Wireless communications are characterized by their unpredictability, posing challenges for maintaining consistent communication quality. This article presents a comprehensive analysis of various prediction models, with a focus on achieving accurate and efficient Wi-Fi link quality forecasts using machine learning techniques. Specifically, the article evaluates the performance of data-driven models based on the linear combination of exponential moving averages, which are designed for low-complexity implementations and are then suitable for hardware platforms with limited processing resources. Accuracy of the proposed approaches was assessed using experimental data from a real-world Wi-Fi testbed, considering both channel-dependent and channel-independent training data. Remarkably, channel-independent models, which allow for generalized training by equipment manufacturers, demonstrated competitive performance. Overall, this study provides insights into the practical deployment of machine learning-based prediction models for enhancing Wi-Fi dependability in industrial environments.

Wireless communication is increasingly used in industrial environments, since it supports mobility of interconnected devices. Among the transmission technologies operating in unlicensed bands available to this purpose, Wi-Fi is certainly one of the most interesting, because of its high performance and the relatively low deployment costs. Unfortunately, its dependability is often deemed unsuitable for real-time control systems.In this paper, the use of parallel redundancy is evaluated from a quantitative viewpoint, by considering a number of performance indices that are relevant for soft real-time applications. Analysis is carried out on a large dataset acquired from a real setup, to provide realistic insights on the advantages this kind of approaches can provide. As will be seen, deferred parallel redundancy provides clear advantages in terms of the worst-case transmission latency, at limited costs concerning the amount of consumed spectrum. Hence, it can be practically exploited every time a wireless connection is included in a control loop.

Wi-Fi is currently considered one of the most promising solutions for interconnecting mobile equipment (e.g., autonomous mobile robots and active exoskeletons) in industrial environments. However, relability requirements imposed by the industrial context, such as ensuring bounded transmission latency, are a major challenge for over-the-air communication. One of the aspects of Wi-Fi technology that greatly affects the probability of a packet reaching its destination is the selection of the appropriate transmission rate. Rate adaptation algorithms are in charge of this operation, but their design and implementation are not regulated by the IEEE 802.11 standard. One of the most popular solutions, available as open source, is Minstrel, which is the default choice for the Linux Kernel.In this paper, Minstrel performance is evaluated for both static and mobility scenarios. Our analysis focuses on metrics of interest for industrial contexts, i.e., latency and packet loss ratio, and serves as a preliminary evaluation for the future development of enhanced rate adaptation algorithms based on centralized digital twins.

QR codes are nowadays customarily used for embedding static data such as web hyperlinks or plain text. The sQRy technology (executable QR codes) permits to embed executable programs in QR codes, enabling people to interact with them even without an internet connection.In this work we present QRmap, a specific dialect that permits the inclusion of geographic maps in sQRy and supports interaction with the user to provide indications to reach the destination of interest. The QRmap technology facilitates navigation in large industrial plants where internet connectivity is absent, due to either environmental limitations or company policies. The proposed technology can have interesting applications in non-industrial contexts as well.

Wireless sensor networks (WSNs) are employed across a wide range of industrial applications where ultra-low power consumption is a critical prerequisite. At the same time, these systems must maintain a certain level of determinism to ensure reliable and predictable operation. In this view, time slotted channel hopping (TSCH) is a communication technology that meets both conditions, making it an attractive option for its usage in industrial WSNs.This work proposes the use of machine learning to learn the traffic pattern generated in networks based on the TSCH protocol, in order to turn nodes into a deep sleep state when no transmission is planned and thus to improve the energy efficiency of the WSN. The ability of machine learning models to make good predictions at different network levels in a typical tree network topology was analyzed in depth, showing how their capabilities degrade while approaching the root of the tree. The application of these models on simulated data based on an accurate modeling of wireless sensor nodes indicates that the investigated algorithms can be suitably used to further and substantially reduce the power consumption of a TSCH network.

Precise clock synchronization protocols are increasingly used to ensure that all the nodes in a network share the very same time base. They enable several mechanisms aimed at improving determinism at both the application and communication levels, which makes them highly relevant to industrial environments. Reference Broadcast Infrastructure Synchronization (RBIS) is a solution specifically conceived for Wi-Fi that exploits existing beacons and can run on commercial devices. In this paper, an evolution of RBIS is presented, we call DOMINO, whose coverage area is much larger than the single Wi-Fi infrastructure network, potentially including the whole plant. In particular, wireless stations that can see more than one access point at the same time behave as boundary clocks and propagate the reference time across overlapping networks.

To ensure an unprecedented degree of flexibility, next-generation Industry 4.0/5.0 production plants increasingly rely on mobile devices, e.g., autonomous mobile robots and wearables. In these cases, a major requirement is getting rid of cables through the adoption of wireless networks. To this purpose, Wi-Fi is currently deemed one of the most promising solutions. Achieving reliable communications over the air for distributed real-time control applications is, however, not devoid of troubles. In fact, bounded transmission latency must be ensured for most of the exchanged packets. Moreover, for devices powered on batteries, energy consumption also needs to be taken into account. In this paper, a joint simulated analysis of these aspects is carried out to quantitatively evaluate what we can practically expect from Wi-Fi technology.

Ensuring reliable and predictable communications is one of the main goals in modern industrial systems that rely on Wi-Fi networks, especially in scenarios where continuity of operation and low latency are required. In these contexts, the ability to predict changes in wireless channel quality can enable adaptive strategies and significantly improve system robustness. This contribution focuses on the prediction of the Frame Delivery Ratio (FDR), a key metric that represents the percentage of successful transmissions, starting from time sequences of binary outcomes (success/failure) collected in a real scenario. The analysis focuses on two models of deep learning: a Convolutional Neural Network (CNN) and a Long Short-Term Memory network (LSTM), both selected for their ability to predict the outcome of time sequences. Models are compared in terms of prediction accuracy and computational complexity, with the aim of evaluating their applicability to systems with limited resources. Preliminary results show that both models are able to predict the evolution of the FDR with good accuracy, even from minimal information (a single binary sequence). In particular, CNN shows a significantly lower inference latency, with a marginal loss in accuracy compared to LSTM.

