modularity offered by the description of a service as a composition of NFs and their interconnecting virtual links, enables tuning of the service's logical structure. A major challenge with respect to the management of these environments is the placement of service requests in the physical infrastructure, comprising NFs that can only be instantiated on particular hosts, due to latency, legislation, or hardware-related constraints. These location constraints are known to complicate the resource allocation. This paper proposes two service placement algorithms with better coordination between the composition and embedding phases. We formulate the combined service embedding and chain composition problem as an integer linear program and then, propose a fast heuristic based on greedy chain selection that iteratively adds virtual NF chains to the composition and embeds them at the same time. Simulation experiments show that the proposed cross-layer optimization can significantly increase the acceptance ratio and reduce the embedding cost.

The use of different modalities improves the perception of the environment in situations where the conventional sensors fail (camera and LiDAR). The inclusion of these modalities, such as sonar or radar, is however difficult as existing methods for the conventional sensors usually do not generalise well on these different environment representations. We experiment with a modality prediction method to keep using the existing methodologies and allow to separate the sensing system from the navigation stack of an autonomous agent. In previous work, we used a convolutional stacked autoencoder to predict LiDAR point cloud data using the data from our 3D in-air acoustic ultrasonic sensor (eRTIS). In this paper, we investigate the usability of the predicted data in off-the-shelf algorithms to safely navigate environments where visual modalities become unreliable and less accurate.

Embodied AI, learning through interaction with a physical environment, typically requires large amounts of interaction with the environment in order to learn how to solve new tasks. Training can be done in parallel, using simulated environments. However, once deployed in e.g., a real-world setting, it is not yet clear how an agent can quickly adapt its knowledge to solve new tasks. In this paper, we propose a novel Hierarchical Reinforcement Learning (HRL) method that allows an agent, when confronted with a novel task, to switch between exploiting prior knowledge through temporally extended actions, and environment exploration. We solve this trade-off by utilizing the disagreement between action distributions of selected previously acquired policies. Selection of relevant prior tasks is done by measuring the cosine similarity of their attached natural language goals in a pre-trained word-embedding. We analyze the resulting temporal abstractions, and we experimentally demonstrate the effectiveness of them in different environments. We show that our method is capable of solving new tasks using only a fraction of the environment interactions required when learning the task from scratch.

IEEE 802.11 (Wi-Fi) is one of the technologies that provides high performance with a high density of connected devices to support emerging demanding services, such as virtual and augmented reality. However, in highly dense deployments, Wi-Fi performance is severely affected by interference. This problem is even worse in new standards, such as 802.11n/ac, where new features such as Channel Bonding (CB) are introduced to increase network capacity but at the cost of using wider spectrum channels. Finding the best channel assignment in dense deployments under dynamic environments with CB is challenging, given its combinatorial nature. Therefore, the use of analytical or system models to predict Wi-Fi performance after potential changes (e.g., dynamic channel selection with CB, and the deployment of new devices) are not suitable, due to either low accuracy or high computational cost. This paper presents a novel, data-driven approach to speed up this process, using a Graph Neural Network (GNN) model that exploits the information carried in the deployment’s topology and the intricate wireless interactions to predict Wi-Fi performance with high accuracy. The evaluation results show that preserving the graph structure in the learning process obtains a 64% increase versus a naive approach, and around 55% compared to other Machine Learning (ML) approaches when using all training features.

By using communication between multiple agents in multi-agent environments, one can reduce the effects of partial observability by combining one agent's observation with that of others in the same dynamic environment. While a lot of successful research has been done towards communication learning in cooperative settings, communication learning in mixed cooperative-competitive settings is also important and brings its own complexities such as the opposing team overhearing the communication. In this paper, we apply differentiable inter-agent learning (DIAL), designed for cooperative settings, to a mixed cooperative-competitive setting. We look at the difference in performance between communication that is private for a team and communication that can be overheard by the other team. Our research shows that communicating agents are able to achieve similar performance to fully observable agents after a given training period in our chosen environment. Overall, we find that sharing communication across teams results in decreased performance for the communicating team in comparison to results achieved with private communication.

