Search

Summary: This open access proceedings presents new approaches to Machine Learning for Cyber Physical Systems, experiences and visions. It contains selected papers from the fifth international Conference ML4CPS - Machine Learning for Cyber Physical Systems, which was held in Berlin, March 12-13, 2020. Cyber Physical Systems are characterized by their ability to adapt and to learn: They analyze their environment and, based on observations, they learn patterns, correlations and predictive models. Typical applications are condition monitoring, predictive maintenance, image processing and diagnosis. Machine Learning is the key technology for these developments.

In order to enable the autonomic network resource management, we resort to game theory to facilitate the spectrum sharing between heterogeneous wireless networks. We set the spectrum sharing objective to improve spectrum efficiency and maximise network revenue. Based on this goal, we propose a revenue sharing bargain game to model the spectrum sharing behaviour of intra-operator radio access networks (RANs) and a price bargain game to model the spectrum trading behaviour between inter-operator RANs. Furthermore, we adopt multi-agent architecture to implement the spectrum sharing schemes autonomously. Both bargaining mechanisms are analysed using bargaining game theory, and consequently the implementations are refined and simplified based on the analysis. Simulation results show that the proposed mechanisms outperform the conventional fixed spectrum management method in network revenue, spectrum efficiency and call blocking rate. [ABSTRACT FROM AUTHOR]

"The next thing in technology is not just big but truly huge: the conquest of complexity", Andreas Kluth, The Economist [7]. The future of the Information Society draws a world where devices with processing capabilities are ubiquously connected to information sources. Due to the diversity of user devices, technologies, communication accesses, etc. operators do really need a technology able to manage such a complex environment giving a service end to end vision. In this sense, the set of technologies that arise from Autonomic Communication appears to be the cornerstone in order to manage the complexity of these new networks and services infrastructure. Competitive market is hard and telecommunication operators must face a double challenge, on one hand they must offer powerful services that must be managed, on the other hand they must address the reduction of service margins. Only really efficient operators will survive in such an environment. The I2OSS (Intelligence to Operational Support Systems) model is thought to provide operators with management infrastructures showing the advantages of Autonomic Communications. The reader will find in this article a technological framework for innovation in the field of Autonomic Communications. The I2OSS model establishes a methodological approach for innovation on OSS (Operation Support Systems) systems that support the management processes of a telco operator. No concrete or specific solutions will be found, but indications on how a telco operator must drive innovation to take advantage of "Autonomic Technologies". [ABSTRACT FROM AUTHOR]

MPLS DiffServ-TE combines the advantages of DiffServ and MPLSTE by allowing a differentiation of services and a traffic engineering based on a fast packet switching technology. However, such MPLS DiffServ-TE network needs an efficient method for Quality of Service (QoS) guaranteeing. In addition, the management of such a network is not a simple function and could not be done manually. In fact, it would be much more economic and effective to let equipments handle a part of the tasks attributed currently to the human administrator. In this paper, a novel architecture based on a Multi-Agent System is proposed to automatically manage MPLS DiffServ-TE domains. To meet the QoS requirements, an LSP creation strategy, based on a traffic-driven approach and depending on the traffic load, is proposed. It determines when to set up a new LSP and when to forward a new traffic in an already existing one. This strategy reduces the number of LSPs and the number of signalling operations in the network. Simulation results are provided to illustrate the efficiency of our proposition. [ABSTRACT FROM AUTHOR]

Autonomic communication aims to build more reflective systems with properties like self-healing, self-organisation, self-optimisation and so forth — the so-called "self-*" properties. To attain this within the existing strictly-layered approaches to network software may be possible to certain extent, but will not leverage all the possible optimisations, and we suggest that cross-layer architectures are better-suited to achieving the self-* properties. This paper explores the possibilities of cross-layering approaches in autonomic networks, reviews and compares the different cross-layer approaches to network architecture, observing that most current approaches depend purely on local information and provide only poor and inaccurate information gathering at the network level. [ABSTRACT FROM AUTHOR]

