Ubiquitous Computing: Smart
Devices, Environments and Interactions
Wiley, ISBN: 978-0-470-03560-3, 2009
Ubiquitous computing, pervasive computing, distributed system, autonomous system, context-aware, artificial intelligence, smart devices, mobile devices, smart environments, smart interaction, tabs, pads, and boards, smart dust, smart skin, smart clay, HCI.
A novel holistic framework has been proposed for Ubiquitous Computing (UbiCom) called the Smart DEI (Devices, Environments and Interaction) model which is based upon three interlinked system viewpoints: a set of internal UbiCom system properties; a set of distinct architectures based upon smart mobile devices, smart environments and smart interaction; system interaction in three distinct environments (ICT or virtual computing, physical world and human world).
No single formal definition for UbiCom systems is given because a diverse range of UbiCom systems is needed. Instead, a set of properties is defined which can be combined in different ways to support this diverse range of systems. Five core properties for UbiCom systems, distributed ICT, context-awareness, intrinsic human computer interaction, autonomy and artificial intelligence are proposed. A classification (of over seventy terms) which supports different combinations of UbiCom system properties and sub-properties and which defines synonyms for many different types of ubiquitous computing is given. These core properties overlap and are interlinked, e.g., iHCI is interlinked to user context-awareness; autonomous system, anticipatory HCI systems and intelligent systems are interlinked.
Mark Weiser proposed three main forms for ubiquitous computing: Tabs (small, inch or centimetre sized, planar, wearable, handheld devices, e.g., smart phones, smart cards), Pads (medium, foot or decimetre sized, planar, two handed operation, multi-application, hub devices, e.g., laptops, touchscreen computers etc) and Boards (large, yard or metre sized, planar devices often fixed to physical environments, often used for interactive display, e.g., smart whiteboards etc). This set of forms is extended to include three additional forms: Dust (micro and nano sized devices); Skins (non-planar surfaces which can be used to cover natural and artificial physical world objects) and Clay (programmable synthetic matter).
Three basic designs for UbiCom based upon smart devices, smart environments and smart interaction are proposed. Smart devices are characterised as being mobile, personalised, planar, macro sized MTOS (Multi Task Operating System) devices which can be used as multi-application platforms, Web portals and combined multimedia players, recorders and communicators. These tend to be designed to access remote rather than local services. Smart environments are environments in which static macro devices are embedded into it, or untethered micro and nano-sized devices are scattered into social and public spaces. Local interaction dominates the use of smart environments. Smart interaction is concerned with combining multiple individual smart devices and environments in order to interact in richer, flexible ways, such as supporting orchestrated, choreographed, competitive and cooperative interaction in dynamic virtual systems. Hybrid designs can also be used where systems combine smart devices, smart environments and smart interaction.
UbiCom systems use covers a range of interaction: between two or more UbiCom devices (C2C or CCI); between devices and people (HCI); between devices and the physical world (CPI). Some of CPI involves sensing the physical environment, performing tasks which are situated in it, affect it and may control it. Physical environments which are smart environments can act in part as collections of smart devices so CPI involving smart environments approximates towards CCI. Interactions between humans (HHI) and between humans and the physical environment (HPI) are often mediated using computer devices. If humans interact with physical environments that are designed as smart environments then human to smart environment interaction approximates towards HCI combined with CPI.
Key Words: Smart devices, smart environments, applications, everyware, smart vehicles, smart buildings, ebooks, ambient intelligence, cyborg, wearable computers, tangible interfaces, smart badge, calm computing, smart space, smart dust, Cooltown, pervasive games.
The vision of ubiquitous computing proposed by Mark Weiser about twenty years ago is a grand vision for computing that is gradually being fulfilled as devices, environments and interactions between humans, between humans and computers, between computers and the physical environment, are becoming digitally endowed. For example, location-awareness can enhance many physically situated activities for mobile users and services that involve mobile business assets.