Batteryless Internet-of-Things (IoT) devices need to schedule tasks on very limited energy budgets from intermittent energy harvesting. Creating an energy-aware scheduler allows the device to schedule tasks in an efficient manner to avoid power loss during execution. To achieve this, we need insight in the Worst-Case Energy Consumption (WCEC) of each schedulable task on the device. Different methodologies exist to determine or approximate the energy consumption. However, these approaches are computationally expensive and infeasible to perform on all type of devices; or are not accurate enough to acquire safe upper bounds. We propose a hybrid methodology that combines machine learning-based prediction on small code sections, called hybrid blocks, with static analysis to combine the predictions to a final upper bound estimation for the WCEC. In this paper, we present our work on an automated testbench for the Code Behaviour Framework (COBRA) that measures and profiles the upper bound energy consumption on the target device. Next, we use the upper bound measurements of the testbench to train eight different regression models that need to predict these upper bounds. The results show promising estimates for three regression models that could potentially be used for the methodology with additional tuning and training.

Indoor temperature modeling is a crucial part towards efficient Heating, Ventilation and Air Conditioning (HVAC) systems. Data-driven black-box approaches have been an attractive way to develop such models due to their unique feature of not requiring detailed knowledge about the target zone. However, the noisy and non-linear nature of the problem remains a bottleneck especially for long prediction horizons. In this paper, we introduce a Convolutional Neural Networks-Long Short Term Memory (CNN-LSTM) architecture to combine the exceptional feature extraction of convolutional layers with the Long Short Term Memory (LSTM)’s capability of learning sequential dependencies. We experimentally collected a dataset and compared three approaches: Multi-Layer Perceptron (MLP), LSTM and CNN-LSTM. Models are evaluated and compared with 1-30-60-120 min horizons with a closed-loop prediction scheme. The CNN-LSTM outperformed all other models for all prediction horizons and showed a better robustness against error accumulation. It managed to predict room temperature with R 2 > 0.9 in a 120-min prediction horizon.

Causal structure discovery in complex dynamical systems is an important challenge for many scientific domains. Although data from (interventional) experiments is usually limited, large amounts of observational time series data sets are usually available. Current methods that learn causal structure from time series often assume linear relationships. Hence, they may fail in realistic settings that contain nonlinear relations between the variables. We propose Neural Additive Vector Autoregression (NAVAR) models, a neural approach to causal structure learning that can discover nonlinear relationships. We train deep neural networks that extract the (additive) Granger causal influences from the time evolution in multi-variate time series. The method achieves state-of-the-art results on various benchmark data sets for causal discovery, while providing clear interpretations of the mapped causal relations., 11 pages, 5 figures

This paper presents the first comprehensive tutorial on a promising research field located at the frontier of two well-established domains, neurosciences and wireless communications, motivated by the ongoing efforts to define the Sixth Generation of Mobile Networks (6G). In particular, this tutorial first provides a novel integrative approach that bridges the gap between these two seemingly disparate fields. Then, we present the state-of-the-art and key challenges of these two topics. In particular, we propose a novel systematization that divides the contributions into two groups, one focused on what neurosciences will offer to future wireless technologies in terms of new applications and systems architecture (Neurosciences for Wireless Networks), and the other on how wireless communication theory and next-generation wireless systems can provide new ways to study the brain (Wireless Networks for Neurosciences). For the first group, we explain concretely how current scientific understanding of the brain would enable new applications within the context of a new type of service that we dub brain-type communications and that has more stringent requirements than human- and machine-type communication. In this regard, we expose the key requirements of brain-type communication services and discuss how future wireless networks can be equipped to deal with such services. Meanwhile, for the second group, we thoroughly explore modern communication systems paradigms, including Internet of Bio-Nano Things and wireless-integrated brain-machine interfaces, in addition to highlighting how complex systems tools can help bridging the upcoming advances of wireless technologies and applications of neurosciences. Brain-controlled vehicles are then presented as our case study to demonstrate for both groups the potential created by the convergence of neurosciences and wireless communications, probably in 6G. In summary, this tutorial is expected to provide a largely missing articulation between neurosciences and wireless communications while delineating concrete ways to move forward in such an interdisciplinary endeavor.

