Why is Ambient Intelligence Challenging?
Ambient Intelligence (AmI in short) is a multidisciplinary approach aimed at enriching physical environments with a network of distributed devices, such as sensors, actuators, and computational resources, in order to support humans in achieving their everyday goals. These can range from helping businessmen in increasing their productivity to supporting activities of daily living for elderly and people with special needs, such as people suffering from cognitive impairments or physical injuries.
From a technological perspective, AmI represents the convergence of recent achievements in communications technologies (e.g., sensor networks and distributed computing), industrial electronics (e.g., the miniaturization process affecting computational devices, such as wearable sensors), Pervasive Computing, intelligent user interfaces (e.g., alternative communication media such as eyes’ gaze and speech) and Artificial Intelligence (e.g., knowledge representation and context or situation awareness), just to name but few. AmI is not only the superimposition of these disciplines, but it is a major effort to integrate and make them really useful for the everyday human life.
The Handbook is aimed at covering cutting-edge aspects involved in the effective design, realization and implementation of comprehensive AmI applications. In particular, it can be considered as a conceptual path originating from the very basic consideration of human aspects that are involved in AmI systems design, heading to the effective reasoning and acting capabilities that AmI systems must be characterized with to operate in the real-world.
Figure 1. Information Flow in Ambient Intelligence.
In particular, eight key aspects have been identified as being fundamental in the design of AmI systems (see Figure 1).
The first one is human aspects: AmI systems are expected to work in symbiosis with humans, and therefore they must understand not only human actions, but also moods, desires and feelings. In other words, AmI systems should encode models of human behaviour (in the mid term possibly in well-defined scenarios, such as people with dementia or typical home activities) and cognition in order to make predictions and to take purposive actions. These models can be either encoded by designers or learned by the system itself through observation.
Models of behaviour and cognition must be grounded with respect to actual data provided by sensors monitoring the given physical space: sensing is therefore a major topic to address when classifying actual human behaviour with respect to encoded models. Many sensing techniques have been investigated in literature. However, what is important is the relationship between sensory information and the algorithms adopted to process it: among the different issues involved in this process, we can identify data fusion, reliability, real-time responsiveness, symbol grounding and anchoring with respect to knowledge representation structures.
Knowledge representation (KR in short) is a central topic in AmI applications and intelligent systems in general. Borrowing techniques developed by research in Artificial Intelligence, it is now considered a core part of any AmI system. In particular, it is at the basis of many other important aspects to consider: models of human behaviours and cognition can naturally be represented using KR techniques, topological and physical aspects of the environment can be easily encoded, reasoning schemes are structures whose templates are stored in knowledge bases (otherwise called ontologies), just to name a few. However, the main feature that makes KR techniques suitable for use in AmI applications is the fact that they can associate semantic information to represented knowledge, i.e., a sort of meaning characterizing symbols and their relationships.
Many techniques for detecting human activity have been developed and appeared in literature in the past few years. Activity recognition is strictly related to sensing and knowledge representation: on the basis of sensory data and structured knowledge, a major dependence exists with respect to sensory modalities and knowledge representation techniques used to manage information. This issue aroused – and still foments – many philosophical and technological debates about the very notion of “activity”, from which completely different approaches originated.
The recognition of human activities, coupled with semantic knowledge represented within the system, forms the basis of context-awareness. A context-aware system is “something” that is aware of what sensory data mean. In other words, a context-aware system is able to associate meaning (useful for humans as well as for automated systems) to observations, and to make the best use of sensory data once the meaning has been assessed. During the past decade, context-aware applications have been one of the hottest research topics worldwide. Several techniques have been proposed to design aware AmI systems: from this experience, it emerged a tight relationship with respect to the capabilities of activity recognition and knowledge representation.
Once a situation has been assessed (possibly involving many humans, their performed activities and related events), the AmI system is expected to purposively act in order to reach a given objective. Depending on the techniques used to provide the AmI system itself with context-aware capabilities (and, ultimately, on knowledge representation strategies), many methods to support reasoning have been proposed in literature. However, research in AmI has been geared more towards representation than action and task planning: as a consequence, the use of reasoning techniques in AmI applications is still in its infancy.