Many researchers and early projects made significant contributions to the field including: Weiser’s smart tab, pad and board and his calm computing ideas; active badge, bat and projects at Olivetti Labs, Cambridge University with Roy Want; early robot development such as the Unimate and MH-1; Steve Mann’s WearComp and WearCam projects in wearable computing; Kevin Warwick’s work on computer body implants in his Cyborg projects; Sony DataTiles project led by Jun Rekimoto, HP’s Cooltown project; the Smart Dust project led by Kris Pister University of California, Berkeley; Things That Think program at MIT involving Prof Ishii’s Tangible Bits project and work by Neil Gershenfeld; the Homelab and Ambient Intelligence project at Philips, NIST’s Smart Space project and various context-aware project led by Geoffrey Abowd such as the Aware Home and Classroom 2000. These are only a sample of the many noteworthy UbiComp projects undertaken
The current status is that UbiCom systems is that there is a wide variety of systems and applications in route routine use, including smart environments containing smart doors, lighting, taps and air ventilation which can detect the presence of humans and adapt to them. Smart transport systems can predict when they will arrive and notify passengers about this. They can predict collisions, failures and adapt their behaviour. These systems, however, tend to exist more as islands of interoperability rather than supporting seamless interoperability. No single grand design will work, many more years and many more cycles of innovation are required to lead to more seamless interoperable UbiCom environments. The disruption to the social interaction and to our interaction with the physical world needs to be carefully evaluated if UbiCom technology is to be harnessed to enhance rather than to degrade human interactions, activities and experiences.
Distributed system, abstraction, partitions, client-server, proxy, middleware, service oriented computing (SOC or SOA), Grid computing, peer to peer (P2P), service discovery, message-oriented middleware, remote procedure call (RPC), cache, delayed write, on-demand, event driven architecture, enterprise service bus, volatile service access, service composition, virtual machine, multi-tasking, operating system (OS, MTOS), BIOS.
A more lateral and holistic view of information and task access via a variety of tab and pad sized ICT devices is discussed. Smart devices embody access to distributed ICT system services and support: multiple applications, ease of use, minimum user management, dynamic heterogeneous services, mobility, volatile service access, local resource constraints, and user and physical context awareness. Key design principles are to support abstractions to simplify access and to use virtualisations to support interoperability across heterogeneous service components.
A range of service architectural models for device access to services have been proposed that define different notions of service. Service architectures include RPC and client-server interaction over MTOS platforms. The application service provider model focuses on distributed services that are hosted by providers, accessed by browser devices and leased to users and managed by providers on behalf of users. In the middleware service model, generic services such as communication and data management are factored out of specific application servers and hosted in computer nodes so that they can be shared across multiple application services. SOC defines explicit models of service interfaces and service-oriented interaction. P2P service models support more flexible service availability over more dynamic service infrastructures such as service access over mobile ad hoc networks.
There are a range of designs for partitioning and distributing services that depend upon the application’s use of resources and processing, the type of communication service and upon the type of access device used. Resource poor access devices, such as light-weight mobile devices are often designed so that service execution largely occur over the network, relying on external servers - a utility model (thin client-server design). However, the need to cope with unreliable and low-performance networks, as well as the need to adapt to limited power, supports a case for some degree of self-reliance and local processing support, i.e., elements of an appliance model (thick-client server design). Additional service designs include the use of client proxies (to mask the complexity of communication for client access) and middleware (to factor out common service actions).
The provision of application services for smart devices entails the management of distributed services throughout the whole of their life-cycle. There are four main phases for service management: service creation, service execution, service maintenance and service dissolution. Some of these phases such as creation are complex and consist of a number of sub-phases such as service announcement, discovery, selection, configuration and composition. XML based Web services and RFD-S and OWL based Semantic Web services support the service life-cycle to varying degrees. The main designs to support volatile service invocation include: the use of asynchronous and unreliable communication protocols; messages caches, read ahead and delayed writes; on-demand service access; event-based notification and shared data repositories. Open service access models can be based upon universal data exchange models and virtual machines.
High resource smart devices can support a full general purpose operating system but there are variations in OS design depending on monolithic versus microkernel design, how the BIOS bootstraps the OS and how resource constrained devices are supported.
Mobile devices, mobile code, mobile user, SMS, WAP, J2ME, .NET CF, microkernel, resource-constrained, power management, smart card, device network, device discovery, OSGi, X10.
High resource smart devices can support a full general purpose operating system but there are variations in OS design depending upon: monolithic versus microkernel design, how the BIOS bootstraps the OS and how resource constrained devices are supported.
Users are naturally mobile, e.g., users can move in between Internet nodes, to log on and to access Web based content and email, anywhere, anytime. Users can carry personalised mobile networked devices with them to access services, filtered according to their personal preferences and that are aware of their location context and can adapt to it. Each of the main components of an UbiCom system can be mobile such as processing at the operating system and application level, code, service access and presentation, resources and services. In order to simplify access by applications and users, mobility in terms of how to locate and address mobile users and how to route data to mobile receivers, should be made transparent. Three kinds of transparency for mobile services are considered: user virtual environments, mobile virtual terminals and virtual resource management.
