Volha Bryl's research interests
The research areas I am currently interested in (and the corresponding projects and results) are the following
- Agents and multi-agent systems, agent-oriented software engineering
(Tropos and MEnSA projects)
- Goal- and agent-oriented requirements analysis and organizational modeling
(MOSTRO project, METTEG'07 paper)
- Automation support for requirements analysis (CoopIS'06 paper and
future PhD thesis)
- Requirements analysis vs. business process modeling
(CLIMA-VIII paper)
- Modeling and analyzing requirements for secure and trusted systems
(CAISE'06 and AOSE'06 papers,
SERENITY project)
- Self-configuring agent systems design (SOAS'06 paper)
- AI planning and its applications (CoopIS'06,
CAISE'06 and SOAS'06 papers)
- Game theory and mechanism design and their applications (CoopIS'06 paper)
Agents are known not only as new programming paradigm but also as a tool to model and analyze organizations and
complex socio-technical systems. Agent-oriented programming, or, more generally, agent-oriented software engineering is
developing and gaining its popularity both in academia and industry, though it is not yet as mature as the
object-oriented approach is. There exist a number of proposals of agent-oriented methodologies (e.g. Gaia, MESSAGE,
Prometheus, PASSI, etc.) which aim at supporting the whole software development process from early requirements to
implementation. The problem I work on currently for my future PhD thesis
lies in context of Tropos, an agent-oriented methodology which gives a
particular attention to early stages of software development cycle, namely, requirements analysis. In Tropos, requirements
to a system and its organizational environment are modeled and analyzed in terms of agents, their goals, and social
dependencies among agents.
During requirements analysis and design of a complex socio-technical system one has to cope with a fundamental problem
of finding optimal/good-enough delegations to a set of system agents which collectively fulfill a given set of goals.
These goals are initially assigned to the agents who may not have enough capabilities to satisfy them, so they are
decomposed and delegated to other agents, thereby creating networks of delegations. The process ends when all initial
goals can be fulfilled if all system agents deliver on their delegations. Exploring the space of alternative dependency
networks is a difficult design task, and there are no generic criteria to guide the design process by determining whether
a solution is good-enough, or even optimal.
The main objective of my research is to build a framework for the automatic selection and evaluation of alternative
dependency networks. This includes
- formalizing an organizational setting of a system and the desired properties of a solution
(i.e. a delegation network) in terms of first-order logic,
- applying AI planning techniques to generate alternative delegation networks with the help of off-the-shelves
planning tools,
- constructing evaluation metrics to assess the generated solutions, including the use of game theory ideas to
determine whether an alternative is an equilibrium (i.e. no agent can do better with respect to its own goals by
adopting a different strategy for delegating and accepting delegations).
The key ideas of this approach and the current results are formulated and illustrated in
CoopIS'06 paper.