In the research of intelligent agents working group has focused on three areas:architecture, development methodologies and applications of agents in networks andtelematic services.
The group has extensive experience in the development of agent-based applications in a great variety of fields, but with a special emphasis in: network management and services; educational settings; personal assistance; and, social simulation.
In the recent past, agent-based software systems have been increased their presence in many industry sectors, such as e-market, economics, surveillance or risk analysis. The importance of intelligent agents and multi-agent systems lays in their suitability in a great variety of computer systems such as the Grid, the Semantic Web, peer-to-peer systems, pervasive computing, and ambient intelligence systems. In this scope, the GSI group focuses on studying and developing architectures, methodologies, and applications of agents in networks and telematic services as well as expert systems. The definition of a model of autonomous agent architecture (proactive, responsive and communicative) was the basis for the realization of an agent development platform in Java (project JAEN). This platform was used in developing educational and research projects aimed at agents (Robocup projects and Disaster 2.0). Currently, our interest in agent architectures focuses on component-oriented models. In this vein, the GSI group has developed several architectures for telecommunication systems Jain SLEE type (TelcoBlocks projects and T2C2) or for network surveillance using uncertain reasoning techniques for network diagnosis based on JADE/WADE.
The group also contributes with several methodologies to facilitate the development of agent-based applications. In this field, the methodology MAS-CommonKADS has gained wide international circulation: it has become a subject of study at several universities and has been successfully used in several developments undertaken. Our current interest in the methodological area is the modeling and definition of testing techniques to develop multi-agent systems in agile environments. BEAST methodology meets these requirements focusing on the communication between the stakeholders and the development team. As agile methodology, it allows a continuous improvement of requirements and takes care about the traceability of these requirements to executable tests. The group is also works on modeling and defining patterns of agreement between agents, i.e., the formalization of shared conceptualizations about the mechanisms to reach satisfactory agreements among agents.
Recently, the Group has undertaken several projects aligned on Personal Assistance research line. Web4.0 aims to mine the user social network and local context information to assist him in common situations such as arranging a meeting, planning a trip or booking a flight. CALISTA brings agent systems to ambient intelligence. In this field, the information from a large number of different sensors is collected to assist the user in a seamless and unobtrusive way. In CALISTA, personal assistants automate multiple processes from the digital home relieving the user of these chores. In both projects, Natural language understanding (NLU) and chatter-bot technologies are included to show the user a natural language interface.
Another application of agent technologies that the GSI group researches on is Agent Based Social Simulation (ABSS). ABSS has become one of the most popular technologies to model and study complex adaptive systems. This technology is being employed in more and more scientific domains: sociology, biology, physics, chemistry, ecology, economy, etc. ABSS combines computer simulation and social science by using a simple version of the agent metaphor to specify single components and interactions among them. The GSI group employs ABSS to test, evaluate and optimize computational systems where the key characteristic is their adaptation to users’ behaviours, i.e. Ubiquitous Computing and Ambient Intelligence. Under the THOFU project, two examples of development that the group is conducting by ABSS are: (1) an adaptive algorithm for emergency evacuation in scenarios where the escape routes change quickly; and, (2) consensus mechanisms in an intelligent hotel to maximize users’ satisfaction when accessing shared services such as a TVs in halls.
Summarising, the group has extensive experience in the development of agent-based applications in a great variety of fields, but with a special emphasis in: network management and services (Project PROTECT, SCARAB, ITECBAN); educational settings (MASTER.web projects, Robocup, Disasters, 2.0); personal assistance (project collaborator, PLEIADES, BOT, Improvise, T2C2, etc.); and, social simulation (THOFU).
You can find further information in our YouTube channel.