Code mobility is an important enabler for system extensibility, for supporting operation in open dynamic environments and in particular for use in resource constrained access devices. It allows service access devices to download new operational capabilities at run-time without requiring a capacity to store all possible needed service support in advance and it enables providers to maintain, i.e., to upgrade and fix code in, consumer devices with a network connection without the provider having to ship physical media to customers. However, there are increased security risks concerned with potentially malicious or malfunctioning code being downloaded onto devices and frequent disruptions to consumer ICT activities when multiple applications on multiple devices upgrade themselves.
Device mobility can be viewed from several dimensions: in terms of whether or not the device is mobile or some kind of host to which it is attached, is mobile; in terms of what kind of hosts, mobile devices can be bound to; in terms of how devices are attached to a host and in terms of when the mobility occurs. There are three main designs for mobile information services, SMS, WAP and i-mode. Mobile Web Service design often use a three tier thin-client, client-proxy, server architecture because a client-proxy is used in order to off-load dealing with the heterogeneity of adapting content to heterogeneous terminals, micro browsers and content languages rather than using a passive approach to content adaptation. Three main technology models for mobile data communication devices are Mobile Web services, e.g., WAP, J2ME and .NET CF.
Even smaller mobile devices are smart cards. Many things found in a person’s wallet have the potential to be represented as a smart card, including a driver’s license, insurance information, Chip and Pin type bank cards, travel cards and tickets etc. Smart cards potentially represent a virulent form of Privacy-Invasive Technology.
One last device model discussed here is the home device network. The design requirements are for home network and service infrastructures that are easy to install, configure, maintain and are low cost. If device networks are to support open, dynamic, heterogeneous device access and interoperability, then automatic device discovery is necessary to simplify use. Device discovery standards and technologies include Sunís Jini, UPnP forum headed by Microsoft, IETF’s SLP, DNS Service Discovery and Bluetooth’s SDP. One of the most prominent current technologies here is the OSGi model.
The design of an OS for mobile use is challenging in order to deal with the higher prevalence of heterogeneous mobile access terminals; the need to dynamically route messages as the user moves, the need to deal with resource constrained devices and the need to conserve energy in a mobile terminal with finite energy supply.
HCI, user interface, WIMPS, multi-modal, gesture, virtual reality, augmented reality, touchscreen, tangible interface, organic interface, posthuman,, wearable computers, heads-up display, virtual retinal display, brain computer interface, human centred design, user context, usability, conceptual model, mental model, user modelling, stereotype, task model, situated action, implicit HCI, personalisation, affective computing, interaction design patterns.
There is a range of human interaction depending to what degree the user is in control of the interaction and is active compared to that of the system they are using. In totally passive systems, users are the active ones and explicitly interact with the system. Semi-active systems take some explicit interaction from users but may also actively model users to anticipate their actions based upon context (implicit interaction). Automated systems are where they system is active during normal operating conditions and users are passive.
As the sheer number and variety of devices increase, the potential for explicit interaction increases. Increasingly, human activities also involve human interaction with multiple devices, multiple participants, multiple tasks and in multiple environments. For many applications, implicit interaction is useful to complement explicit interaction. Implicit interaction is a way for the system to take away some of the control of the interaction from the user in order to reduce the interaction needed by the user to operate the system. The use of full user control via explicit HCI (eHCI) versus reduced user control via the addition of implicit HCI (iHCI) and full system control, is a design decision which depends in part of the type of task and application.
Four main types of interaction with common ICT devices are analysed, the personal computer, hand-held mobile devices used for communication, games consoles and remote controlled AV devices. It is seen that much effort has been used in the design of user interfaces to make these more usable, e.g., to support natural interaction using touch screens, gestures and speech input. Richer and more natural interfaces are being developed including wearable computers, tangible and organic user interfaces and using devices implanted into humans.
A human centred design process for interactive systems specifies four principles of design: the active involvement of users and a clear understanding of user and task requirements; an appropriate allocation of function between users and technology based upon the relative competence of the technology and humans; iteration is inevitable because designers hardly ever get it right the first time; a multi-disciplinary approach to the design. Human centred design life-cycle involves user participation throughout four main sets of activities: defining user tasks and the (physical, ICT) environment context; defining user and organisational requirements; iterative design prototyping and validation against the requirements.