During the last decade, networks have been growing dramatically on several dimensions: the number of users has been multiplied by more than 15, the amount of traffic by more than 100 and the number of network technologies and usages has also been diversified. One of the main consequences of this growth is the increase in dynamicity of the network behaviour. Therefore, it is more and more difficult for human operators to manage the network in an efficient way. That is why current research in the field of network management tends to automate network control and management. A large initiative has been proposed by IBM to give more autonomy to computing systems: Autonomic Computing Initiative. The paradigm of Autonomic Networking was introduced as an extension of Autonomic Computing, focusing on the autonomy of networks. At the centre of it lies knowledge management. In fact, to be self-managed, a network element needs to know a lot about its environment. But sharing a large amount of knowledge over a large network is very expensive in terms of resources. Therefore, new knowledge management mechanisms are required. A good example of networks that need to be autonomic is military networks. In fact, on one hand, military communication environment is very dynamic, and very uncertain; and on the other hand, quick decisions have to be triggered to guarantee -whenever it is possible- the connectivity for strategical communications. This paper proposes an interpretation of the Autonomic Networking paradigm, a description of a situatedness-based knowledge plane, and an instantiation of these concepts to a concrete application: resource regulation in IP military networks. [ABSTRACT FROM AUTHOR]

This chapter presents the Living Systems Autonomic Service Access Management Suite, LS/ASAM, a comprehensive middleware solution enabling adaptive connectivity management of nomadic end hosts across heterogeneous access networks with autonomic optimisation of network performance and availability. [ABSTRACT FROM AUTHOR]

This paper presents an approach to manipulating available hybrid connectivity to autonomically maximise the potential for sustained connectivity in the event of path disruptions. The approach is documented in terms of the features, architecture and deployment modes of an autonomic communications module, termed RASCAL. This module employs software-agent logic supported by a state-of-the-art policy engine to dynamically determine best options for packet transmission over available infrastructure and ad-hoc connections. [ABSTRACT FROM AUTHOR]

Current network management systems have been impeded by a scarcity of open standards for interoperable management solutions. Information models have made progress in promoting interoperability of traditional "centralised" networks but have still to significantly address the proliferation of next generation networks such as autonomic networks. Such networks impose challenges which include distributed self-control and self-management. The goal of the Madeira project was to utilise novel technologies and methodologies, based on an underlying P2P paradigm, for a logically meshed, distributed Network Management System (NMS) that facilitates dynamic behaviour of transient network elements. In this paper, we describe a solution for a meta-model that attempts to capture the key concepts behind the task of network management of a mesh network. A case-study focusing on the fault management of such a network will be presented with the purpose of verifying the applicability of such a meta-model. [ABSTRACT FROM AUTHOR]

The 2006 MACE workshop [1, 28] has presented the drivers and challenges of Autonomic Networking [2] and fostered an understanding of emerging principles for this new type of networks [3]. In this paper, we present concepts and principles that define a technological neutral, architectural perspective of Autonomic Networks. The work presented in largely based on work within the Architecture team of the TeleManagement Forum (Technological Neutral Architecture, [4]) and joined research work of industrial and academic research teams (for example Ericsson [5], HP & QinetiQ [6] and Motorola [7]). The goal of this paper is to provide the reader with manageability guidelines and architectural patterns leading to the development of manageable autonomic software and communication systems. We present a component-based, distributed system architecture and an associated set of critical system services that Autonomic Networks require. Since we tackle the problem from a technologically neutral angle, this paper will not prescribe a single new technology, but rather provide a means that allows for federating different technologies, each of which offers particular advantages at business and system levels. In particular, it enables business concepts and principles to drive system design and architectures. This may be further implemented using currently available distributed systems information technologies. [ABSTRACT FROM AUTHOR]

You may be getting anxious to get started on your podcast, but before you take a stab at creating one of your own, it's a good idea to first spend some time listening and watching. One of the best aspects of learning to podcast is sampling a variety of the available programs. Not only is this fun, but it's a great opportunity to get "plugged into" the podcast community—to really find out what is available and what void may be left to fill. Your experience exploring podcasts will help tremendously when it comes time to create your own show. By checking out an assortment of podcasts, you'll get a sense of what you like and dislike. You'll also start to form some ideas about how you want to do your show. As with most anything else, you can learn a lot from the people who are currently producing podcasts. Above all else, you'll have fun. [ABSTRACT FROM AUTHOR]