The fifth generation of mobile networks (5G) and the Software-Defined Radio Access Networks (SD-RAN) architecture envision to support lower latency, enhanced reliability, massive connectivity, and improved energy efficiency. In this context, low latency is considered crucial and Ultra-Reliable Low Latency Communication (URLLC) as one of the key enablers. Currently, IEEE 802.11 networks cannot be programmed fine-grained enough nor manage multiple networks at runtime. Besides, in such scenarios, the coarse-grained level of monitoring information has been hindering troubleshooting and management. In this paper, we present an SDN-based framework where fine-grained End-to-End (E2E) network statistics can be gathered using Inband Network Telemetry (INT) and used for network control and management. With such fine-grained network information, we show how our system can enhance the Quality of Service (QoS) delivery through slice orchestration in IEEE 802.11 Radio Access Networks (RANs).

The number of connected devices has reached 18 billion in 2017 and this will nearly double by 2022, while also new wireless communication technologies become available. Since these modern devices support the use of multiple communication technologies, efforts have been made to enable simultaneous usage and handovers between the different technologies for these devices. However, existing solutions are missing the intelligence to decide on fine-grained (e.g. flow or packet level) optimizations that can drastically enhance the network's performance (e.g., throughput) and user experience. To this extent, we present a multi-technology flow-management load balancing approach for heterogeneous wireless networks that dynamically re-routes traffic through heterogeneous networks, in order to maximize the global throughput. This dynamic approach can be deployed on top of existing solutions and takes into account the specific characteristics of the different technologies, as well as station mobility. We both present a mathematical problem formulation and a heuristic that ensures practical scalability. We demonstrate the heuristic's ability to increase the network-wide throughput by more than 100% across a variety of scenarios and scalability up to 10,000 devices.

Nowadays, shared network infrastructures are being designed for supporting various services that may be subject to diverse failures and disruptions due to the malfunctioning of network component(s) or human-caused disasters. As the delivery of these services may be subject to multiple correlated failures and disruptions, the present chapter develops a reliable model for multi-objective design problems. This model finds applicability in content distribution, storage as a service, computing as a service and their possible combinations (including virtual machine service) as well as resource installation, dimensioning, and allocation for virtual networks. This chapter further elaborates on the motivation underlying the development of such a mixed-integer optimization model from the operational perspective compared to current practices. Indeed, for such problems, the objective function includes in addition to the installation and the transport/delivery cost, terms such as sizing, production, and supply costs. These optimization problems are usually modelled and solved independently though, as proposed in this chapter, it is more realistic and effective to model and solve them simultaneously because related decisions are interdependent. Next, using this generic model, we demonstrate the gain that could be achieved by resilience scheme(s) involving cooperation between the network infrastructure and the service level against conventional protection schemes upon independent and conditionally dependent service-level failure events.

The MAC layer of the IEEE 802.11 standard deploys a CSMA/CA protocol to regulate the access to the shared medium. Not only other IEEE 802.11 stations but also non-IEEE 802.11 devices can be a source of interference, causing collisions and, therefore, re-transmissions, leading to an increased packet latency and a decrease of throughput. As in general, the non-IEEE 802.11 devices do not employ the IEEE8 802.11 CSMA/CA protocol (in particular carrier sensing and the behavior when the medium is sensed busy), the impact of the interference they cause may lead to vital performance degradation of the IEEE 802.11 network. This impact of non-IEEE 802.11 interfering sources on the network performance has not been accurately modeled yet in literature. In this paper, we first characterize a non-IEEE 802.11 interfering source by employing an on-off process. Then we propose, based on earlier results from Bianchi, an analytical model to predict latency and throughput in a saturated as well as an unsaturated network, considering that an interfering source is present. The model utilizes a Markov chain to correctly characterize the behavior of the back-off algorithm of a tagged station in the presence of an interfering source, as well as a Quasi Birth-Death (QBD) process to model the station’s packet queue behavior. We show that we can accurately estimate the impact of non-IEEE 802.11 inference on IEEE 802.11 performance. The model is validated through a comparison with measurements in a real IEEE 802.11 network as well as with an ns3-simulation, whereby a very good agreement is achieved (a difference less than 2%).

