Our group is working now on the following main areas:
Multi-agent systems paradigm is a growing interest area in AI. It is due to its application to complex problem resolution, where classical techniques falls in the obtention of a satisfactory solution. Multi-agent systems (MAS) face up to the necessity of communication and collaboration among autonomous agents (entities which behaviour is guided by themselves)..
- Control processes
- Management of complex industrial environments
- Mobile robotics
- Information search and retrieval
Within the MAS research line, the GTI-IA is working in the following areas:
1.1) MAS architectures:
It specifies how the MAS can be decomponed into a set of independent modules (agents) and how these agents can interact. The total set of agents and interactions must provide a reponse to the problem of determining actions and future internal states from the current world situation and the internal state of each agent.
1.2) Methodology and Software Engineering for MAS
Principles and basic lessons from Software Engineering and Knowledge Engineering must be applied to development and implementation of MAS. Currently, almost all agent-based software is developed by following non-rigurous, design methodologies and using limited specifications depending on the design requirements. Some considered topics in this area are:
- Analysis and design techniques for MAS
- Specific ontologies for different requirements and models of agents
- Tools supporting the development process of MAS
1.3) Communitacion, coordination and negotiation among agents
Modular and open design of intelligent distributed systems make the resulting system to be composed of multiples autonomous intelligent agents. Each agent has different capabilities. Efficient use of such agents in distrubuted problem-solving requires mechanisms for controlling and coordinating the behaviour of individual agents. The final goal is the fullfilment of the whole system goal. The interest is focused on the social behaviour of intelligent entities. It deals with the investigation of behaviour models, co-operation strategies, negotiation models, etc.
1.4) Holonic systems:
Holonic systems are intelligent production systems, formed by autonomous and auto-configurable units, called holons, which collaborate to reach the global goal of the production system. The main purpose of holonic systems is to obtain stability in the presence of disturbs, adaptability and flexibility before changes and efficient use of available resources.
1.5) Agent-Based Simulation for Manufacturing Systems:
The manufacturing field is an area where the application of simulation is an essential tool to validate methods and architectures before applying them on the factory floor. Despite the great number of simulation tools that are available, most of them do not take into account the specific requirements for the "new manufacturing era". Features such as proactivity, reactivity, flexibility and sociability which are provided by the Multiagent System Technology may be useful to fulfill the specific simulation needs of the new manufacturing requirements.
1.6) Learning Meta-Reasoning Knowledge in Multi-Agent Systems
One of the main problems of any agent regarding its efficiency is the named "meta-level control problem". This is the problem of how to sequence control activities (planning, coordination, ...) and domain activities (actions over the environment) without wasting too much resources. The process of reasoning and acting over this sequencing process is called the meta-reasoning process, and it is a fundamental piece of the adaptive trait of an agent. This work deals with adding a learning algorithm to this meta-reasoning process so that some of this meta-reasoning knowledge may be learnt by the agent.
1.7) Multi-Agent Systems in Virtual Reality Environments
The main purpose of this work is to integrate Artificial Intelligence and Virtual Environments, taking into account that the main purpose is to reach an acceptable feeling of immersion for the user, to get an illusion of credible behaviour on computer-controlled beings.
One of the most interesting AI technique for this purpose, due to its scalability feature, is the Multi-Agent Systems (MAS) approach. In this way, this work pretends to establish a framework that integrates a MAS and a Virtual Environment for developing Intelligent Virtual Environments, so that a designer will be able to not worry about the low-level management and interaction with the virtual world. This will allow him to focus in the implementation of the Artificial Intelligence peculiarities of his agents, that is, in their deliberation process whatever technique he will use for them (neural networks, FSM, rules, etc ...).
Real-Time Artificial Intelligence is a discipline that incorporates problem-solving techniques used in AI environments with real-time constraints. These environments need a valid response in bounded time intervals for guaranteing the correct working of the system. Classical AI techniques must be adapted to be applied in such environments.
- Industrial processes control
- Aeronautic systems control
- Monitoring and responsiveness in health-care systems
- Real-time traffic control
- Resource management for telecommunications
- Distributed sensor control
- Mobile robotics
Autonomous systems in robotics are designed to realise tasks without the supervision of human controllers. There are several advantages: to minimise the controller fatigue, to minimise the dangerous woks risks, to minimise the operational costs and to improve the products and operations quality control.
- Underwater works
- Space exploration
A Real-Time system is a computing system in which the accuracy of the response depends on not only the logical accuracy, but also the time instant in which it is obtained.
- Industrial processes control.
- Signal treatment.
- Multimedia applications
The traditional languages for developing rule-based systems are not appropriate for real-time environments, due to the enormous difficulty that implies their timing analysis. For this reason we have developed a new language, and its pattern matching algorithm that fulfills the necessary conditions for allowing its timing analysis.
- Industrial processes control
- Mobile robotics
6) Automatizing the chromatic sensorization process from works of art and their analysis using AI techniques:
The analysis of chromatic data from works of art (paintings, murals, ...) is important to know how these works are affected by the ilumination of the place where they are exhibited, the passing of the public, ... Until now, data was collected manually, implying few data to be analysed by the art expert. The purpose of this research line is to automatize this data collecting, by means of using cartesian robots. This robots allow to make mesurements in a bigger quantity of zones of the work of art. Therefore, more information will be available to extract conclusions. On the other hand, it is intended to use AI techniques to analyze these data facilitating the work to the art expert who has now to manage a quite bigger information volume.