Achieving a sensible interaction level with humans is probably the main goal in AmI. This objective involves a tight – yet well-defined – coupling between reasoning (and the involved context awareness techniques) and the considerations of relevant human aspects. In a sense, it can be argued that sensible interaction is what is actually perceived by humans when dealing with AmI systems, therefore encompassing not only what kind of information is presented or what action is performed by the system, but above all how this is achieved. Without any doubt, this is considered the most challenging part of an AmI system, and therefore it is far from being clearly addressed in literature.
It is not possible to implement real-world AmI systems (involving all or part of the preceding topics) without taking architectural aspects into account. These refer to the underlying sensory machinery, to the middleware, to real-time and operating systems requirements, to guarantee up-to-date information when needed, and to enforce scalability and robustness, just to name a few. One of the goal of this Handbook is to provide the reader with a comprehensive overview of an AmI system as a whole (although made-up of many interacting parts) thus showing how the various modules can effectively interact.
The Ambient Intelligence Slogans
Ambient Intelligence is a novel and groundbreaking paradigm emerged during the past few years as a consequence of important achievements in various research and more technical fields. As a matter of fact, AmI has tremendous implications in such fields as design, healthcare, tele-medicine and Robotics, just to name a few.
Defining AmI is a rather difficult task, for it encompasses a broad range in the spectrum of interaction modalities between humans and their environment. In spite of this fuzzyness, it is possible to roughly define AmI as a computational paradigm aimed at designing, developing and realizing “augmented” environments (i.e., environments provided with electronic devices, such as sensors and actuators) that prove to be sensitive, responsive and aware of both humans and their activities. In a world where the AmI vision is realized in its entirety, distributed devices seamlessly cooperate with each other and with humans, in order to support them in better carrying out their everyday activities.
The AmI vision is characterized by a number of key ideas which distinguish AmI itself from similar paradigms aroused in the past few years, all of them closely related to the relationships among humans, their activities and their environment.
No more users, but humans! Traditionally, the generic notion of “computation” is meant at providing support for activities people must carry out either for work or leisure. However, it requires purposive and well-defined interaction modalities on the user side and, specifically, users should be skilled enough to benefit from the interaction. On the contrary, AmI envisions a future where purposive interaction on the user side will be kept to a minimum, whereas the “system” will provide information and assistance when needed. In other words, people will no longer need to be users: in order to interact with intelligent systems they will simply act, and the system will be able to autonomously understand their needs and goals.
Non only smart homes, but smart environments! Early research in AmI concentrated on specific types of augmented environments, and in particular on Smart Homes (intelligent apartments devoted to enforce specific aspects of domestic life). As a matter of fact, this approach helped in assessing first results and in conceiving new ideas to pursue. However, smart homes are only one side of the whole AmI vision: outside homes, AmI systems should be silent companions to human activities when travelling, sitting in a cafeteria, at work or in the pool. In other words, each accessed service in everyday life should be personalized and tailored especially for the single individual.
No more visible computational devices!. The fact that AmI shares with other paradigms, such as Ubiquitous and Pervasive Computing, the idea that computational devices (including sensors and actuators) should disappear from sight is an important concern. In particular, service provisioning and specific information should be conveyed through a hidden network connecting all these devices: as they will grow smaller, more connected and more integrated into our environments, the technology will disappear into human surroundings until only the (possibly implicit) interface will be directly perceivable. However, differently from Ubiquitous and Pervasive Computing, the network is the means by which data are gathered by a system that intelligently operate on them in order to provide directed, specific and targeted assistance.
Ambient Intelligence can have a tremendous impact on the future of society: from disaster assessment to elderly care, from surveillance activities to tele-medicine, from the tracking of patients in hospitals to package tracking in factories, just to name a few. AmI techniques can be exploited for the benefit of humans.