Executable QR codes, also known as sQRy, are a technology aimed at inserting executable programs in a QR code.Through a concrete example, in this paper, we demonstrate their usage in the context of industrial networks in order to assess the operation of a TSN switch by analyzing its status LEDs even in the absence of an internet connection.The entire generation chain that is used to create the sQRy, as well as the corresponding execution chain that, starting from the sQRy, runs it on a mobile device, has been detailed through examples.

Executable QR codes, or sQRy, is a technology dated 2022 that permits to include a runnable program inside a QR code, enabling interaction with the user even in the absence of an Internet connection. sQRy are enablers for different practical applications, including network equipment configuration, diagnostics, and enhanced smart manuals in industrial contexts. Many other non-industry-related fields can also benefit from this technology. Regardless of where SQRy are used, text strings are among the most commonly embedded data. However, due to strict limitations on the available payload, the occupancy of strings limits the length of the programs that can be embedded. In this work, we propose a simple yet effective strategy that can reduce the space taken by strings, hence broadening sQRy applicability.

The radio spectrum is characterized by a noticeable variability, which impairs performance and determinism of every wireless communication technology. To counteract this aspect, mechanisms like Minstrel are customarily employed in real Wi-Fi devices, and the adoption of machine learning for optimization is envisaged in next-generation Wi-Fi 8. All these approaches require communication quality to be monitored at runtime. In this paper, the effectiveness of simple techniques based on moving averages to estimate wireless link quality is analyzed, to assess their advantages and weaknesses. Results can be used, e.g., as a baseline when studying how artificial intelligence can be employed to mitigate unpredictability of wireless networks by providing reliable estimates about current spectrum conditions.

Besides extremely high throughput, Wi-Fi 7 is also aimed at providing users a more deterministic behavior, characterized by shorter average latency and smaller jitters. A key mechanism to achieve this is multi-link operation, which brings simultaneous multi-band communication to client stations as well. In this paper, traffic steering policies are briefly reviewed and grouped into general classes, each one with its advantages and limitations. A basic mechanism for supporting dynamic steering is then described, which is simple enough to allow implementation in real Wi-Fi chipsets but highly flexible at the same time. Its operation can be driven by the host on a per-packet basis, and this permits to optimize spectrum usage depending on the requirements of applications and the traffic pattern they generate.

Modern Artificial Intelligence (AI) research is having a huge impact in many technological domains. As in many other research areas, the application of AI in smart factories has been a key factor in the contribution to the “smartness”. Every aspect of industrial automation has been affected by the introduction of AI: the usage of AI solutions allows to introduce advanced capabilities for optimizing processes, increasing efficiency, and reducing costs. This paper analyzes some relevant aspects of the application and the impact of AI solutions on the current scenario of smart factories and industrial automation. We identify a list of significant topics related to this domain, and we report the main aspects related to them. The dissertation includes an initial quantitative analysis of the relevance of these topics in the scientific publications, a detailed description of the characteristics of the topics, and a discussion of the related challenges.

One of the main goals of wireless sensor networks is to permit the involved nodes to communicate with low energy budgets, as they are typically battery-powered. When such networks are employed in industrial scenarios, constraints about latency may have a significant role, too. The TSCH mechanism, and more in general TDMA schemes, rely on traffic scheduling, and consequently they can feature low power consumption and more predictable latency. Some recent proposals like PRIL-M enable further consistent energy savings, but unfortunately they cause at the same time a dramatic increase in latency. This work presents an extension of PRIL-M, we named PRIL-ML, that achieves a significantly shorter latency in exchange for a slight increase in power consumption. Its operating principles are first illustrated, then some approximate equations are provided for assessing analytically the improvements it achieves, starting from simulation results obtained for both standard TSCH and the original PRIL-M technique.

Executable QR codes, also known as eQR codes or just sQRy, are a special kind of QR codes that embed programs conceived to run on mobile devices like smartphones. Since the program is directly encoded in binary form within the QR code, it can be executed even when the reading device is not provided with Internet access. The applications of this technology are manifold, and range from smart user guides to advisory systems. The first programming language made available for eQR is QRtree, which enables the implementation of decision trees aimed, for example, at guiding the user in operating/maintaining a complex machinery or for reaching a specific location. In this work, an additional language is proposed, we term QRind, which was specifically devised for Industry. It permits to integrate distinct computational blocks into the QR code, e.g., machine learning models to enable predictive maintenance and algorithms to ease machinery usage. QRind permits the Industry 4.0/5.0 paradigms to be implemented, in part, also in those cases where Internet is unavailable.

Multi-Link Operation (MLO) in Wi-Fi 7 is expected to tangibly boost throughput while lowering transmission latency at the same time. This is very relevant in industrial scenarios and makes MLO suitable, e.g., to support seamless device mobility. Benefits depend on the ability of multi-link devices to select at run-time the best link, among the available ones, in order to maximize both communication performance and reliability.In this paper an experimental platform is proposed, with the aim of leveraging commercial hardware and open source software, and easing prototyping and evaluation of MLO techniques. The platform has been employed to analyze the transmission quality of two pairs of non-overlapping channels, and in particular to assess whether or not adequate diversity is provided, so that those channels can be exploited to improve reliability. Results point out that correlation between different links is, in most cases, limited, which makes MLO a valuable approach.