To enable humans to effectively interact with devices to perform tasks and to support human activities, systems need to be designed to support good models of user interfaces and processes of human computer interaction. Users can be modelled directly andindirectly. User task models can be modelled as task plans or as situated actions. iHCI design concerns three additional concerns: support for natural (human computer) interaction; user models including models of emotions which can be used to anticipate user behaviour and user context awareness including personalisation. Some design patterns and heuristics oriented towards iHCI are described.
Tags, RFID, virtual tags, sensors, sensor nets, controller, proportional integral differential (PID) controller, robot, embedded system, Micro-ElectroMechanical System (MEMS), nano technology, nanobots, operating system, Application Specific Operating System (ASOS) , Real-Time Operating System (RTOS), smart dust, smart paint, smart skins, smart matter,
Computer (device) to Physical Environment Interaction involves both interaction with dumb natural physical environments and smart (physical) environments, the latter is the focus in this chapter. Smart environments considered to be physical environments which are embedded with fixed, macro-sized devices or scattered with micro untethered devices. The environment can then acquire and apply knowledge about itself and its inhabitants in order to improve the experience of its inhabitants.
Important enablers for ubiquitous computing are the reduction of cost, size maintenance and interdependencies of new computing devices. The realisation of this vision introduces its own secondary set of problems in terms of establishing ownership of micro items, coping with the flood from many thousands of objects triggering signals, interpreting ambiguous and redundant signals and disposing of micro devices.
Several complementary enabling trends for pervasive computing are described, objects in the physical world can be physically tagged, e.g., using RFID tags, or virtual views of the physical world can be recorded and tagged. Both of these enable physical world features to form part of a discourse in a virtual computer environment. Tags may be collocated with a linked physical world object. The link between the tag and object maybe direct or indirect. There are many applications of tagging systems including tracking of business assets and enhancing personal interaction with the physical world and personal memories of the physical world.
To interact with the physical world, systems need to sense, measure and adapt to its state. To get a true picture of a region of physical environment may entail the use of multiple sensors which measure multiple physical phenomena, situated at multiple physical locations, which are networked into a sensor net. This sensed data about the physical world needs to be processed, related to an application context, stored and retrieved.
Electromechanical devices that act on the physical world can be macro sized, e.g., robots, micro sized, e.g., MEMS devices, or nano-sized, e.g., nanotechnology and nanobots. Whereas MEMS devices tend to be fabricated in a similar way to IC chips, nano-sized devices involve molecular engineering of more varied materials. Although, the microscopic world is governed by the same physical laws as the macroscopic world, the relative importance of physical and chemical laws at this scale and how these affects the mechanics and the electronics of systemschanges.
Smart physical environment devices tend to use a different design for managing processes, memory and ICT resources compared to a personal, MTOS computer. Such embedded devices tend are designed to be customisable, reconfigurable and low cost and to use an ASOS (application specific) OS.
The simplest type of ASOS is a controller which is activated only when defined thresholds are crossed. More complex types of controller include PID control, adaptive control and robust control. An important type of system which combines the use of controllers and sensors are robots. Robots consist of sensors, actuators, a locomotion and drive unit and a controller. Robots differ in focus from embedded systems in that the programmable control for the latter is fixed and task specific. Robots can be along classified into robot manipulators or robot arm, mobile robots and biologically inspired robots and nanobots. Practical issues in developing robot applications are also considered.
Context aware, adaptive system, composition, mobility-aware, location-aware, spatial-aware, trilateration, triangulation, GPS, geographical information system (GIS), geocoding, user-context, ICT-aware, content-adaptation, temporal-aware.
A basic context-adaptation process model involves acquiring the current state of the environment through sensors, mapping it to some current application context and adapting the application to the current context. The use of the current context is often driven through its relationship to some (user or application) goal context. Two main types of context adaptation can be distinguished depending on whether or not systems actively and automatically adapt applications to some well-defined physical environment context, such as location and temperature, or whether systems support more passive adaptation. Passive adpation by the system may be a better option because of the complexity of context-awareness and adaptation and when theere exists a need for humans to retain some decision making capability.
More complex context-ware systems involve adapting not just to current crisp contexts but also to past, less precise, uncertain and composite user, including social (groups of users), contexts. On occasion, no predefined user context may exist. Instead, changes in an environment context may spontaneously trigger a new user goal context and trigger adaptation, i.e.,situated actions and goals. The more complex the adaptation is, the more challenging it is to do and the increased likelihood that undesirable and incorrect adaptation can result. The ability to accurately determine a user or physical context from partial information, often indirectly, may not be possible with certainty, e.g., weather prediction or users who perform subjective and variable actions. A context-aware system should not necessarily be regarded as a replacement for human-decision-making but an assistant or advisor that has different degrees of fallibility. As computers become more human-like it is also a question of whether humans can become more fault-tolerant of computers.