Traffic recognition is commonly done using deep packet inspection or packet‐based approaches. However, these methods require the listening device to be part of the network and raise privacy concerns. Traffic recognition models that operate directly at the spectrum level could, for instance, be used for smart spectrum management. To this extent, we present such an approach using deep learning methods. In particular, we present a convolutional neural network architecture that forms the basis for prediction models to recognize different transport protocols, burst traffic with different duty cycles, and different transmission rates. These models are trained with pure synthetic data to lighten the burden of data collection. As such, we validate recent successes in the area of robotics in the context of wireless networks. We compare the performance of two different datasets that contain spectrum images in either time or time‐frequency domain. Our evaluation shows that using time domain data results in an accuracy of at least 96% across all models. Time‐frequency information improves this accuracy even further. Furthermore, a validation with real‐life data shows that it is still possible to discriminate between different transmission rates with an accuracy of around 87%, while the detection of duty cycles and transport protocols takes place with accuracies of, respectively, around 73% and 78%. Finally, we also present a small‐scale real‐life prototype.

Wireless networks are ubiquitous in today’s world and consist of an ever-expanding number of heterogeneous consumer devices and communication technologies. As modern devices support multiple communication technologies, efforts have been made to efficiently manage this plethora of technologies by supporting functionalities such as simultaneous usage or handovers.However, existing solutions are missing both the fine-grained control and intelligence to offer seamless inter-technology management and network optimizations. As a result, technologies operate in an isolated manner, network management is inefficient, and the requirements of users and modern applications are not being met. In contrast, we present intelligent and dynamic multi-technology network management that breaks this isolation, abstracting the connectivity decisions from the user and application level. Different contributions are made: first of all, we introduce a framework for inter-technology management that enables, among others, seamless handovers and packet-based load balancing. Next, we propose different algorithms that can be deployed on top of the novel, or existing, management solutions to increase the network-wide throughput by providing more intelligent network configurations, taking into account mobility and real-time requirements. Finally, as management approaches rely on an accurate overview of the network state, we also consider the monitoring aspect and investigate the detection of traffic patterns in the radio spectrum. Our contributions are evaluated through practical implementations in real-life prototypes.

Low-power wireless mesh networks provide connectivity for a wide range of applications in industrial scenarios. For many years, IEEE 802. 15.4e Time-Slotted Channel Hopping (TSCH) networks have proven their efficiency in such environments, providing high reliability and low-power operation. TSCH networks run on top of one physical (PHY) layer and are thus limited by the characteristics of the chosen PHY layer in terms of, among others, data rate, reliability and energy efficiency. To tackle these limitations and to improve network performance and flexibility in those challenging industrial environments, this work explores the simultaneous use of multiple PHYs, and more specifically multiple modulation and coding schemes (MCSs), in a TSCH network. Traditionally, TSCH relies on fixed-duration slots, large enough to send a packet of any size given the fixed data rate. In order to avoid wasting airtime when simultaneously using multiple PHYs or MCSs with different data rates, we first introduce the concept of slot bonding. This allows the creation of different-sized bonded slots with a duration adapted to the data rate of each chosen PHY. Afterwards, we formally describe TSCH slot bonding using a Mixed Integer Linear Program (MILP) model. Finally, we use this model to determine the optimal MCS configuration with a short slot frame length that causes network saturation and show the scalability advantage of slot bonding in terms of packet delivery ratio.

