Advancements in self-driving vehicles, smart factories, multi-robot systems in logistics, distributed computing, financial technology, and many other fields are rapidly expanding and deep impacting individuals and society. Autonomous and Multi-Agent Systems (MAS) are a well-established research area concerned with modeling and reasoning on these kinds of systems.
When MAS models are employed in safety-critical contexts such as healthcare, computer security, and advanced manufacturing, their correct behavior is a strictly fundamental requirement, as the fault of an even single component might irreparably compromise integrity and security. Nevertheless, the current designing and adjustment techniques are still sub-optimal as for their efficiency and quality. This is due to the fact that they are still manually dealt case-by-case and not yet fully understood and systematized.
BeGood aims at developing the theory and the computational tools for the automated production of the just rules that will guide designers of MAS to a correct specification and behavior of their systems. Remarkable discoveries by the applicant in the context of Rational Synthesis and Strategic Reasoning, and their connection with Verification and Synthesis in Formal Methods, uncovered a novel scientific direction and a potential breakthrough in the automated production of rule-making mechanisms. This will bring the theory and practice in multi-agent systems to the next level, thus impacting the cutting-edge technology based on these powerful models.
The outcome of the project will realize self-governing mechanisms for the correct design and
implementation of behavioral rules in Multi-Agent Systems. This will impact the practice in
designing networks and intelligent systems in several contiguous research areas such as: planning,
robotics, and process management.
BeGood is a high risk, high gain project. It is high gain, as it will introduce radical changes in synthesis and rational synthesis, whose correct application is already revolutionizing the practice in proximal science and technology domains. Our knowledge of behavioral rules in MAS is much behind what is needed to fully employ synthesis techniques. The project will introduce a new generation of self-programming mechanisms, which will shift the design paradigm to a next level of understanding and implementation of Multi-Agent Systems. The notion of behavioral rules belongs to different areas: formal methods, multi-agent systems, game theory, logistics, robotics, Internet of Things. The ambition of BeGood is to find a unifying theory that will contribute to interconnect and facilitate the transfer of knowledge in all these disciplines. BeGood is also high risk, as we need to combine stand-alone mechanisms from different disciplines, which might have heterogeneous representations. Particularly challenging, then, is the task of composing and/or integrating them into effective and readily implementable approaches. This risk is mitigated by the fact that the integration of such techniques has been proven efficient already. The extension from the synthesis to rational synthesis itself is an excellent example of how such coalescence between formal methods and multi-agent systems is successfully achieved. Another big challenge is in the complexity of the solutions. When dealing with synthesis problems, even the most basic ones, research has shown theoretical lower-bounds for its complexity. However, their implementations still have good performances, and impact positively in practice. This is also because such theoretical bounds are based on the worst-case complexity analysis, in which usually real-world scenarios do not fall in. As a matter of fact, the very same synthesis is a well-established method for the practical implementation of correct-by-construction algorithms.
The rational synthesis vision is becoming central to the AI community. I am one of the few researchers employed at the intersection of the Formal Methods, Multi-Agent Systems, and Artificial Intelligence communities. My background in logic and synthesis, combined with my experience in multi-agent systems produced important scientific results, witnessed in my publication record in top AI and Formal Methods venues. This makes me outstanding in the pool of potential researchers that can deliver this project.