The next generation of Wi-Fi is meant to achieve ultra-high reliability for wireless communication. Several approaches are available to this extent, some of which are being considered for inclusion in standards specifications, including coordination of access points to reduce interference.In this paper, we propose a centralized architecture based on digital twins, called WiTwin, with the aim of supporting wireless stations in selecting the optimal association according to a set of parameters. Unlike prior works, we assume that Wi-Fi 7 features like multi-link operation (MLO) are available. Moreover, one of the main goals of this architecture is to preserve communication quality in the presence of mobility, by helping stations to perform reassociation at the right time and in the best way.

Wireless sensor networks are characterized by low power consumption because motes are typically battery-powered. Time slotted channel hopping (TSCH) relies on a fixed transmission schedule, which enables the receiver module of wireless motes to be switched off every time it is not needed. Unfortunately, in many practical contexts most of the reserved slots remain unused, which leads to appreciable energy waste. For periodic traffic, proactive reduction of idle listening (PRIL) techniques have been proven able to mitigate this problem. In this paper, PRIL multi-hop (PRIL-M) is introduced with the aim to improve existing PRIL techniques, by lowering energy waste further in large real-world mesh networks. PRIL-M is advantageous in all those contexts where ultra-low power consumption is more important than end-to-end latency. Applications that can benefit from PRIL-M include, e.g., environmental monitoring, where sensors are deployed over the target area and must operate for years without maintenance. A thorough simulation campaign showed that, in these scenarios, energy consumption of PRIL-M is 75% less than standard TSCH, while the average latency is about 20 times larger.

QR codes are increasingly used in a plurality of scenarios, and research activities are being successfully carried out to improve this technology and widen its contexts of applicability. After an extensive survey of the state-of-the-art on the subject, this work presents the new, promising possibility to embed a programming language in a QR code. This new kind of executable QR codes, we named eQR codes, enable interaction with end users even in the absence of an Internet connection, and provide a sort of IoT paradigm where intelligence is embedded in the object tag in the form of a program. Among all the possible languages that can be embedded, this work focuses on a powerful but compact (in terms of QR code storage occupation) dialect, termed QRtree, which is aimed at implementing decision trees. The eQR code technology makes a new class of applications possible, e.g., providing hints for navigation or instructions for using rescue devices in places with no network coverage like mountains and caves. Smart interactive user manuals are enabled as well. Besides defining the QRtree language and eQR code structure, this paper describes all the steps needed to generate eQR codes and to manage their execution in end-user devices. A simple yet realistic example and the related code are also presented, to practically show how this technology can be used to solve real-world problems. For the example, the QRtree version of the code takes 234B, less than one-half the size of an equivalent program in Python bytecode (634B).

Nowadays, large enterprises, critical infrastructures, and utilities are complex systems, and consist of a number of sub-systems, often located far away from each other, that coordinate their operations over geographic networks. Typically, they are interconnected through the Internet and communicate by means of TCP/IP-based protocols. When system-level resilience is required, extremely high dependability is demanded from the network. In this paper a solution is described and evaluated that loosely resembles SDN architectures. A redundant version of MQTT is used for the data plane, whereas an adaptive mechanism is implemented in the control plane that evaluates the quality of paths and dynamically selects the best choice. To maximize dependability, yet keeping resource consumption to acceptable levels, concepts borrowed from Reinforcement Learning are exploited. An experimental campaign corroborated our expectations and showed the practical feasibility of our proposal.

The ability to predict the behavior of a wireless channel in terms of the frame delivery ratio is quite valuable, and permits, e.g., to optimize the operating parameters of a wireless network at runtime, or to proactively react to the degradation of the channel quality, in order to meet the stringent requirements about dependability and end-to-end latency that typically characterize industrial applications.In this work, prediction models based on the exponential moving average (EMA) are investigated in depth, which are proven to outperform other simple statistical methods and whose performance is nearly as good as artificial neural networks, but with dramatically lower computational requirements. Regarding the innovation and motivation of this work, a new model that we called EMA linear combination (ELC), is introduced, explained, and evaluated experimentally. Its prediction accuracy, tested on some databases acquired from a real setup based on Wi-Fi devices, showed that ELC brings tangible improvements over EMA in any experimental conditions, the only drawback being a slight increase in computational complexity.

By removing wire harness, Wi-Fi is becoming increasingly pervasive in every aspect of our lives, in both the consumer and industrial worlds. Besides flexibility, the recent high efficiency and extremely high throughput versions managed to close the performance gap with Ethernet. However, it still lags behind Ethernet for what concerns dependability. To this aim, the ultra high reliability study group has been recently formed. This paper reports on some preliminary ideas and proposals about the ways seamless redundancy can be exploited to make Wi-Fi more reliable, yet retaining a good degree of backward compatibility with existing network infrastructures.

New generation electrified and self-driving vehicles require much higher performance and flexibility for onboard digital communications than Controller Area Networks may offer. For this reason, automotive Ethernet is often regarded as the next de facto standard technology in these contexts, and by extension for networked embedded systems as well. However, an abrupt and drastic move from CAN to Ethernet is likely to cause further cost increases, which can be hardly tolerated by buyers. This paper analyzes the third generation of CAN, termed CAN XL, and studies how interoperability can be ensured with Ethernet. Likely, composite CAN XL-Ethernet networks are the key for getting the best of both worlds, not only in the automotive domain but also for sensing and control in scenarios like building automation, wired sensor networks, and low-cost networked embedded systems with real-time constraints.