Spectrum sharing is a reality closer than one might think. Dynamic and intelligent spectrum allocation, where different networks collaborate to optimize spectrum usage jointly, is required to overcome spectrum scarcity. Artificial Intelligence (AI) can play a major role to solve this complex problem in dynamic environments with continuously changing data requirements. A part of the problem can be solved by using smart AI-enabled flow control. Smart flow control can have a major impact on the spectrum footprint of mobile networks where it can optimize the Quality of Service for the own and neighboring networks. This paper presents the architecture and the basic principles of the dynamic spectrum footprint control based on flow prioritization of the SCATTER radio system, a wireless endto-end communication system that participated in the DARPA Spectrum Collaboration Challenge. The flow control mechanism is a policy-based framework.

Recurrent Neural Networks such as Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRUs) are often deployed as neural network-based predictors for time series data. Recently, Hierarchical Temporal Memory (HTM), a machine learning technology attempting to simulate the human brain's neocortex, has been proposed as another approach to time series data prediction. While HTM has gained a lot of attention, little is known about the actual performance compared to the more common RNNs. The only performance comparison between the two, performed at the company behind HTM, shows they perform similarly. In this article, we present a more in-depth performance comparison, involving more extensive hyperparameter tuning and evaluation on more scenarios. Surprisingly, our results show that both LSTM and GRUs can outperform HTM by over 30% at lower runtime. Furthermore, we show that HTM requires explicitly timestamped data to recognize daily and weekly patterns, while LSTM only needs the raw sequential data to predict such time series accurately. Finally, our experiments indicate that the temporally aware components of all considered predictors contribute nothing to the prediction accuracy. We further strengthen this claim by presenting equally or better performing Multilayer Perceptrons conceptually similar to the HTM and LSTM, disregarding their temporal aspects. (C) 2019 Elsevier B.V. All rights reserved.

Many Internet of Things (IoT) applications require a distributed architecture for decision making: either because of a lack of a centralized system, failure-prone connectivity to a centralized system, or because the imposed latency to contact such a system is too high for real-time applications. Often, these IoT applications fall in the domain of Reinforcement Learning (RL), e.g., autonomous robot navigation in Smart Factories and Traffic Signal Control in Smart Cities. However, RL-based applications require a long learning time. To overcome this limitation and scale with the number of agents, Parallel Reinforcement Learning (PRL) algorithms run multiple RL agents in parallel and on distributed environments. However, deploying PRL algorithms in such environments entails a communication overhead that increases the (actual) execution time. State-of-the-art PRL algorithms are designed for reducing the learning time while assuming no (or limited) communication overhead. In this work, we present a novel partitioning algorithm that minimizes the communication overhead in PRL running on IoT environments. To the best of our knowledge, this is the first work that focuses on solving the communication overhead of distributing PRL algorithms algorithm without requiring any a priori knowledge about the structure of the problem. The proposed algorithm intelligently combines a dynamic state partitioning strategy, which exploits the agent’s exploration capabilities to build partition knowledge while learning, with an efficient mapping of agents to partitions, which reduces the communication among agents. Performance evaluations show that the proposed algorithm can achieve almost no communication among PRL agents at the converged state.

Predicting cycling race results has always been a task left to experts with a lot of domain knowledge. This is largely due to the fact that the outcomes of cycling races can be rather surprising and depend on an extensive set of parameters. Examples of such factors are, among others, the preparedness of a rider, the weather, the team strategy, and mechanical failure. However, we believe that due to the availability of historical data (e.g., race results, GPX files, and weather data) and the recent advances in machine learning, the prediction of the outcomes of cycling races becomes feasible. In this paper, we present a framework for predicting future race outcomes by using machine learning. We investigate the use of past performance race data as a good predictor. In particular, we focus on the Tour of Flanders as our proof-of-concept. We show, among others, that it is possible to predict the outcomes of a one-day race with similar or better accuracy than a human.