Now, here's where the fun stuff begins. It's time to slide on those headphones, clear your throat, and begin. But before you hit the Record button, let's first go over a few things about recording audio, using the microphone, and making sure your recording is clear with good "levels." The podcast you record today may still be getting hundreds (thousands?) of downloads in six months, so let's ensure you'll still be proud of the sound quality that far down the road. Still being proud of what you said is completely up to you. [ABSTRACT FROM AUTHOR]

After all the hard work you've put into creating your show, it's difficult and possibly upsetting to imagine your podcast just sitting there on the server, being downloaded by only the two or three family members you can get to actually understand what you're spending your time doing. Let's be very clear about something right here and now: just because you think you're "podcasting" and there are millions of podcast listeners out there doesn't mean your podcast is going to instantly have an audience. People can't listen to a podcast they don't know about or can't find. You need to get your podcast out to the podcast community and out to the world. [ABSTRACT FROM AUTHOR]

It's time to do some soul searching and some serious shopping. It's time to put together your podcast studio. Now, as we said before, the price of entry is relatively low. You can record and prep your podcast with nothing more than a microphone and a handful of free software applications. On the other hand, you can very easily spend thousands of dollars on a "power podcaster" rig to ensure you have broadcast-quality sound. [ABSTRACT FROM AUTHOR]

In the past couple years, YouTube has exploded into a "consumer-generated media" (and copyright infringement) phenomenon. Untold hours of video are being produced and uploaded to YouTube (and its illegitimate offspring) all day, every day. Whether it's kids in basements lip syncing to Maroon 5 or corporate concerns trying desperately to hit upon a viral marketing phenomenon, it seems everyone with access to a camcorder is shooting, cutting, and uploading. As far as we're concerned, anyone producing video with any kind of regularity should be enabling it for subscription. Yes, podcast it, kids! [ABSTRACT FROM AUTHOR]

Before too long, your podcast will be up and running. You'll have your production process down to a science, your content will be focused and inspired, and your audience will be growing at a nice, steady pace. There's nothing like the days of your podcast's youth, watching what was recently a mere concept bloom into a full-fledged ongoing concern. You're busy now with your new offspring, checking download stats, answering e-mails, and worrying over your ranking on your favorite podcast directory. [ABSTRACT FROM AUTHOR]

It's time to get your podcast on the Internet and establish its web presence. The process is really pretty straightforward. First, you'll find a place on the Internet to host your files. Next, you'll upload your files and set up a website or blog to tell everyone what your show is about. Finally, you'll create the feed to let subscribers know that there is a new podcast available for download. [ABSTRACT FROM AUTHOR]

The last few steps before you upload your podcast for the world to download involve converting your file to a podcast-appropriate format and tagging it so people know where the file came from, who was involved, and how to find more. Luckily, the hardest parts of preparing your show are over. Now you just have a little bit of processing to do, and you're ready to send your podcast off into the wild cyber yonder. [ABSTRACT FROM AUTHOR]

So you want to do a podcast, huh? Well, we're not going to stop you. In fact, we'll help! But, before you begin, we advise taking a hard look at why you want to podcast. [ABSTRACT FROM AUTHOR]

So, it's finally time to do it yourself. You've checked out a few shows, gotten yourself all worked up, and now it's time to fire up that equipment and start a podcast. Right? Well, not quite. [ABSTRACT FROM AUTHOR]

There are probably as many ways to handle the editing of audio as there are podcasters. Some podcasters approach postproduction by simply hitting the Stop button on their recording device, while others labor for hours, fine-tuning their audio until it is perfect. (Of course, "perfect" is a completely subjective assessment.) Whether you're working on an audio or a video podcast, the approach you take toward editing and processing the sound will depend, like so many other things, on you and your show. You may just want to record the show "live in studio" where whatever happens, happens. No editing, no assembly after the fact. Or you may want to assemble your podcast from various individual pieces, weaving the elements together in the editing process. [ABSTRACT FROM AUTHOR]

Podcasting is one of the most exciting and wonderfully disruptive technologies to emerge in recent history. [ABSTRACT FROM AUTHOR]