The ability to reliably predict the future quality of a wireless channel, as seen by the media access control layer, is a key enabler to improve performance of future industrial networks that do not rely on wires. Knowing in advance how much channel behavior may change can speed up procedures for adaptively selecting the best channel, making the network more deterministic, reliable, and less energy-hungry, possibly improving device roaming capabilities at the same time. To this aim, popular approaches based on moving averages and regression were compared, using multiple key performance indicators, on data captured from a real Wi-Fi setup. Moreover, a simple technique based on a linear combination of outcomes from different techniques was presented and analyzed, to further reduce the prediction error, and some considerations about lower bounds on achievable errors have been reported. We found that the best model is the exponential moving average, which managed to predict the frame delivery ratio with a 2.10% average error and, at the same time, has lower computational complexity and memory consumption than the other models we analyzed.

Lifetime of motes in wireless sensor networks can be enlarged by decreasing the energy spent for communication. Approaches like time slotted channel hopping pursue this goal by performing frame exchanges according to a predefined schedule, which helps reducing the duty cycle. Unfortunately, whenever the receiving radio interface is active but nobody in the network is transmitting, idle listening occurs. If the traffic pattern is known in advance, as in the relevant case of periodic sensing, proactive reduction of idle listening (PRIL) noticeably lowers energy waste by disabling receivers when no frames are expected for them. Optimal PRIL operation demands that, at any time, the transmitter and receiver sides of a link have a coherent view of its state (either enabled or disabled). However, this is not ensured in the presence of acknowledgment frame losses. This paper presents and analyzes some strategies to cope with such events. An extensive experimental campaign has been carried out through discrete event simulation to determine what consequences above errors may have from both a functional and performance viewpoint. Results show that, although no strategy is optimal in all circumstances, different solutions can be profitably adopted depending on the specific operating conditions.

Embedding a programming language in a QR code is a new and extremely promising opportunity, as it makes devices and objects smarter without necessarily requiring an Internet connection. In this paper, all the steps needed to translate a program written in a high-level programming language to its binary representation encoded in a QR code, and the opposite process that, starting from the QR code, executes it by means of a virtual machine, have been carefully detailed. The proposed programming language was named QRscript, and can be easily extended so as to integrate new features.One of the main design goals was to produce a very compact target binary code. In particular, in this work we propose a specific sub-language (a dialect) that is aimed at encoding decision trees. Besides industrial scenarios, this is useful in many other application fields. The reported example, related to the configuration of an industrial networked device, highlights the potential of the proposed technology, and permits to better understand all the translation steps.

Communication quality of wireless links unavoidably varies over time, for a number of reasons that are mostly unknown in the design phase of distributed systems. This is a severe limitation when such network technologies are employed to connect devices and subsystems in industrial applications, where high reliability and timeliness are customarily demanded. The ability to foresee, to some extent, the quality of a link in the near future is likely to provide clear benefits in many such cases.In this paper, we analyzed and compared the performance of two approaches, based on simple moving averages and artificial neural networks, respectively, to carry out predictions for a real Wi-Fi link. Results confirm that reliable estimates about the packet delivery ratio can be obtained in realistic operating conditions by averaging the recent past, and that the use of machine learning may improve prediction further.

Controller Area Network is still the most popular solution for in-vehicle communications in automotive scenarios. Recent solutions like CAN FD and CAN XL manage to reduce the performance gap with Ethernet-based solutions. Nevertheless, the priority-based access mechanism they rely on is not completely adequate to support fragmented transfers of large blocks of data, as happens, e.g., for driver assistance and the Internet of Things. In this paper, a slight modification is proposed for the basic CAN access mechanism that permits a group of objects to be served by the network with about the same quality of service. This provides fair treatment to a number of block transfers initiated by different nodes and taking place contextually over the bus.

The use of white and black listing techniques in Wireless Sensor Networks (WSN), and in particular those which are based on the Time Slotted Channel Hopping (TSCH) op-erating mode of IEEE 802.15.4, permits to improve reliability and latency by performing transmissions on the best channels. Techniques that operate on a per-link basis are deemed quite effective, but proper operation requires that the two end points involved in the communication agree on the channels to be used for transmission. On the contrary, communication in the network can be prevented, eventually leading, in the worst cases, to the disconnection of part of the nodes. This paper presents CONSIP, a technique aimed to ensure strict consistency in the information exchanged between the nodes and used to drive communication, by preventing a priori the aforementioned problem from occurring. Results show a slight increase in energy consumption, due to the use of a backup cell, whereas communication latency does not worsen. The effectiveness of CONSIP was assessed by means of an experimental campaign, and the only drawback we found is that the backup cell, which is required to be reserved per link, may limit the number of nodes in dense networks.

Among the application layer protocols that support the Internet of Things, a very popular option is MQTT. Its main advantages reside in the inherent simplicity, which makes it suitable for embedded devices too, and the publisher-subscriber model it relies on, which decouples data sources and sinks and enables efficient multicast streaming over the Internet. In this paper a proposal is described that exploits redundant MQTT brokers, deployed in different geographic locations, whose aim is to provide uninterrupted service in spite of failures that may be affecting both the servers executing the middleware and the Internet core. Unlike routing algorithms and protocols, seamless redundancy features zero intervention time, making this solution suitable for critical infrastructures where timings are relevant, like public utilities and some large enterprises.

Saving motes energy as much as possible is a main requirement in wireless sensor networks. Time Slotted Channel Hopping (TSCH) is an effective solution, as the radio module of receivers is enabled only when a frame reception is scheduled. Unfortunately, this often leads to non-negligible idle listening, i.e., the receiving interface is switched on but no incoming frame is detected. In cases where the traffic pattern can be somehow predicted, Proactive Reduction of Idle Listening (PRIL) techniques can reduce energy wastes significantly by temporarily switching receivers off. In this paper two specific strategies are proposed, which are aimed at improving PRIL robustness in the case acknowledgment frames are lost, hence causing the transmitter and receiver to have different perceptions of the link state. They were encoded by means of a discrete event simulator and their behavior analyzed. Results obtained through an extensive experimental campaign show that there is no clear winner. Instead, each proposed strategy is suitable for specific application contexts.

