Adina Magda Florea

A Context-Aware Multi-Agent System as a Middleware for Ambient Intelligence

There is currently a lot of work in Ambient Intelligence—or AmI—reporting on specific scenarios, or on implementations of particular cases. In the same time, there is a common agreement of the fact that AmI applications should be pervasive, covering a large number of devices, assisting a large number of people, and serving a large number of purposes. In an attempt to achieve scalable scenarios and implementations, we have focused our research on the development of a generic middleware layer for the context-aware transfer and exchange of information between devices. This paper presents a model for a such middleware, based on software agents, in which context-awareness is implemented both in the agent’s representation of context information, and in the logical topology of the agent system. The model is oriented towards decentralization of the system and relies mostly on local behavior. The paper also reports on several proof-of-concept applications that have been developed and tested using the proposed model, proving thus the validity of the approach.

A Dynamic Strategy Agent for Supply Chain Management

The international trading agent competition (TAC) provides a dynamic trading environment aimed at testing different strategies to be used in the design of autonomous agents for e-commerce and automated marketplaces. The TAC supply chain management (TAC/SCM) scenario defines a complex six-player game with severely incomplete and imperfect information, and high-dimensional strategy spaces. The paper presents PhantAgent, a successful trading agent in the 2006 TAC/SCM competition. The paper discusses PhantAgent architecture, the basic ideas and strategies that were used in its design, and emphasizes that the unpredictable nature of the environment is the main challenge in building autonomous agents in real-world environments

Reconsidering the social web of things: position paper

The notion of a Social Web of Things (SWoT) appears in recent works at the convergence of the Social Web and the Web of Things. In our vision, a third dimension is needed: pro-activeness. We propose to extend and transform social networks by integrating autonomous and proactive things. In this paper, we discuss the evolution of the Web on several dimensions, leading to our vision for the SWoT. We discuss the challenges that need to be addressed, a possible approach for addressing them and we illustrate the applicability of the SWoT through a motivating scenario.

Graphs and Patterns for Context-Awareness

A central issue in the domain of Ambient Intelligence is context - awareness. While previous research in the field presents complex context-aware infrastructures, but with little flexibility and fixed context representations, this paper presents a simple, flexible and decentralized representation of context, for the detection of appropriate context-aware action. This representation is inspired from notions like concept maps and conceptual graphs. A formalism for context patterns, that allows the detection and solution of problems, based on the user’s context, is also proposed.

Multi-Agent Systems: A Paradigm to Design Ambient Intelligent Applications

In this paper we present a Multi-Agent System (MAS) paradigm and discuss how it can be used to design intelligent and distributed systems. The main features of this MAS, such as natural distribution of the system, inherent intelligence of its agents, and their mobility help address a large scope of distributed applications including the domain of ambient intelligence. Other features of the MAS, like multiagent planning, context-awareness and adaptation are also very useful since they bring added value, by allowing to implement intelligent and collective behavior. The paper also presents a scenario of ambient intelligence and shows how it could be designed using the MAS paradigm.

Responsive Decentralized Composition of Service Mashups for the Internet of Things

Applications envisioned for the Internet of Things (IoT) would generally have to fulfill their design goals by mashing up devices and digital services in a manner that is both flexible, such that they can adapt to dynamic environments, and responsive, such that they can react to sensor and user input in a timely fashion. Most existing approaches for the development of IoT applications rely on precompiled mashups that are highly responsive, but inflexible due to their static nature. At the other end of the spectrum, fully automatic composition of services results in IoT mashups that are highly flexible, but responsive only for small numbers of IoT services. This paper presents a middle ground approach: goal-driven software agents are equipped with precompiled mashups and cooperate with one another to compose their mashups at runtime in pursuit of their goals. Agents are interconnected via relations that enable them to discover and interact with one another in a flexible manner. To support our approach, we provide an open-source platform that facilitates application development. We used this platform to implement a realistic IoT application that achieves its design goal by mashing-up multiple heterogeneous devices at runtime. Evaluation results suggest that applications remain responsive when scaling to many devices and for relatively large mashup compositions.

CONSERT: Applying semantic web technologies to context modeling in ambient intelligence

Representation and reasoning about context information is a main research area in Ambient Intelligence (AmI). Context modeling in such applications is facing openness and heterogeneity. To tackle such problems, we argue that usage of semantic web technologies is a promising direction. We introduce CONSERT, an approach for context meta-modeling offering a consistent and uniform means for working with domain knowledge, as well as constraints and meta-properties thereof. We provide a formalization of the model and detail its innovative implementation using techniques from the semantic web community such as ontology modeling and SPARQL. A stepwise example of modeling a commonly encountered AmI scenario showcases the expressiveness of our approach. Finally, the architecture of the representation and reasoning engine for CONSERT is presented and evaluated in terms of performance.

Multimodal Indoor Tracking of a Single Elder in an AAL Environment

In this paper we present a proof-of-concept practical architecture of a system for tracking older adults living alone in an ambient intelligence environment by using multi-modal sensory information, such as audio and video input from Kinect and Arduino devices. This AAL (Ambient Assisted Living) system collects data into a Database of Trajectories, used to automatically generate training examples. We then show how the Database of Trajectories can be used in conjunction with a set of well-trained algorithms to alert remote caregivers in case of an unfortunate event, such as the elder falling down - one of the leading causes of injury and death.

Embedding emotions in an artificial tutor

Emotions have been shown to have an important role on several human processes such as decision-making, planning, cognition, and, in particular, learning. However, many e-learning systems disregard the emotional aspects of the learning process and the impact of these aspects on the students performance. The paper presents an intelligent agent with synthesized emotions that influence its beliefs, desires, and intentions in order to drive the agent behavior. Based on this model, we propose an artificial tutor, integrated in an e-learning system, which develops lectures plans and tailors its behavior towards the student according to the student reactions and his/her presumed affective states.

A Multi-agent System for Human Activity Recognition in Smart Environments

Activity recognition is an important component for the ambient assisted living systems, which perform home monitoring and assistance of elderly people or patients with risk factors. The paper presents a prototype system for activity recognition based on a multi-agent architecture. In the system, the context of the person is first detected using a domain ontology. Next, the human position is obtained and together with the context forms a sub-activity. The sequence of successive sub-activities is then assembled in a human activity, which is recognized using a stochastic grammar.