Multi-Agent and Multi-Robot Systems
The Department of Computer Science at Bar Ilan University is home to the largest artificial intelligence group in Israel. The group is known worldwide for its many extraordinary scientific achievements over the past two decades. It includes 11 senior faculty members and dozens of graduate students, postdoctoral fellows, programmers and computer engineers.
One of the main strengths of the group is intelligent robotics, and in particular the study of multi-robotic systems, led by Prof. Gal Kaminka and Prof. Noa Agmon. Among the many interesting topics explored by the two are problems of multi-robot planning in adversarial environments, multi-robot reconnaissance, navigation, coverage, task allocation, and formation. Much of the group's work on reconnaissance, exploration, and multi-robot configurations has been widely publicized, resulting in several technology transfer programs and patents. Agmon's focus is primarily on the strategic behavior of robots in adversarial environments, and the use of theoretical means to model, analyze and solve these and other realistic robotic problems. Kaminka is more concerned with controlling teams of robots and agents on the one hand, and monitoring its complements by the agents themselves on the other. He initiated the annual ARMS (Autonomous Robots and Multi-Robot Systems) workshop at AAMAS, which is now the main meeting point for all robotics from the AAMAS community. One of the founding results of his research is the understanding that teamwork, as a system of general mechanisms for cooperation, can be automated and computerized. This allows robots to cooperate well, cheaply, and to do so with humans as well.
Another area that attracts much of the group's attention is the study of the theoretical foundations of interactions between multiple software agents, focusing on settings where the agents represent different entities, with conflicting interests. With the help of mathematical and algorithmic tools, Prof. Yonatan Uman analyzes such definitions, and develops methods to promote social goals, such as fairness and well-being. In an interesting recent project, Uman and his colleagues extend the decades-long study of so-called ``pie cutting'' from the one-dimensional to two-dimensional case. In the process, they developed the first ever effective protocols for the fair distribution of land. Various other aspects of multi-agent systems are studied in the AIM laboratory led by Prof. Sarit Kraus. Kraus is recognized for her many contributions to artificial intelligence subfields such as strategic negotiation, collaborative planning, human-agent interaction, coalition building and non-monotonic logic. Today it is engaged, together with many partners from both industry and academia, in developing systems and technologies that will revolutionize various traditional professions. These include a virtual communication therapist, culturally sensitive systems that collaborate, negotiate and argue proficiently with people, systems for training law enforcement officials to interview witnesses and suspects, and persuasion systems that generate advice for drivers about various decisions involving conflicting goals. Other active members in the field of multi-agent systems are Prof. Avinathan Hasidim and Prof. David Sarna. Hassidim is engaged in designing algorithmic mechanisms, developing new algorithms and implementing them in real world markets. The list of his latest projects includes impressive applications of new methods for drawing medical interns in Israel, university admissions for psychology and designing a national school selection system. Sarne focuses on exploring the role that information plays in multi-agent settings and in particular the dynamics that arise from the introduction of information intermediaries/platforms into such environments. Another subject that interests him is the design of methods for providing intelligent information in the interaction between a person and an agent, and in particular methods for providing intelligent advice.
The Bar-Ilan group is also very active in the field of natural language processing (NLP). The leading group members in the field are Professor Ido Dagan and Professor Yoav Goldberg. Dagan and his group focus on applied semantic processing of texts, using an empirical approach based largely on unsupervised and supervised learning. His current work involves developing a generic framework for representing the knowledge expressed in texts. This framework is based on the use of natural language constructions (rather than synthetic languages) as the basic building blocks for knowledge representation, with the goal of an open and extensible schema for unlimited textual knowledge representation. While Dagan's work focuses on higher-level semantic tasks, Goldberg's research involves the lower-level building blocks of natural language processing: parsing sentences into their syntactic structures, deriving continuous vector representations for words, and building tools and methodologies that enable cross-domain and generalization between languages.
Last Updated Date : 22/11/2022