The past few decades saw factory and process automation move from monolithic to fully distributed architectures, where the communication system plays a key role in providing higher flexibility and availability, yet fitting the peculiar requirements of such contexts. As part of the Industry 4.0 (r)evolution, the Industrial Internet of Things is meant to achieve full convergence toward well-agreed non-proprietary digital networking technologies. In this scenario, OPC UA is becoming increasingly relevant as a high-level protocol to exchange information in heterogeneous distributed industrial applications. In this paper, the performance of OPC UA has been evaluated by means of an experimental measurement campaign carried out on real devices connected to real networks. In particular, we analyzed how much the end-to-end notification latency achieved by the subscription mechanism is affected by the position and interconnection of devices, their specific implementation, and configuration parameters. Outcomes can be helpful in designing systems where industrial devices interact with local workstations, the edge, and the cloud, to understand whether or not timing constraints of applications will be met.

Among the many enabling technologies Industry 4.0 relies on, the Industrial Internet of Things is a relevant one that enables both horizontal and vertical integration in modern enterprises, to accomplish complex functions that exploit either cloud or edge processing. In particular, the OPC UA protocol is knowing an ever increasing attention for interconnecting devices, equipment, and facilities in next-generation plants. This work focuses on the characterization of the notification latency in distributed systems that rely on the OPC UA publish-subscribe paradigm for interactions. In particular, starting from a black-box experimental evaluation carried out on real setups, we tried to explain the seemingly peculiar behavior of some popular solutions by means of theoretical models complemented with code inspection.

Time slotted channel hopping provides reliable and deterministic communication in IEEE 802.15.4 mesh networks. Although slotted access is able to lower energy consumption drastically by reducing the duty cycle of the radio module, it usually leads to significant idle listening experienced by receivers, which makes it a sub-optimal solution when ultra low-power wireless is sought for.In this paper a listening suspension mechanism is described, which operates at the MAC layer and is part of a more general approach aimed at cutting down energy consumption by proactively reducing idle listening. Links can be temporarily disabled, that convey slow-rate data streams whose characteristics, e.g., the generation period, are either known in advance to some extent or can be inferred by traffic inspection.

TSCH is emerging as a reliable access mechanism for wireless sensor networks that, thanks to time slotting, suits the needs of battery-powered devices deployed in industrial scenarios. To take into account the traffic needs of the different links in a mesh network and the related variability, in such a way to prevent instability and butter overruns, overprovisioning can be exploited. It consists in allocating more cells than strictly needed for communication between a node and its neighbors. Unfortunately, overprovisioning in TSCH leads to higher power consumption, because of idle listening. In this paper, two approaches for saving energy are proposed, which temporarily disable listening when no packets are available at the transmitter. Such mechanisms show good backward compatibility with existing TSCH devices, and may coexist with them.

This special section on heterogeneous industrial networks of the current and next-generation factories tackles a wide range of the issues mentioned above. Its main objective is both bringing together ideas from the worldwide research community about the directions in which the industrial networks’ development is headed and presenting the latest advances in the works related to industrial communication technologies to be potentially used in the future. All these concepts have to be evaluated keeping in mind that the evolution of these technologies is a fluid and continuous process. Probably, in the near future, industrial networks will be more and more heterogeneous for the introduction of new technologies with even more variegated characteristics, and application boundaries will be able to space from a single room to the whole world. Only a few articles are currently addressing intelligent and self-adaptive networks, and some of these works are included in this Special Section. Certainly, the introduction of relatively new paradigms for industrial communication such as SDN is the first important step for a network of the future on which to lay intelligent algorithms. These paradigms will have to also be extended to the wireless counterpart, bearing in mind that the adaptability to the intrinsic variability of the wireless medium and nodes mobility will remain research challenges for a long time. The eight articles selected for publication in this special section are briefly summarized here.

Industrial Internet of Things demands high reliability and determinism on the one side, and easy interconnection with the existing network infrastructures on the other. A very promising class of solutions combines IEEE 802.15.4 Time Slotted Channel Hopping, to meet above communication requirements, and IPv6, to provide universal connectivity. Products compliant to this approach have recently begun to appear off-the-shelf.In this paper, two such solutions have been taken into account, namely SmartMesh IP and 6TiSCH. Their performance has been experimentally evaluated by taking into account metrics relevant to industrial applications. Since we aimed at providing a practical feedback to designers of industrial systems, devices were tested using their default configuration. Results provide valuable experimental figures on what engineers can expect from these technologies, which includes seamless connectivity and a reasonably high resilience to interfering traffic.

The Industrial Internet of Things is currently gaining momentum in production plants as a way to efficiently support a number of functions, e.g., preventive maintenance. To this extent, a number of solutions have been proposed in the past few years, which typically rely on a layered approach. In this paper, a general-purpose architecture is presented for factory automation systems, which mimics the industrial networks used at the shop-floor. In particular, it relies on a modular structure for devices and protocols aimed at reducing implementation effort, so as to ease its adoption in real-world applications. A preliminary experimental testbed based on such architecture has been set up, and it operations verified.

In distributed control systems where devices are connected through Wi-Fi, direct access to low-level medium access control (MAC) operations may help applications to meet their timing constraints. In particular, the ability to timely control single transmission attempts on air, by means of software programs running at the user space level, eases the implementation of mechanisms aimed at improving communication timeliness and reliability. Relevant examples are deterministic traffic scheduling, seamless channel redundancy, rate adaptation algorithms, and so on. In this paper, a novel architecture is defined, we call software-defined MAC (SDMAC), which in its current embodiment relies on conventional Linux PCs equipped with commercial Wi-Fi adapters. Preliminary SDMAC implementation on a real testbed and its experimental evaluation showed that integrating this paradigm in the existing protocol stacks constitutes a viable option, whose performance suits a wide range of applications characterized by soft real-time requirements.