Few Keywords
Human aspects. Humans are the central components of AmI. In a sense, an AmI system is meant to wrap humans within an augmented physical space, where implicit and sensible interaction can occur and made effective. Modelling human capabilities is therefore a central topic that AmI must address in depth. Human models can be considered under many perspectives. Among them, we can think of cognitive models, interaction models, or preference models. This is a topic becoming increasingly central in AmI and in Artificial Intelligence in particular, thereby deserving great attention in future treatises.
Sensing. In order to realize the aforementioned implicit, purposive and sensible interaction, an AmI system must be provided with the necessary sensory information. Different sensors have been traditionally used to detect human activity: from the rich, highly expressive but prone to errors information provided by cameras, to the simple, poorly expressive but extremely reliable information provided by binary sensors, a plethora of techniques have been presented in literature. It seems reasonable to assume that different approaches will be required in different scenarios, thereby enforcing modularity, scalability and integration.
Activity recognition. Sensory information must be directed and manipulated in order to make sense of it. Specifically, recognizing human activity within a given physical environment is the first step to support purposive interaction. Many techniques have been proposed in literature during the past few years, all of them characterized by benefits and drawbacks: probabilistic, logic-based or ontology-based techniques to achieve activity recognition are currently hot topics in this area. Cross-fertilization among different approaches is expected to greatly improve the overall system capabilities.
Knowledge representation. It is widely recognized that AmI needs to efficiently and reliably represent the information gathered by distributed sources. During the past few years, in the Artificial Intelligence community, many techniques have been proposed to this aim: ontologies (conceptualization of a given domain) clearly emerged as a powerful modelling tool to encode possibly the overall behaviour of an AmI system, but other approaches, possibly taking relational as well as temporal information into account, are suitable as well. To date, there is no agreement about standard techniques and models, therefore requiring a deep analysis in AmI research.
Context-awareness. A system that is aware of what is going on within a given physical environment is closely part of the AmI vision. Assessing a situation means to characterize the relevant properties of an environment, the activities currently carried out within the environment and the possible consequences. In particular, the notions of “context” and “situation”, when related to humans, aroused many harsh debates about philosophical and practical aspects of AmI since, from them, originate completely different approaches to user modelling, activity recognition and knowledge representation. In particular, context-awareness identifies modalities under which AmI system components cooperate with each other: it requires, from the designer’s perspective, a description of what the system can do, and how to do it. This is probably the hottest topic of research in AmI in the latest few years.
Reasoning. Once the current situation has been assessed, the AmI system is expected to purposively interact with humans. In AmI systems, this interaction ultimately depends on many factors, such as the kind of requested interaction (e.g., physical for people suffering from physical disabilities or injuries or at the information level for businessmen with busy schedule), the kind of information needed (and the related reliability, obtained through merging different data sources), and associated priorities (e.g., understanding which, among many information request, is more relevant for the user given his/her current situation). Reasoning in AmI systems poses unique challenges to techniques developed in Artificial Intelligence so far: real-time requirements, human constraints, plausibility and compliance with common sense requirements. This will be for sure the next hot topic to deal with in AmI systems.
Sensible interaction. This topic describes how to couple context-awareness with reasoning in order to unleash the full potential of AmI applications. Sensible interaction policies must be either encoded within the system or learned by the system itself over time. In both cases, these policies should represent when and why the AmI system should operate, what to do and where to do it. Since this research direction is aimed at encoding how the AmI system interacts with humans, further research in this direction is expected to push the field to its very limits.
Architectural aspects and infrastructure. During the past few years it clearly emerged that AmI applications are complex systems composed of many different interacting parts. This difference is related to both conceptual and technological aspects: information flow should be clearly modelled in order to enforce reliability and efficiency of the whole system, whereas many technologies must be put together, thereby requiring complex interfaces at the software level. For instance, think about providing differentiated services to a businessman travelling from home to the airport: different information must be made available at different times, whereas how information itself is conveyed changes according to businessmen location. To date, these aspects have been mostly disregarded (due to the fact that AmI systems have been rather small), but nonetheless it is for sure one of the main challenges to face.
What is this Book for?
