Toward Flexible and Scalable Sensor Network Systems with Evolving Intelligence

http://www.mdpi.com/journal/jsan/special_issues/TFSSNSEI

Dear Colleagues,

As sensor network organization is increasing in complexity, the ability to efficiently utilize the system requires a better understanding of the functional components and their interrelations. The connectivity of wireless sensor networks to the cloud environment has opened the door for a range of options towards system operation, scalability, and flexibility. This has encouraged research work to develop at both the physical sensor network level, as well as the cloud and related virtualization level. This could lead to a highly complex system with an environment that could have the ability to evolve with intelligence and offer a wide range of reliable and efficient services.

At the sensor network level, various researchers have looked into the formation of network organization in capturing the dynamics of the phenomenon. Here, for example, the grouping of network nodes—or what we call the formation of physical sensor clouds (PSC)—on functional, spatial, temporal, and/or spectral basis is ongoing research. This targets unfolding best operational practices for cyber physical systems. On the other hand, a significant amount of research work has been targeting the area of virtualization of the PSC and their related cloud services. The term “xx as a service” has been used extensively to reflect the range of services offered. Aspects like the process, the data, the infrastructure, the software, or even the QoS as a service have been explored. With a rich digital environment, the cloud is hosting aspects of intelligence, including ambient intelligence, opening the door for managing system behaviour and related training at large and, thus, allowing for better interaction with the process dynamics.

Further system complexity results from the inclusion of social and scientific data. They have started to play an important role in understanding the complex dynamic process behaviour, through cross-correlation among the behaviours in each of these domains. These aspects, among others, contribute to the increasing complexity of sensor network systems. The challenge here is the need to understand and manage the behaviour at both the component and system levels. This should then encourage the formation of an organic organisation that is manageable, dependable, and lends itself to better operational efficiency, flexibility, and scalability.

This Special Issue is trying to focus the efforts toward unfolding the key parameters and system organisations that could help in formulating an evolving system that is dependable, scalable, and flexible. The goal is to encourage new ideas that reflect potential in supporting a system organisation that could have the ability to adapt flexibly in capturing the dynamics of a physical phenomenon efficiently, and offer the related quality of services with the least human intervention. This could be at the sensor network level, IoT level, sensor cloud level, and cloud services level.

Topics:

Cyber Physical System (CPS) and Physical Sensor Cloud (PSC) formation.
IoT based PSC and related data management.
Virtual sensor clouds and cloud services.
Data organization, analysis and visualization.
Cloud based sensor network architectures.
Evolving sensor network organization and dynamic system training.
Sensor network and ambient intelligence. Sensor Network QoS as a service (QoSaaS).
Dynamic interaction of sensor network data, social data and/or scientific data.
Adaptive sensor networks, self-organization, and trust
Big data and event detection
Case studies and applications in various challenging areas of smart cities: e.g., eHealth, Ambient Assisted Living, Intelligent Transportation Systems, logistics of perishable goods, and others.
Other related areas.

Prof. Dr. Adnan Al-Anbuky
Prof. Dr.-Ing. Christian Müller-Schloer
Prof. Dr. Antonio Puliafito
Prof. Dr. Franco Davoli
Guest Editors
Public
All users can post
Moderated
7 followers

Write a new post to this group

 
Anonymous  |  Now

Latest Discussions in the Group

Intelligent sensor networks from the MAS perspective

 
Christian Müller-Schloer  |  25 August 2015 14:41  |  3 replies  |  View topic in separate thread
Dear colleagues,
here are a few thoughts to illustrate my perspective on Intelligent Sensor Networks (from the point-of-view of adaptive and self-organizing systems).
With best regards
Christian