Seamless redundancy layered atop Wi-Fi has been shown able to tangibly increase communication quality, hence offering industry-grade reliability. However, it also implies much higher network traffic, which is often unbearable as the wireless spectrum is a shared and scarce resource. To deal with this drawback, the Wi-Red proposal includes suitable duplication avoidance mechanisms that reduce spectrum consumption by preventing transmission on air of inessential frame duplicates. In this paper, the ability of such mechanisms to save wireless bandwidth is experimentally evaluated. To this purpose, specific post-analysis techniques have been defined, which permit assessment on a simple testbed that relies on plain redundancy and require no changes to the adapters' firmware. As the results show, spectrum consumption decreases noticeably without communication quality is impaired. Further saving can be obtained if a slight worsening is tolerated for latencies.

Applying link-level seamless redundancy to Wi-Fi noticeably ameliorates communication quality, but also doubles spectrum consumption. The adoption of mechanisms, operating at runtime, aimed at preventing the transmission on air of inessential frame copies, permits to lower the generated traffic tangibly and also improves behavior further. In this paper, a prototype setup is described, based on redundant Wi-Fi, which carries out reactive duplication avoidance completely in software. This makes it mostly independent from wireless adapters hardware. An experimental evaluation of its performance showed that, although implementation is preliminary, this solution is able to outperform plain seamless redundancy.

Reliability and determinism of Wi-Fi can be tangibly improved by means of seamless redundancy, to the point of making this technology suitable for industrial environments. As pointed out in recent papers, the most benefits can be achieved when no phenomena can simultaneously affect transmissions on all channels of a redundant link. In this paper, several aspects are analyzed which, if not properly counteracted, may worsen seamless redundancy effectiveness. Effects they cause on communication have been experimentally evaluated in real testbeds, which rely on commercial Wi-Fi devices. Then, practical guidelines are provided, which aim at preventing joint interference through a careful system design. Results show that measured communication quality can be made as good as expected in theory.

Seamless redundancy can be profitably exploited to improve predictability of wireless networks in general and, in particular, IEEE 802.11. According to this approach, packets are transmitted by senders on two (or more) channels at the same time and duplicate copies are discarded by receivers. As long as the behavior of physical channels is uncorrelated, communication quality improves noticeably, in terms of both transmission latencies and percentage of dropped frames. In this paper, communication over redundant links has been analyzed by means of a thorough experimental campaign, based on measurements carried out on real devices. Results confirm that, under typical operating conditions, the assumption of independence among channels in properly designed systems is verified reasonably well. Indeed, in our experiments, measured link quality indices did not differ more than$10\%$from what we expected from theory. This grants for redundant solutions tangible advantages over conventional Wi-Fi networks.

Communication over the ether is by nature erratic, and thus wireless networks are often deemed unsuitable to support time-critical distributed control applications, like those at the shop-floor of industrial plants. However, scheduling techniques derived from real-time operating systems, along with channel redundancy and runtime retransmission management, can be profitably applied to wireless networks to improve their determinism tangibly. This paper presents a novel approach that combines such solutions. A preliminary prototype implementation, which relies on IEEE 802.11 (Wi-Fi), has been set up to assess feasibility and performance of the approach. Results show that determinism of data exchange can be improved to a degree that makes wireless networks actually suitable for many soft real-time applications.

Seamless redundancy has been recently proven to increase the quality of wireless links communication noticeably. As such, it can be exploited to improve Wi-Fi reliability and make it suitable for industrial environments. Higher benefits can be achieved when phenomena affecting transmissions on the two channels of a redundant link are independent. In this paper, a number of aspects that may lead to a reduction of the effectiveness of seamless redundancy are analyzed. In particular, the quality of communication has been experimentally evaluated in a real testbed made up of conventional Wi-Fi devices. Guidelines are then provided, which are aimed at preventing joint interference through a careful system design.

The current version of IEEE 802.11, based on EDCA, is unable to support fine-grained assignments of priorities to messages. Instead, such an ability could be advantageous in distributed real-time control systems communicating over the air, as it would enable the exploitation of feasibility analysis to assess whether or not timing constraints are met. In this paper, a proposal is made for implementing a simple fixed-priority access scheme on conventional, commercial Wi-Fi devices. Basically, this is obtained by disabling the random backoff mechanism and suitably managing interframe gaps only. In particular, each real-time message is assigned a unique gap duration, so as to help in preventing collisions. An experimental campaign has been carried out on real devices aimed at assessing the performance this approach can achieve.

Communication over the ether is by nature erratic and, consequently, wireless networks are unsuitable for hard real-time distributed control applications typical of industrial plants. However, scheduling strategies employed by real-time operating systems, along with channel redundancy and advanced bandwidth management, can be used to improve their determinism and make them suitable for soft real-time applications. This paper presents an approach that combines such techniques and a preliminary implementation, which relies on IEEE 802.11 (Wi-Fi), employed to assess its feasibility. Results show that data exchange determinism can be substantially improved towards the level required by many soft real-time applications.

Seamless redundancy can be profitably exploited to improve predictability of wireless networks, including IEEE 802.11. Packets are sent on two channels at the same time and duplicate copies are discarded by receivers. As long as behavior of channels is reasonably uncorrelated, both transmission latencies and the frame loss ratio improve noticeably. In this paper, such an assumption has been analyzed by means of experimental measurements carried out on real devices. Results confirm that, in typical operating conditions, independence reasonably holds.

