Distributed Artificial Intelligence
Future Research Themes
We identified a number of problems to be solved in order to make progress in the domains of multi-agents.
Agents offering deep technical expertise
Systems developed in industrial environments, for example in concurrent engineering, include very technical agents, e.g., using complex numerical computation methods. We selected the case of agents working in the domain of finite element methods, which already have a sophisticated object structure (active objects). We would like to see how such agents can be inserted in systems like DIDE, and what are the missing mechanisms necessary to allow the agent to capitalize knowledge associated with the problems that it must solve.
The work is currently done by Bomme, a PhD student at the EPFL (Ecole Polytechnique Fédérale de Lausanne).
Agent systems for capitalizing knowledge
We have developed tools and methods which allow us to tackle large project. In particular, we are interested by systems of agents capable of capitalizing knowledge in big technological projects, like the development of a new car, or the management of knowledge available inside a laboratory.
A new project in cooperation with PSA and the GSM department at UTC is currently being started, another one is in progress with the Petrobras group in Brasil within an Alfa European project (X-MAS project).
Mechanisms for integrating a human user
We proposed the concept of proxy to integrate human actors into agent systems. It will be necessary to explore two functions: filtering information, and offering proactive behavior (which requires modeling intentions). This should be done with the department of Human Sciences (TSH) at UTC.
Automatic acquisition of implicit knowledge
Using the proxy allows to record temporal graphs. We need now to develop tools and methods for segmenting and indexing such graphs, and link them with explanation modules.
Return to main page