Wireless network technologies are becoming more and more popular. Because of this, important parts of the wireless spectrum become overloaded. Static spectrum allocation, which has been the norm for decades, is not suitable anymore. To maintain the high demand for spectrum and the continuous development of new wireless technologies, there is a need for an intelligent, dynamic spectrum allocation mechanism, where different network technologies collaboratively optimize the spectrum usage. New wireless network paradigms, such as Neutral Host Networks (NHNs) and private 5G, require a smart, spectrumfootprint-aware flow control algorithm to overcome the spectrum scarcity in collaborative way. This article presents a strategy, vision and flow control mechanism to implement collaboration in a Quality of Service (QoS)-driven way. The solution in this article is based on policies which may activate depending on its current and neighbor’s network states. Through a flow ordering and selection strategy, these policies optimize the spectrum footprint, based on the performance and QoS-requirements of the own and surrounding networks. The proposed algorithm is tested extensively and validated on a large scale during the DARPA Spectrum Collaboration Challenge (SC2) competition. The results of the SC2 final event and intermediate scrimmages showed that the proposed approach increased the score, indicating increased inter-network collaboration was achieved.

Recently, the operation of LTE in unlicensed bands has been proposed to cope with the ever-increasing mobile traffic demand. However, the deployment of LTE in such bands implies sharing spectrum with mature technologies such as Wi-Fi. Several studies have discussed this coexistence problem by suggesting that LTE implements different adaptation mechanisms that allow transmission possibilities to Wi-Fi. While such adaptation mechanisms exist, they still negatively impact Wi-Fi performance, mainly due to the lack of collaboration/coordination mechanisms that inform about the co-located networks’ activities. In this paper, we propose a distributed spectrum management framework that enhances the performance of Wi-Fi, as a particular case, by detecting harmful co-located wireless networks and changes the Wi-Fi’s operating central frequency to avoid them. The framework is based on a Convolutional Neural Network (CNN) that can identify different wireless technologies and provides spectrum usage statistics. Experiments were carried out in a real-life testbed, and the results show that Wi-Fi maintains its performance when using our framework. This translates in an increase of at least 40% on the overall throughput compared to a non-managed operation of Wi-Fi.

Current Software Defined Networking (SDN) techniques allow improving network control and flexibility. However its use in IoT is not trivial because IoT networks are unreliable and highly resource-constrained. Among some of the existing solutions proposed in the literature, Whisper enables SDN-like control over the packet forwarding and cell allocation of IoT devices by injecting in the network artificial, but still standard compliant messages that alter the default protocol behavior. Since Whisper uses carefully computed routing and scheduling messages that are compatible with the distributed protocols run in the network, it reduces the overhead in the network and operates without modifying the IoT devices' firmware. However, as other SDN-on-IoT technologies, Whisper is currently limited to the IoT network scope and remains as yet another independent network management silo. In this paper we propose a new higher-level architecture that allows to fully integrate the IoT SDN network management into a network operating system, such as ONOS, by using Whisper in order to provide an integral endto-end softwarization. We also describe the interaction between the Whisper platform and the orchestrator and test our solution with real 6TiSCH-compatible hardware in the ONOS platform. Finally, we discuss the requirements and technical challenges to fully leverage Whisper to provide an efficient and programmable end-to-end control over an heterogeneous network domain.

Due to the varying conditions in traffic and resource availability in networks nowadays, maintaining continuity of network service and satisfying QoS and QoE requirements became a challenging task. If considered in a highly diversified environment in terms of technology and administration, it gets even more complicated, and appropriate service and resource management solutions are mandatory. Thus, the aim of this paper is to present our specific perspective of the ongoing European H2020 5G-CARMEN project, addressing the importance of proactive reconfiguration of network services and their migration between different domains. Leveraging 5G technology and MEC platform, our management platform for automated low-latency-aware VNF placement and migration will enable orchestration of network services and resources across different administrative and technology domains.