The adoption of wireless communications and, in particular, Wi-Fi, at the lowest level of the factory automation hierarchy has not increased as fast as expected so far, mainly because of serious issues concerning determinism. Actually, besides the random access scheme, disturbance and interference prevent reliable communication over the air and, as a matter of fact, make wireless networks unable to support distributed real-time control applications properly. Several papers recently appearing in literature suggest that diversity could be leveraged to overcome this limitation effectively. In this paper, a reference architecture is introduced, which describes how seamless link-level redundancy can be applied to Wi-Fi. The framework is general enough to serve as a basis for future protocol enhancements, and also includes two optimizations aimed at improving the quality of wireless communication by avoiding unnecessary replicated transmissions. Some relevant solutions have been analyzed by means of a thorough simulation campaign, in order to highlight their benefits when compared with conventional Wi-Fi. Results show that both packet losses and network latencies improve noticeably.

Worst-case synchronization error is probably the most interesting performance index of clock synchronization protocols when used in real-time control applications. It depends on several factors, among which jitters on timestamp measurements and thermal variations in oscillators, which are very difficult to replicate in real systems. As a consequence, it makes little sense comparing the errors measured in different experiments, which means that assessing synchronization quality in real-time systems in a meaningful way is not a trivial task. In this paper, a methodology is described that permits different clock discipline algorithms to be experimentally evaluated and reliably compared. It was applied to a testbed where the reference broadcast infrastructure synchronization protocol is implemented over Wi-Fi, and permitted to define some guidelines for the optimal selection of the algorithm and its parametrization.

This paper describes reference broadcast infrastructure synchronization (RBIS), a clock synchronization protocol for IEEE 802.11 infrastructure wireless networks. The protocol is especially tailored for industrial and home automation networks, and in many application contexts, it offers several advantages compared with other solutions targeted at similar purposes. RBIS has been conceived to rely on conventional Wi-Fi equipment and, in particular, on unmodified access points. It is based on the master/slave approach and follows the receiver/receiver paradigm. An implementation of RBIS-carried out completely in software and based on timestamps taken at the interrupt handler level-has been developed, which achieves a synchronization error below 3 μs. Then, a simple distributed hard real-time control application has been setup, which consists in two PCs running real-time application interface for Linux (RTAI) and connected through Wi-Fi. The actuation error, measured on the generation of synchronous pulses, is strictly below 13 μs.

Wireless communications suffer from disturbance and interference, which prevent reliable data exchanges and preclude their adoption in the real-time applications found at the lower levels of factory automation systems. This is particularly true for Wi-Fi, for which mechanisms like channel hopping and blacklisting are not available. Several solutions were defined in the past few years, which range from the adoption of general-purpose redundancy protocols over conventional communication equipment to proposals like Wi-Red, which offers link-level redundancy. In this paper, an enhancement is presented for the latter solution, which achieves better performance by deferring transmission of duplicate frames. The net effect is a sort of load balancing between wireless channels, which reduces network traffic and increases reliability further.

In this paper the RBIS synchronization protocol is presented, which has been specifically conceived for industrial infrastructure WLANs. It combines the concepts of the sender/receiver and receiver/receiver approaches: similarly to RBS, timestamps are recorded only on received frames while, like IEEE 1588, it permits slaves to synchronize to the same master clock. Compared to RBS, RBIS has several advantages. Every single wireless sub-network, in fact, can be easily synchronized with an external time source located on the wired backbone. Moreover, when software implementations are considered that rely on unmodified conventional WiFi equipment, RBIS is able to feature noticeably better timing accuracy than IEEE 1588.

Recently, the open-source EtherCAT master developed by the IgH company, has been enhanced with the introduction of the EtherCAT library, which allows to develop control applications in user space. This feature makes it possible to build complex applications that need to use the floating point unit or existing user space libraries. Unfortunately, the EtherCAT library does not seem to be able to guarantee a high degree of determinism. In this paper, first the real-time performance of the user space implementation of the EtherCAT master is evaluated and compared with the one that can be obtained by implementing the same control application at the kernel level. Then, we propose an alternative hard real-time approach based on RTAI, that allows to exploit all the advantages of user space development, with a negligible impact on performance.

EtherCAT is a real-time Ethernet protocol conceived explicitly for industrial applications. It is characterized by high communication efficiency, which permits control loops to be closed with short cycle times, and is provided with a suitable mechanism, known as distributed clock (DC), that enables synchronized operations to take place across the controlled system. These features can be profitably adopted, for instance, to support motion control applications. In this paper, the performance of the DC mechanism is evaluated by means of a thorough campaign of experimental measurements carried out on a real network setup. A number of factors have been taken into account that can affect accuracy and precision, and their effects studied in depth.

At present, two important trends in industrial environments are the adoption of Ethernet-based technologies to interconnect devices and the use of PCs provided with open-source real-time operating systems to execute control applications. This makes the implementation of distributed control systems easier and inexpensive. In these cases, it is of utmost importance that response times concerning the messages exchanged over the network be deterministically bounded. In this paper, the contribution to the end-to-end transmission delay which is due to the latencies introduced by the local protocol stack is analysed through experimental evaluation. In particular, two popular stacks have been taken into account, that is the one provided in conventional Linux and RTnet.

The adoption of open-source operating systems for the execution of real-time applications is gaining popularity, even in the networked control systems domain, due to cost and flexibility reasons. However, as opposed to their commercial counterparts, the actual performance level to be expected from them is still little known and may often depend on the kind of real-time extension being used, its configuration, and the overall software load of the system, including best-effort components. In this paper, an open-source EtherCAT master supported by a popular real-time extension for Linux, the RT Patch, is thoroughly evaluated with long-term measurements, which build confidence on the suitability of the proposed approach for real-world applications. Special attention is devoted to the unexpected, adverse effect that some best-effort components, for instance, graphics applications, may have on the overall real-time characteristics of the system. For reference, the proposed approach is also compared with RTAI, a more traditional and well-known real-time extension for Linux already in use for demanding applications.