Social Sensor Networks

Sensor networks are becoming more complex, in terms of the number of participating nodes, their connectivity, and the intelligence of the nodes themselves. This is a challenge because the highly dynamic organization of such a network cannot be managed manually any more. But it gives us the chance as well to exploit the higher intelligence of the nodes for their ability to accomplish the frequent re-organizations themselves. Self-organization is therefore a necessity for such systems, and it is feasible as well.
This complexity increase is acerbated by the fact that future sensor systems (and this is true for all distributed systems) are becoming open systems. We cannot control the composition of the total system any more nor the single systems themselves because they are autonomous and might have been built by unknown programmers. This means we can judge them only from their externally observable behaviors. So we have to resort to social mechanisms such as reliability, dependability, trust, reputation, or forgiveness.
Sensor networks consisting of thousands of intelligent and largely autonomous nodes can be viewed like a social system. This means that we can learn from other social systems such as our human one and try to copy some of the successful control mechanisms developed there. It has been shown that agent systems become more robust against external disturbances, malevolent agent attacks and abnormal systems states when we add trust and reputation to the agents’ abilities. But it has also become clear that a flat P2P organization is not sufficient in all cases. Certain global system states like the so-called trust break-down or overload situations can be detected from the single agent point-of-view only with difficulties. System-level observers are much more effective in this context. Along the same line, distributed control (e.g. to accomplish a re-assignment of tasks to the nodes) might be possible but it is less efficient than a limited control interaction by a system-level controller.
It is interesting that real (i.e. human) social systems have developed such partially hierarchical control mechanisms in the form of global rules, goals, laws, norms, or conventions. Nevertheless, the single human agents remain largely autonomous. While in Multi Agent Systems research normative control has been explored we see an interesting potential to practically apply normative control to networks of intelligent sensors or sensor-actor systems like robots. The challenge will be to find architectures and interaction schemes that allow limited top-down control while leaving the agents (semi-)autonomous.
A number of challenging research questions results from this analysis. I will list just a few of them.

  • How can we balance local control of an autonomous agent with external control applied from the outside?
  • How can we construct these systems such that the additional technical effort for the social system is justified by an increased performance or increased robustness of our target system?
  • How can norms be designed such that the enforced micro-behavior leads to the desired macro-behavior of the system?
  • How can we prevent the harmful exploitation of social mechanisms by intruders? Here collusion attacks present a special challenge.
  • How can we guarantee that these systems converge to a stable state in bounded time?
 
Adnan Al-Anbuky  |  27 August 2015 04:59
Dear Christian
The argument you are putting on the table is quite valid. Looking at the sensor network and related services as an autonomous unit, that have to integrate with a bigger community of sensor networks and services, and from its external behaviour point of view, is something that will eventually happen. We may have two basic scenarios here:
1. workable model formulated by a group of sensor networks and related services
2. A new network and/ or service trying to merge into the existing group in 1 above
So how would a community of sensor networks formulate a group? And how that organization are going to be governed?
On the other hand what virtual environment should be in place for the new member to test their ability to join the group smoothly?
While these are all good at the federated system level, we still need to dig deep at the autonomous unit level (or if the term allow the molecular Level). In this context and if we assume that we are constructing a new sensor network, we need to identify what are the key features that offers flexibility and training to this network and how this is going to be interfaced with the external system. The current literature offers rich research information on various aspects of designing a sensor network, can we extract some form of intelligent interface out of this that could be used for manipulating the sensor network system or even bound a given system and allow for smooth merging or prevent conflict with others? At the sensor network level this in effect will be the key for the cyber physical system where one try to optimize the engagement of the sensor network with the related phenomena dynamically. On the remote server or cloud level this may relate to the number of agents such as network virtualization, training and implementation. The questions that remain to be answered yet are:
1. What is a typical architecture of a sensor network system that can organically evolve and dynamically change with the associated phenomena?
2. How would the operational flexibility of the system be expressed such that it allows for an effective interoperability interface? In another word how effective we could get in accessing the system potential without having to know the internal design?
3. What would be the mechanism for merging multiple systems that have overlapping behaviours and my result with compound services?
4. What would be the mechanism for formulating an open community of sensor network systems? And how would member joins and leave this open community?
5. What live examples are available that can be used as use cases for any of 1 to 4 above? Possibly the context of smart cities may bring about numerous examples.
6. Should there be a standard that defines the growth of systems, knowledge and technology here?
The top down approach that you are reflecting here could be quite beneficial when looking into the system complexity as it grows in number of overlapped or collaborating networks and offers various possible services. So we may be having wide range of networks that are associated with various phenomenon spread over the planet but can deal with newly defined phenomenon (or new services) through the available infrastructure.
 