Since the early days of wireless communication, wireless spectrum has been allocated according to a static frequency plan, whereby most of the spectrum is licensed for exclusive use by specific services or radio technologies. While some spectrum bands are overcrowded, many other bands are heavily underutilized. As a result, there is a shortage of available spectrum to deploy emerging technologies that require high demands on data like 5G. Several global efforts address this problem by providing multi-tier spectrum sharing frameworks, for example, the Citizens Broadband Radio Service (CBRS) and Licensed Shared Access (LSA) models, to increase spectrum reuse. In these frameworks, the incumbent (i.e., the technology that used the spectrum exclusively in the past) has to be protected against service disruptions caused by the transmissions of the new technologies that start using the same spectrum. However, these approaches suffer from two main problems. First, spectrum re-allocation to new uses is a slow process that may take years. Second, they do not scale fast since it requires a centralized infrastructure to protect the incumbent and coordinate and grant access to the shared spectrum. As a solution, the Spectrum Collaboration Challenge (SC2) has shown that the collaborative intelligent radio networks (CIRNs) -- artificial intelligence (AI)-based autonomous wireless networks that collaborate -- can share and reuse spectrum efficiently without any coordination and with the guarantee of incumbent protection. In this article, we present the architectural design and the experimental validation of an incumbent protection system for the next generation of spectrum sharing frameworks. The proposed system is a two-step AI-based algorithm that recognizes, learns, and proactively predicts the incumbent's transmission pattern with an accuracy above 95 percent in near real time (less than 300 ms). The proposed algorithm was validated in Colosseum, the RF channel emulator built for the SC2 competition, using up to two incumbents simultaneously with different transmission patterns and sharing spectrum with up to five additional CIRNs.

Industrial Internet of Things (IoT) calls for not only highly reliable, quasi-deterministic and low-power networks, but also for more flexible and programmable networks to cope with operator's dynamics demands. Software Defined Networking (SDN) offers the high levels of flexibility and programmability that traditional distributed protocols cannot offer. In between a fully centralized SDN-on-IoT management solution and a traditional fully distributed one, Whisper stands out as a tradeoff solution that has the robustness, scalability and low-overhead of distributed solutions and the flexibility and programmability of centralized ones. In this demo we present a hands-on experience of how Whisper can be jointly used with traditional SDN solutions, such as ONOS, in order to extend the already existing network programmability in wired domains to 6TiSCH-based Industrial IoT segments. We deploy and test such architecture in real-world large-scale testbeds and demonstrate to be feasible and beneficial to provide an efficient and programmable end-to-end control over a heterogeneous network.

Today's local area networks consist of an ever-expanding number of heterogeneous consumer devices and communication technologies. Despite supporting multiple technologies, those devices tend to connect to the Internet using a single technology, based on predefined priorities. This static behavior does not allow the network to unlock its full potential, which becomes increasingly more important as the quality of service (QoS) requirements of services grow. Moreover, existing approaches make use of theoretical models that assume, unrealistically, full knowledge of the network. To this extent, we present a multi-technology flow-management load balancing framework that dynamically re-routes traffic through heterogeneous networks, in order to maximize the global throughput, based on changing network conditions and QoS demands. Along a problem formulation, we focus on the estimation of wireless and dynamic network characteristics and provide a thorough evaluation through simulations and a prototype implementation. We show that our framework is indeed capable of responding to dynamic network events in real-time and offers an increased overall throughput by optimally using the network's capacity. This results in a throughput increase of around 20% on average.

The recent increase of connected devices has triggered countless Internet-of-Things applications to emerge. By using the Time-Slotted Channel Hopping (TSCH) mode of the IEEE 802.15.4e MAC layer, wireless multi-hop networks enable highly reliable and low-power communication, supporting mission-critical and industrial applications. TSCH uses channel hopping to avoid both external interference and multi-path fading, and a synchronization-based schedule which allows precise bandwidth allocation. Efficient schedule management is crucial when minimizing the delay of a packet to reach its destination. In networks with recurrent sensor data transmissions that repeat after a certain period, current scheduling functions are prone to high latencies by ignoring this recurrent behavior. In this article, we propose a TSCH scheduling function that tackles this minimal-latency recurrent traffic problem. Concretely, this work presents two novel contributions. First, the recurrent traffic problem is defined formally as an Integer Linear Program. Second, we propose the Recurrent Low-Latency Scheduling Function (ReSF) that reserves minimal-latency paths from source to sink and only activates these paths when recurrent traffic is expected. Extensive experimental results show that using ReSF leads to a latency improvement up to 80% compared to state-of-the-art low-latency scheduling functions, with a negligible impact on power consumption of at most 6%.