At present, an increasing number of distributed control systems are based on platforms made up of conventional PCs running open-source real-time operating systems. Often, the need arises in these systems to have networked devices supporting synchronized operations. In this paper, an inexpensive solution is introduced, described, implemented and evaluated that relies on standard software and protocols such as RTAI, RTnet and IEEE 1588. The main goals of this architecture are reducing design and development costs, ensuring adequate synchronization accuracy, and easing the porting of control applications to different H/W and S/W configurations.

EtherCAT is a popular, real-time Ethernet network that features very high communication efficiency. However, because of the ring topology it relies on, cycle times increase when the network size grows larger (considering both the amount of process data and the network extension). In this paper a solution based on commonly available off-the-shelf switches is presented and analyzed that can improve performance for large automation systems. Such an approach (almost) does not require new custom h/w to be purposely developed.

In this paper we describe the implementation of a complete ANN training procedure for speech recognition using the block mode back-propagation learning algorithm. We exploit the high performance SIMD architecture of GPU using CUDA and its C-like language interface. We also compare the speed-up obtained implementing the training procedure only taking advantage of the multi-thread capabilities of multi-core processors. Our approach has been tested by training acoustic models for large vocabulary speech recognition tasks, showing a 6 times reduction of the time required to train real-world large size networks with respect to an already optimized implementation using the Intel MKL libraries.

In this paper, we propose a word confidence measure based on phone durations depending on large contexts.The measure is based on the expected duration of each recognized phone in a word.In the approach here proposed the duration of each phone is in principle context-dependent, and the measure is a function of the distance between the observed and expected phone duration distributions within a word.Our experiments show that, since the "duration confidence" does not make use of any acoustic information, its Equal Error Rate (EER) in terms of False Accept and False Rejection rates is not as good as the one obtained by using the more informed acoustic confidence measure.However, combining the two measures by a simple linear interpolation, the system EER improves by 6% to 10% relative on an isolated word recognition task in several languages.

This paper presents a front-end consisting of an Artificial Neural Network (ANN) architecture trained with multilingual corpora.The idea is to train an ANN front-end able to integrate the acoustic variations included in databases collected for different languages, through different channels, or even for specific tasks.This ANN front-end produces discriminant features that can be used as observation vectors for language or task dependent recognizers.The approach has been evaluated on three difficult tasks: recognition of non-native speaker sentences, training of a new language with a limited amount of speech data, and training of a model for car environment using a clean microphone corpus of the target language and data collected in car environment in another language.

Training Artificial Neural Networks (ANNs) with large amounts of speech data is a time intensive task due to the intrinsically sequential nature of the back-propagation algorithm.This paper presents an approach for training ANNs using sentence and frame selection.The goal is to speed-up the training process, and to balance the phonetic coverage of the selected frames, trying to mitigate the classification problems related to the prior probabilities of the individual phonetic classes.These techniques, together with a three-step training approach and software optimizations, reduced by an order of magnitude the training time of our models.

A technique is proposed for the adaptation of automatic speech recognition systems using hybrid models combining artificial neural networks with hidden Markov models. The application of linear transformations not only to the input features, but also to the outputs of the internal layers is investigated. The motivation is that the outputs of an internal layer represent a projection of the input pattern into a space where it should be easier to learn the classification or transformation expected at the output of the network. A new solution, called conservative training, is proposed that compensates for the lack of adaptation samples in certain classes. Supervised adaptation experiments with different corpora and for different adaptation types are described. The results show that the proposed approach always outperforms the use of transformations in the feature space and yields even better results when combined with linear input transformations

Many techniques for speaker or channel adaptation have been successfully applied to automatic speech recognition. Most of these techniques have been proposed for the adaptation of hidden Markov models (HMMs). Far less proposals have been made for the adaptation of the artificial neural networks (ANNs) used in the hybrid HMM-ANN approach. This paper presents an adaptation technique for ANNs that, similar to the framework of MAP estimation, tries to exploit in the adaptation process prior information that is particularly useful to deal with the problem of sparse training data. We show that the integration of a priori information can be simply achieved by linear interpolation of the weights of an "a priori" network and of a speaker specific network. Good improvements with respect to the baseline results are reported evaluating this technique on the Wall Street Journal WSJ0 and WSJ1 databases and on TIMIT corpus using different amounts of adaptation data

In connectionist learning, one relevant problem is "catastrophic forgetting" that may occur when a network, trained with a large set of patterns, has to learn new input patterns, or has to be adapted to a different environment. The risk of catastrophic forgetting is particularly high when a network is adapted with new data that do not adequately represent the knowledge included in the original training data. Two original solutions are proposed to reduce the risk that the network focuses on new data only, loosing its generalization capability. The first one, conservative training, is a variant to the target assignment policy, while the second approach, support vector rehearsal, selects from the training set the patterns that lay near the borders of the classes not included in the adaptation set. These patterns are used as sentinels that try to keep unchanged the original boundaries of these classes. Moreover, we investigated the extension of the classical approach consisting in applying linear transformations not only to the input features, but also to the outputs of the internal layers. The motivation is that the outputs of an internal layer represent a projection of the input pattern into a space where it should be easier to learn the classification or transformation expected at the output of the network. We illustrate the problems using an artificial test-bed, and apply our techniques to a set of adaptation tasks in the domain of automatic speech recognition (ASR) based on artificial neural networks. Supervised ASR adaptation experiments with several corpora and for different adaptation types are described. We report on the adaptation potential of different techniques, and on the generalization capability of the adapted networks. The results show that the combination of the proposed approaches mitigates the catastrophic forgetting effects, and always outperforms the use of the classical transformations in the feature space.