Guest Editorial: Special Section: Ambient Intelligence for Large Premises

Ambient Intelligence (AmI) relates to smart spaces, where pervasive computing and other information technologies are used to make physical environments responsive to people's presence and needs. The responsiveness means here knowing what information and actions are needed, where and when they are needed, and how to perform them. These objectives depend on the context of the system, which comprises features of the system itself, the physical space, and people present within. The objective of AmI is, hence, to minimize the need of explicit actions by users in their everyday activities. This can be done through the AmI's actuation in the physical environment based on explicit gathering of environment information and user preferences and by understanding implicit user behaviors in a certain context.

AmI systems usually comprehend networks of heterogeneous sensing, processing and actuation devices. Their pervasive technologies need to be constantly available and continuously coordinating their activities to provide a coherent and cohesive service to users.

Despite recent achievements, AmI aims at ever growingly ambitious goals in terms of the size and number of its smart spaces, the number of served users, and the level of adaptation to them. By extending the AmI concept from a single smart space of a limited size to Large Premises (AmILP), the user experience can be significantly improved as the AmI can monitor the users throughout the premise and thus provide them with services of better context-awareness. The available integral information of the constituent spaces and the related user behavior can be used to build richer context-aware models and provide a more meaningful and personalized support to the users in every part of the premise.

AmILP comprises complex and large networked systems, possibly made of multiple different subsystems and devices. These elements need to interact and coordinate with each other and with users, as in conventional AmI. However, the differences in scale pose new and specific challenges.

New requirements emerge related with big groups of people moving in premises that fall beyond the classical closed and controlled environments of most of AmI systems. The ways of interaction, the expected services, and the behaviour of people acquire a new dimension and variability in such interconnected smart spaces. AmILP systems need to adapt to the movements and other actions of space occupants by using large numbers of multiple and heterogeneous resources in frequently uncontrollable and changing physical environments. The complexity and the magnitude of operations and decision making in such a context require innovative computation and simulation tools and methods.

This special section is focused on certain particular challenges for AmILP and their potential solutions. The three papers appearing in the section present recent achievements in the area of AmILP with a common concern on the problem modelling and simulation to address these issues.

The cost of the development, deployment, and testing of AmILP components and the whole AmILP system can be very high. This process usually involves multidisciplinary groups of experts who use computer simulations as a key tool for the development of AmILP systems. However, simulating complex AmILP systems is a hard work since it is difficult to guarantee that the simulated models faithfully mirror the real world and reflect the requirements of all involved experts.

Fernández and Fuentes in “Extending a generic traffic model to specific agent platform requirements” propose a Model-Driven Engineering (MDE) approach to work with simulations in AmILP. Their work focuses on Modelling Languages (MLs) that define systems at different levels of abstraction and for different types of experts. These models are used to produce or modify artefacts (e.g., other models or source code) using automatic transformations. This development process makes explicit all the information required to develop the simulation and system and facilitates understanding it.

The proposed approach is brought out using a running case study on Smart Roads. The paper introduces a high-level ML for road traffic with extension mechanisms. The new languages are used to specify new traffic theories and simulation platforms, while establishing conceptual mappings among them that are implemented with transformations.

Gomez-Sanz et al. in “Requirement engineering activities in smart environments for large facilities” focus on the interplay between systems and the physical and human environment in simulations. In large premises, an AmI system deals with flows of people constrained by the physical environment. Simulations need to consider how these interactions happen and facilitate easy understanding by development team experts with different backgrounds (e.g., computer scientists, civil engineers, architects, sociologists, and psychologists). To this aim, Gomez-Sanz et al. propose an annotation tool and apply it on their simulator for physics of crowd flows in large premises. The tool allows adding comments to the simulation and its review by experts for three different tasks: requirements elicitation, system specification, and verification.

In the case of a hazard in large premises, AmILP systems are a useful solution to minimize evacuation–related casualties. Lujak et al. in “A distributed architecture for real-time evacuation guidance in in-door smart spaces” present an AmILP system application for context-aware coordination of crowd flows in emergency evacuation. The system provides a personalized evacuation route recommendation to each evacuee. The objective is to minimize the overall evacuation egress time considering individual constraints, fairness among routes, and possible congestion. The evacuation is coordinated based on a distributed optimization of evacuation routes considering individual evacuees' mobility constraints.

In relation to context awareness, the proposed approach includes complex event processing for inferring the context by exploiting real–time data. The paper discusses the underlying event models and gives some examples for appropriate rules for achieving situation awareness. Their proposed AmILP model captures relevant features of the situation (e.g. sensor readings), as well as the behaviour of people related to the environmental context and potential mobility constraints. The paper includes a case study that shows its functioning dynamics in several settings.

The guest editors are thankful for the support of all the people who contributed in some way to this special section. We are grateful to the authors that sent their high-quality papers for consideration, and to the reviewers who participated in the evaluation process. We want also to thank for the great work of the journal production team in the organization of this Special Section, particularly to the ComSIS Editor Prof. Mirjana Ivanoviæ. Finally, we acknowledge the support of the MOSI-AGIL-CM project (grant S2013/ICE-3019) supported by the Autonomous Region of Madrid and co-funded by EU Structural Funds FSE and FEDER.

Guest Editor
Marin Lujak, IMT Lille Douai, France

Guest Editor
Rubén Fuentes-Fernández, GRASIA, University Complutense of Madrid, Spain