This Handbook is thought of as the book about AmI “the Editors would have read if it had been available when Ambient Intelligence began to be an interesting research field in its own right”. In spite of the many publications labelled as “AmI-related”, the need arises for an exhaustive treatise of AmI research “as a whole”, and not as being originated by many small (and expensive) gadgets or ad-hoc systems considering mutual interactions as a plus.
In the Editors’ perspective, the Handbook is meant at providing the skilled reader with a reasoned view of what is currently pursued by researchers and practitioners worldwide. In principle, it should be possible to read a chapter, to think about it, and to contact its authors in order to discuss about timely topics. More specifically, it is possible to identify the two following main objectives:
- Comprehensive view of AmI systems. The Handbook is intended to provide a clear view of an AmI system in its entirety, from human modelling to architectural aspects related to system engineering.
- Oriented towards current research and future perspectives. In the spirit of the “Handbook of Research” series, each Chapter provides the reader with a close look at what researchers are doing “here and now”.
In the Editors’ opinion, a book on AmI based on the previous objectives is both timely and needed. It is widely recognized that the next step in AmI will pass through a comprehensive description of human aspects, knowledge representation, context-awareness and reasoning. In a sense, this Handbook could be the first attempt in systematizing the field with a look at future developments in the meanwhile. From human perspective, this is a dramatic paradigm shift with respect to how humans themselves relate to technology and information technology in general. From old Science Fiction books and movies in the Seventies down to European Union strategies for the new century, AmI-related technologies are considered fundamental.
Information Science is perhaps the field that can benefit more from the AmI vision. Information will have to be managed in completely different ways: it will have to be stored for quick retrieval, to be light-weight in order to allow quick delivery across many heterogeneous networks, and to be reliable in that humans will depend on it on many activities. To achieve all these goals, research in AmI is expected to enforce the development of dependable infrastructures, to understand and reason upon human needs, to act proactively in the real world.
Technology development will surely be driven by the AmI vision. From the miniaturization of microelectronic components down to innovations in user interfaces, AmI is expected to provide designers with general guidelines as well as practical directions to follow in the mid and long-term period. Furthermore, AmI is not limited to specific environments and scenarios; on the contrary, its application is only limited by human acceptance and imagination. Automotive industry, building and factory automation, smart homes, intelligent building, energy saving policies, hospitals, tele-medicine, home assistance, gerontechnology, remote assistance, and disaster management are the first fields that push up after a brief brainstorming session. Understanding the potential offered by AmI that can be unleashed on these fields is fundamental for the well-being of our society.
From a management perspective, understanding what technology can offer can lead to better decisions in the mid and long-term period. When dealing with the AmI vision, this need is even more important, since AmI is expected to dramatically change our society in unexpected ways and with unprecedented benefits. As soon as AmI applications will be part of our society, new business models (taking AmI principles into account) will for sure be established.
The present Handbook is targeted to researchers and practitioners in Ambient Intelligence. However, since AmI per se is the conjunction of many research and technological fields, it is possible to assume that the Handbook can be useful for researchers and practitioners in Ubiquitous Computing, Pervasive Computing, Artificial Intelligence, Sensor Network and Sensor Technologies, Knowledge Representation, Automated Reasoning, Automated Learning, System Engineering, Software Engineering, Man-Machine Interfaces. Furthermore, for its practical implications, it is possible to argue that interested readers should also be acquainted with eldercare, tele-medicine, gerontechnology, assisted living technologies for people with disabilities.
Many other books have been published on AmI topics. As previously pointed out, these are most of the time edited collections of chapters devoted to specific aspects, such as privacy, human aspects, architecture, distributed computing, and so on. Although these contributions represent cutting-edge achievements in the field, they are not grounded with respect to a more abstract and looking-forward view of Ambient Intelligence as a whole.
We hope you will enjoy reading this Book as much as we enjoyed collecting all the material presented herein. It has been a daunting task: a special thanks to all the Authors, the Reviewers and the IGI-Global personnel, especially Elizabeth Ardner and Christine Buffon for their endless support and patience.
Fulvio & Nak