Several main applications of context-awareness are discussed in detail including temporal-awareness, spatial-awareness and ICT system awareness, e.g., to support user mobility. Context-aware applications are producers of metadata about the state of the environment and about the state of the user tasks. This metadata may be represented using an expressive semantic data models such as ontologies or using less semantic data models because of the overhead in processing semantic data messages. Context-aware systems have the potential to produce a deluge of metadata to equal the mass of data produced, potentially overloading the user. The main challenges for context adaptation are discussed.
Intelligent Systems, artificial intelligence (AI), software agent, knowledge-base (KB), semantic, ontology, machine learning, reactive, goal-based, first-order logic, propositional logic, reasoning, inferencing, soft computing, probabilistic reasoning, Bayesian network, fuzzy logic, searching, planning, partial order planning, hierarchical task network.
Intelligent systems (IS) are systems which use AI to perceive, reason, learn, and act intelligently. There are multiple ways to classify IS. A basic system is one that senses its environment and can act on it. Design issues for IS include whether a system has a local versus global model of its environment, whether the model is unilateral versus bilateral, whether an external model of the environment exists, if an IS combines an external environment model with an internal model of self, how the model is represented and what the model represents or mean.
The main representations for an IS model include semantic knowledge-based models, classical logical models, and soft computing models. These representations can be combined, for example, strong ontology models combine a semantic with a classical logical representation. An issue here is that multiple types of logic are needed. Iit can be complex to combine multiple logic model representations.
In terms of what the IS model represents, five basic types of IS system architecture are described as proposed by Russell and Norvig (2003). These types of model can directly underpin several types of UbiCom Model. Reactive IS which lack a model of an environment or system can be used as a design for context-aware systems that depend just on the current environment context. In a less basic context-aware system design, the actions will also depend upon the environment model. Some environment context changes, e.g., user context changes, are however less well-defined, less crisp. In this case an environment model type IS which uses soft computing techniques, or uses which aggregates varying values of the environment context into a best estimate, e.g., collaborative filtering systems, seems a suitable design.
Simple control systems could be based upon reactive and utility type IS designs. More complex controllers could use uncertainty type knowledge models. Autonomous systems are similar to IS in the sense that these can be goal-directed and policy constrained although an IS would need to incorporate explicit control algorithms. Autonomous systems can incorporate plan-switching and replanning to achieve a goal when internal system and external environment changes which may otherwise cause a system to fail to achieve its goals. Utility-based designs can allow the system to select from multiple plans to achieve multiple goals.
Architectures for UbiCom systems are sometime needed which can operate in multiple environments. Two different hybrid designs which differ along two different dimensions are given. Firstly, whether or not the different IS models are executed in a chain (vertical layered model) or in parallel (horizontal layered model). Secondly, for the chained model, the order in which the models are executed may greatly affect the computation efficiency, a particular concern for ICT resource constrained systems. The chained design is based upon environment events which are first processed in a human environment model, then in a goal-based model, then in a human reactive model and finally in a physical world reactive model.
Finally, a learning-based IS enables any of the IS system designs to acquire its system model either with the help (supervision) of some active entity in the environment or without (unsupervised). This is a good design to support adaptation in dynamic environments. Often the representation of the model is linked to how the system uses the model (the type of system architecture). Many types of system architecture can be combined to form hybrid models.
Interaction multiplicity, mediator, proxy, matchmaker, broker, cooperation, coordination, cohesion, joint-planning, norms, organisation, competitive, self-interested, electronic markets, micro-economic, agreements, consensus, convergence, negotiation, auction, voting, multi-agent system, speech act, interaction protocol, agent communication language, agent platform, agent-oriented software engineering, social network, Web 2.0, recommender system, referral, collaborative filtering, trust.
As UbiCom devices become better and universally connected, we finally have the building blocks to enable smart interaction. Smart interaction refers to a richer interaction beyond using basic universal network communication protocols. Three basic types of interaction are considered: interaction between relatively dumb participants (peer to peer interaction), interaction between cooperative intelligent participants and interaction between competitive intelligent participants.
Firstly, basic interaction that focuses on interaction involving unknown receivers, transmission of too many messages and transmission by unknown senders, is considered. Interdependencies may exist between senders and between receivers, such as the use of shared resources, e.g., communication channels or shared information resources. Interaction can be controlled at different levels of the communication protocol stack, e.g., use of MAC schemes versus the use of active mediators at the application level.