 |  27 August 2015 08:17
Dear Christian and Adnan,
I fully agree with your observations. I am strongly convinced that future sensor networks will be managed according to Cloud computing principles, that we may call Sensor clouds. This means a very flexible way to organized sensing and actuators resources that may join and leave the systems dynamically. In this sense we have a combination of traditional cloud systems with sensors clouds, that will federate to set up ad hoc infrastructures to provided new advanced services.
Smart cities/communities represent an interesting scenario where new principles can be tested, also involving crowdsensing, and adding citizens into the loop.
We are pursuing such goal in the #SmartME crowdfunded initiative. I summarize the objectives reporting a glimpse of such project:

The #SmartME project was born from a wish of a team of researchers in the Mobile and Distributed Systems Lab (MDSLab) at the University of Messina who, in collaboration with the Industrial Liaison Office and the Center for Information Services of the University (CIAM), are eager to encourage, in an innovative fashion, a “conversation” with the municipality of Messina, based upon the paradigm of the Internet of Things (IoT), in order to spur the creation of a novel, virtual, ecosystem. The project is also going to involve the university spin-off DHLabs, active in the production of innovative solutions for advanced sensor-based systems, and is sponsored by the Municipality of Messina.

At the foundation of #SmartME thus lies the belief that research can and should yield benefits for the community. We are therefore talking about a novel way to play the role of a University with regards to technology transfer, a more “open” one, as it is aimed even at laymen; a “smarter” one, because it enables anyone - citizens, researchers, enterprises - to experiment and test new services for the city.

To morph Messina into a “smart” city, it is essential to set up an Open Data platform that demands the employment of boards based on low-cost microcontrollers, such as, e.g., Arduino Yun, installed onto buses, lamp posts and buildings of local institutions. Moreover such a network will be enlarged with sensors and actuators scattered all over the urban area.

Thanks to such infrastructure, it will be possible to collect data and information for building services for citizens, who may take part in this network through the involvement of smartphones and other mobile devices by which it will be possible to interact with objects and may even themselves turn into data producers. For instance, it will be possible to monitor global indices about environmental quality, which will constitute the initial testing stages of the project, yet an Open Data platform will be available to be leveraged for other services too (e.g., mobility management, district area monitoring and safety, tracking of expenses, reporting of acts that may jeopardize maintenance of public facilities and citizens’ safety).

It follows that several services will concur to make the city of Messina “smart” and help citizens to improve their daily life accordingly, as Smarter People. In that spirit, every year, the MDSLab team, in collaboration with ILO and CIAM, will organize a contest that will assign a prize to the most innovative idea by conferring a Digital Innovation Award.

For the scientific community, #SmartMe will represent an interesting case study, that will supplement the FI-ware project ones, thanks to a development, integration and testing platform, that we are set to establish, integrated with networks of sensors, actuators or other smart devices already deployed to the district area, and to this day employed exclusively in specialized domains of application and for a limited number of purposes.