Cooperative interaction enables multiple systems to work together through sharing information and tasks, and more advanced models involve sharing knowledge, goals, intentions, plans, experiences, etc. Cooperative interaction can be governed using explicit communication synchronisation or using indirect communication such as coordination. Cooperative interaction is also characterised by coherent actions of groups of individuals that can be managed with respect to hierarchical, role-based organisational interaction and using norms and electronic institutions.
As an increasing number of diverse smart autonomous configurable and networked devices are introduced into physical spaces, expanding the virtual ICT space more into the physical world, the degree of competitive interaction that occurs, will increase. Competitive interaction is interaction that is driven by self-interests, on autonomous participants furthering their own goals, rather than by collaborating to further the (organisational and social) goals of others. Design models to solve the associated resource conflicts and resource allocation problems are essential. Three basic design models are considered: market-places (agreements between service requesters and resource providers), negotiation (more general and often more strategic agreement) and voting (consensus type agreements).
There are two basic solutions to support intelligent interaction, either to design system interaction to become more intelligent or to design individual intelligent systems to interact. Both are needed but the focus in this chapter is on the latter. Three basic designs for interaction are considered based upon network protocols, based upon knowledge sharing and based upon agent communication languages or ACLs. The core of the ACL model is based upon a speech act model which interprets communication as actions which change the state of the world. The use of speech acts involves the use of other protocols to communicate content structures, content logic and patterns of interaction. Hence speech acts are defined as part of a agent interaction protocol suite (AIPS), a multi-layer of sub-protocols that operate at the application layer of a conventional network protocol stack.
Multi-agent systems are considered as a core design for multiple interacting IS. There are many types of IS depending on how the individual IS are defined. The type of MAS considered here is the one which supports a AIPS and has some other core middleware or agent platform services. The development process for creating MAS applications is outlined. Many applications and projects have used this type of MAS model for the system design for interaction of IS. Finally, the chapter ends by considering several intelligent interaction applications involving social networking and media exchange, recommender and referral systems, pervasive work flow systems and trust management.
Autonomy, autonomic computing, automatic, self-star property, reflective systems, self-management, self-configuration, self-optimisation, self-healing and self-protection, complex, self-organising, replication, artificial finite state automata models, cell automata, evolutionary computing, genetic algorithm.
Autonomy is considered a core property for UbiCom systems, enabling systems to operate independently and to have some degree of self-control over their own actions. The benefits to the system are to offload tasks from humans and from the physical world, to execute these more optimally and to repeat them more accurately. The disadvantages are that machines may remove too much control from humans and may attempt to control things in open service environments in a non-optimal, less safe, way.
Autonomic Computing focuses on the ability of system to be designed to self manage different properties such as configuration, healing, protection and optimisation, independent of external control. One of the most common types of autonomous systems are automatic systems which operate independently of humans but which are usually closed systems which perform fixed tasks. More general types of automatic systems are needed to support self-operation in open, heterogeneous, dynamic (ICT, Human and Physical) environments. Autonomous systems can be designed to operate to achieve internal goals, independently, without any external control. Autonomous systems need to be designed to operate as part of systems-of-systems in which other systems are also designed to operate autonomously.
Specific models for autonomous systems can be based upon object-oriented systems, event-driven systems, hybrid goal-based and environment model-based IS, and autonomic computing. At a lower level, autonomous systems can be characterised into more basic sets of self-star properties (over twenty properties are described). Three particular types of autonomous system are examined in more detail, reflective systems, autonomic systems and complex systems.
Reflective systems and goal-based IS systems focus on meta computation, computation to reason about their operational computation. Two different kinds of reflective systems are examined. Self-explaining systems reason about how they operate in order to explain their self-operation to external environments including humans. Self-modifying systems reason about themselves for several reasons, e.g., to adapt their behaviour to the environment.
Autonomic or self-management systems consist of groups of locally interacting autonomous entities that cooperate to maintain system wide behaviour without any external control. The motivation for autonomic systems is to deal with the obstacle of IT system complexity. The model is inspired through analogy with the human body’s nervous system. The group of key self-star properties that characterise autonomic computing consists of self-configuration, self-optimisation, self-healing and self-protection. These may be underpinned by an additional self-star property self-awareness. Three major types of internal self-star system design are proposed based upon how they control system resources: global policy driven local self-star control, global policy driven global self-star control, and local policy driven local self-star control. MAS models can also be used as a design for autonomous systems.
In general complex systems are those whose properties are not fully explained by an understanding of its component parts. Complex systems can arise when there are many possible combinations of interactions because of many interrelationships and interdependencies between them. Complexity can also arise through simple interactions and repeated applications of simple rules. This type of complex system uses self-organising, self-creation and self-replication models. An important model for self-organisation, self-creation and self-replication is based upon artificial life models. Two main types of artificial life model are discussed: models based upon finite state automata models and models based upon evolutionary computing.
Ubiquitous Communication, PSTN, intelligent networks, IP multimedia subsystems, asynchronous digital subscriber line (ADSL), digital radio, software radio, WLAN, WiMAX, Bluetooth, ZigBee, infrared, ultrawide-band, satellite, microwave communication, roaming, network convergence, power line communication, personal area networks, body area networks, mobile users, mobile ad hoc networks, NAT, VPN, group communication, rural networks, control plane, service-oriented networks, content-based networks, programmable networks, overlay network, mesh networks, cooperative networks.
A ubiquitous communication network is a key enabler for pervasive computing. Communication networks are generally composed of networks of networks or internets. It can be quite challenging to interlink these into a coherent whole in order to route or switch messages between different parts of the internet. The internetworking is also complex because they often use different physical media networks with different signal characteristics and signalling systems. There are several different dimensions for designing networks with respect to communication: control versus signalling, analogue versus digital, homogeneous versus heterogeneous, wired versus wireless, access networks versus core networks, LAN versus WAN and ad hoc versus static networks.
To support a vision of ubiquitous computing, communication systems need to deliver content in any media, anywhere, anytime to anyone. But distinct networks still exist for sharing content and tasks. Some limited convergence of different networks content is happening, such as data over mobile wireless, VoIP and IPTV but full convergence is still some way off.
A wireless network deserves a special mention because in theory it supports a more natural model for ubiquitous access, inherently supporting anywhere access not just at fixed access points but also through supporting access by mobile users. However, there is a huge choice of wireless network technologies depending upon: user mobility, coverage, access control, power usage and whether a permanent versus temporary or ad hoc network infrastructure is in place. Body area network support a vision of a person as a local smart environment whilst personal area networks enable a local network to follow the user wherever they are. Smart mobile devices tend to be used to enable people to access remote services rather than to directly access local resources.
Service oriented networks enable services to be offered across networks, which are less dependent on specific networks. Here service orientation to the network may be driven by applications on the edge of the network rather than in the core of the network. Service-oriented network designs include content-based, programmable, overlay, mesh, and cooperative networks.
Process-oriented management, network oriented management, FCAPS, monitoring, accounting, ICMP, SNMP, configuration management, security, fault tolerance, performance, metrics, distributed resource management, Grid, Service level agreements (SLA), policy-based management, rich data, soft data, privacy, biometric user identification, privacy-invasive technologies, privacy enhanced technologies, regulation of user privacy, unattended embedded devices
UbiCom systems are basically distributed ICT systems. Hence, the management of smart devices in virtual computing environments is first considered by reusing and then extending existing robust system and network management models for use with smart devices. Basic process and application management can be handled by the device MTOS. This if often referred to as (computer) system management to distinguish it from network management which was based upon models for telecoms network management (TNM) and data network management. However, today, computers and networks are so closely intertwined that these are combined together.
The main management functions, are expressed as FCAPS (Fault, Configuration, Accounting, Performance, Configuration) which models management functions in terms of fault, configuration, accounting, performance and Security. Accounting is interlinked with monitoring, a core enabler for all the FCAPS management functions. Two of the most widely known monitoring models are based upon the ICMP and SNMP models. Two key challenges for configuration management are to ease configuration and reconfiguration. Ideally, flexible configuration should support plug and play; in some cases unplug and (continue to) play is also desirable. Conflicts when multiple devices interfere with each other need to be dealt with. Performance management has its origins in TNM and is based upon quality of service (QoS) models but quality of experience(QoE) is also a concern for end-users. Security management is considered from three different perspectives: securing the middleware and ICT infrastructure; securing the device itself which may be mobile, prone to eavesdropping, have unsecure access etc; securing information which may even prevent the owner of the data from accessing it (restriction management). Fault or safety management overlaps with security management in that both cover denial of service (DoS) attacks, however, fault management also covers inadvertent operator errors and design errors.