 
Franco Davoli  |  29 August 2015 10:39
Dear All,

I have read your comments and thoughts with much interest. You have been indeed raising a number of very interesting and challenging questions. In order to better understand myself some of the issues, let me just try to look at the problem from a number of different points of view, which may correspond to different levels of abstraction in a complex hierarchical control system. At each of these, one could identify control and communication strategies, and communication protocols. As the level of abstraction increases, we are looking more into functionalities and services.
In this context, I envisage the following:
i) a physical control layer, where individual sensors interact directly with neighbours and/or with sink nodes. I am viewing sinks as intermediate nodes that, besides performing data collection, aggregation and, possibly, fusion, help in constructing the abstract views of each sensor network portion that will be the objects handled by the higher abstraction layers (in the cloud, or even in the access network). Protocols and addressing schemes at the physical control layer can be the classical ones mostly used in sensor networks (e.g., IEEE 802.15.4/ZigBee, IEEE 802.15.1/Bluetooth, ISA100/WirelessHart, etc.). They can encompass OSI layers 1 and 2, and even include routing strategies in multi-hop networks. From the control point of view, there are several possible tasks to be accomplished, depending on the goals and capabilities of the sensing objects and on the possible presence of actuators that perform control actions on a physical system of interest. In this latter case, the sensor and actuator network at this level performs “regulatory” control, e.g., with the goal of stabilising the system around set points that have been decided upon by an optimisation process at a higher abstraction layer, or following a trajectory in state space, again determined by a higher control task. This task can be more or less complex, according to the degree of distribution of the system parameters, the size of the network, the degree of cooperation among nodes. Other control task here may be performing tradeoffs between measurement and communication (where and how aggregate data, e.g., by means of estimation procedures), determining network boundaries (and, in this case, handling incoming and outgoing nodes), performing localisation, devising distributed encoding strategies that possibly combine source and channel coding.
ii) An intermediate layer, which should build a first abstraction of the underlying networks, offering a standard interface, possibly in the form of an API, to the higher abstraction levels. In the past, in the context of Remote Instrumentation, we have used the term “Instrument Abstraction Layer” (IAL) to identify this. There are several ways to accomplish this task, with more or less detailed views, but the general goal should be to provide: i) an abstract vision of a sensor network (or a portion thereof) as a manageable object, conveying the control capabilities, manageable parameters, key performance indicators and available commands; ii) a means of identifying, and interacting with, the physical objects inside the network with different levels of granularity; this aspect may include, for instance, also the translation between different addressing schemes (e.g., each sensor may have an IPv6 address, which is used to identify it in the outside world, but not inside the physical network itself, in the sense that no IP routing is present inside the physical network). In general, the IAL should allow the higher layers to identify the portion of sensor network it refers to as a single “instrument”, which can be then managed according to Network Control Policies (NCPs) that address the system as a whole. A similar point of view has been adopted recently in the ECONET project, where we have been working on energy efficiency aspects in networking, in order to convey the effect of Local Control Policies (LCPs) applied inside devices to trade off power consumption and performance, as well as the capabilities of the underlying hardware/firmware, and to transfer commands, between network devices and the control/management planes (possibly residing in a cloud); this development has led, in that context, to the recent standardisation of the Green Abstraction Layer (GAL), and abstract interface adopted as ETSI standard ES 203 237. The general philosophy here is quite similar.
iii) One or more higher control layers acting on abstract objects that represent aggregate views of the underlying physical objects or of their “more detailed” abstract representations. These may be handled in a cloud, but in many cases it might be advisable to have such abstract environments residing closer to the physical objects, e.g., in the access network (this vision is in line with the general trend of many operators, which are moving toward the adoption of SDN and NFV, to deploy general-purpose hardware in the network and to assign it more processing capabilities and tasks, to be able to directly offer services to the users beyond the transfer of information). The control tasks here may be hierarchical, aiming at coordination of several underlying network “islands”. Here again, the possibilities are quite a few, depending on the abstraction level and the representation of system dynamics. They can range from dynamic control strategies over finite horizons, receding horizons or infinite horizons, to simpler parametric optimisation schemes, with the goal of determining set points or trajectories for the underlying objects, which will be conveyed through the IAL (from where, in the other direction, measurements are acquired). The point of view of abstracting a whole sensor network as an object in a “personal cloud” is being pursued currently in the INPUT project, where the goal is to build user personal networks that allow objects (virtual and physical) in a user home network to be virtualised and operated upon within a personal virtual environment, handled in cooperation between the operator’s access network and a cloud.

As a final observation, it may be noted that, within a given abstract hierarchical layer, control and communication schemes may be centralised or distributed, both from the point of view of control and communication. On one hand, the adoption of SDN allows applying centralised control strategies, owing to the possibility of receiving information and acting upon a number of switches at a time; on the other hand, it may be convenient to adopt purely distributed paradigms in some cases (e.g., in the past we have used a p2p protocol environment to perform bandwidth allocation to sensors, by reaching agreement in a completely distributed fashion).

Citations

There are no citations in this group

Files

There are no files in this group

Group information

Created by: Ling Yang
Created on: 21 August 2015

Full group description

http://www.mdpi.com/journal/jsan/special_issues/TFSSNSEI

Dear Colleagues,

As sensor network organization is increasing in complexity, the ability to efficiently utilize the system requires a better understanding of the functional components and their interrelations. The connectivity of wireless sensor networks to the cloud environment has opened the door for a range of options towards system operation, scalability, and flexibility. This has encouraged research work to develop at both the physical sensor network level, as well as the cloud and related virtualization level. This could lead to a highly complex system with an environment that could have the ability to evolve with intelligence and offer a wide range of reliable and efficient services.

At the sensor network level, various researchers have looked into the formation of network organization in capturing the dynamics of the phenomenon. Here, for example, the grouping of network nodes—or what we call the formation of physical sensor clouds (PSC)—on functional, spatial, temporal, and/or spectral basis is ongoing research. This targets unfolding best operational practices for cyber physical systems. On the other hand, a significant amount of research work has been targeting the area of virtualization of the PSC and their related cloud services. The term “xx as a service” has been used extensively to reflect the range of services offered. Aspects like the process, the data, the infrastructure, the software, or even the QoS as a service have been explored. With a rich digital environment, the cloud is hosting aspects of intelligence, including ambient intelligence, opening the door for managing system behaviour and related training at large and, thus, allowing for better interaction with the process dynamics.

Further system complexity results from the inclusion of social and scientific data. They have started to play an important role in understanding the complex dynamic process behaviour, through cross-correlation among the behaviours in each of these domains. These aspects, among others, contribute to the increasing complexity of sensor network systems. The challenge here is the need to understand and manage the behaviour at both the component and system levels. This should then encourage the formation of an organic organisation that is manageable, dependable, and lends itself to better operational efficiency, flexibility, and scalability.

This Special Issue is trying to focus the efforts toward unfolding the key parameters and system organisations that could help in formulating an evolving system that is dependable, scalable, and flexible. The goal is to encourage new ideas that reflect potential in supporting a system organisation that could have the ability to adapt flexibly in capturing the dynamics of a physical phenomenon efficiently, and offer the related quality of services with the least human intervention. This could be at the sensor network level, IoT level, sensor cloud level, and cloud services level.

Topics:

Cyber Physical System (CPS) and Physical Sensor Cloud (PSC) formation.
IoT based PSC and related data management.
Virtual sensor clouds and cloud services.
Data organization, analysis and visualization.
Cloud based sensor network architectures.
Evolving sensor network organization and dynamic system training.
Sensor network and ambient intelligence. Sensor Network QoS as a service (QoSaaS).
Dynamic interaction of sensor network data, social data and/or scientific data.
Adaptive sensor networks, self-organization, and trust
Big data and event detection
Case studies and applications in various challenging areas of smart cities: e.g., eHealth, Ambient Assisted Living, Intelligent Transportation Systems, logistics of perishable goods, and others.
Other related areas.

Prof. Dr. Adnan Al-Anbuky
Prof. Dr.-Ing. Christian Müller-Schloer
Prof. Dr. Antonio Puliafito
Prof. Dr. Franco Davoli
Guest Editors

Group members