Service Oriented Computing (SOC) Management is currently a hot topic. Three specific types of SOC management model are considered, distributed resource management including Grid computing resource models, Service Level Agreement (SLA) management of services, policy-based service management and pervasive work flow management for services. Each aspect of the system, such as application specific functional operations and generic non-functional operations such as configuration, performance and security, can be modelled as information. These can then be managed using information management techniques for a spectrum of information types from rich to lean, from soft to hard and modelled using a variety of different representations. Although hard lean data such as alphanumeric data is processed to support many kinds of quantitative organisational decision making, many human activities involve the use of richer and softer information. The increasing capability to generate digital information from a multitude of sources over a multitude of channels could easily overwhelm our human ability to be able to use for decision making. Often rather than manage data, data can be managed via accessing good metadata descriptions of information instead.
The use of smart devices in user centred and physical world environments requires the use of a wider range of system management models. Seven different system management models are proposed for use in this wider environment. Different designs and metaphors are also reviewed that go beyond the desktop file system metaphor based upon user task and activity based management that could be chronological rather than pseudonym (indirect via file names). Finally, a major concern of smart devices is discussed that of balancing user privacy against system security, i.e., the greater the system security, the more the system knows about you and that can be potentially tied to a user identity. Three main user privacy management techniques are considered based upon: balancing Privacy-Invasive Technologies (PIT) versus Privacy Enhanced Technologies (PET); provider and entrusted regulation of user privacy; legislative and user-centred approaches to privacy.
Management of smart physical environments concern three main facets: the design of physical world context aware systems, the management of small devices and the management of unattended embedded devices. Context awareness can enhance the management of resources in the physical world such as enabling context-aware power management and ICT resources access but the use of contexts itself introduces complexity which must be managed. Micro and nano device management focuses on the characteristics of their tiny dimensions, coupled with being untethered and possibly being deployed massively in parallel.
Challenges, future technologies, smart interaction, context awareness, low power, sustainable energy use, eco friendly, †natural interaction, analogue to digital conversion, form factors, human intelligence, machine intelligence, posthuman, reality blurring, social issues, promise versus peril, virtual social interaction, accessibility, affordability, legislation, professional skills.
Despite the progress achieved in terms of widespread use of UbiCom, many challenges still lie ahead, with some researchers doubting whether or not the full vision of UbiCom can ever be achieved. A full vision of UbiCom may not even be useful or usable depending on the application. Hence, it may be a better option to consider graded levels of support from minimal, through, basic, medium and high to full. Many innovative system and technologies have been studied, each could become mainstream UbiCom systems over the next twenty-five years.
A holistic framework for analysing and designing the complete spectrum of UbiCom device has been proposed. This framework called the Smart DEI (pronounced Smart Day) model and is based upon three core designs: smart devices, smart Environment and smart Interaction. The internal system operation is based upon five key system properties: distributed systems, context awareness, intrinsic HCI, intelligent systems and autonomous systems. The external system operation is characterised by interactions in three main environments: virtual computing, physical and human environments.
In terms of smart devices, it is likely that smaller, more functional smart devices, more fluid ensembles of diverse devices will be used as more functions migrate from analogue to digital devices. Much richer system interaction and interoperability between more sundry digital systems will become common place. This will lead to unexpected connectivity, impromptu service interoperability and accidently smart interactions.
Smart (physical) environment interaction will need to handle the use of ill-defined contexts versus trying to operate in a context free world. As more of the physical world becomes strewn and embedded with digital devices, designs for lower power and sustainable energy use and eco-friendly UbiCom device will become more critical. Human device interaction will involve more diverse human device interaction, more natural HCI and richer multi-function devices.
The balance between the degree of human intelligence versus the degree of machine intelligence that is incorporated in systems will need to be carefully considered as ICT systems become more capable of augmenting human abilities beyond being human, and as reality become increasingly mediated by technology and becomes more blurred. Machines may displace more human tasks and become more subservient to a seamless virtual computer space People may lose the accountability in society as machines make more complex decisions which are incomprehensible by the majority of the human race People may lose their right to privacy as everything can get logged and replayed and analysed by others. The success of AI may lead to the end of the human race as machines rather than humans rule the world, possibly making decisions to rule out humans from the world.
The use of UbiCom offers social promise as well as peril. UbiCom may increase virtual social interaction at the expense of local social interaction. The affordability and accessibility of these systems in one inclusive world will become more important, else a massive digital divide will open up, splitting society. New legislation for the digital world will be needed along with digitising the legislation itself, in order to deploy legislation across more diverse systems and environments.