· Lecture 1: Ana Lucia C. Bazzan (UFRGS, Brasil)
Title: Beyond reiforcement learning in multiagent systems
Abstract: Learning is an important component of an agent’s decision making process. Despite the diversity of approaches in the machine learning area, in the multiagent community, learning is associated mostly with reinforcement learning. Given this background, this talk has two aims: to revisit the old days motivations for multiagent learning, and to describe some of the work addressing the frontiers of multiagent systems and machine learning. The intention of the latter task is to try to motivate people to address the issues that are involved in the application of techniques from multiagent systems in machine learning and vice-versa.
· Lecture 2: Augusto Loureiro da Costa (UFBA, Brasil):
Title: Inteligent System on Chip (IntelliSoC) : A System on Chip to a Cognitive Autonomouos Agent Architecture
Abstract: A System-on-a-Chip (SoC) for the execution of cognitive agents will be peresented in this lecture. The computational architecture of this SoC will be presented using the cognitive model of the Concurrent Autonomous Agente (CAA) as a reference, since it has been successfuly applied in numerous intelligent robotics applications. The cognitive model of the CAA comprises three levels that run concurrently, namely (in the order of growing cognitive complexity) the reative level, the instinctive level and the cognitive level. The reactive level consists of a collection of behaviours and executes a fast perception-action cycle. The instinctive level receives perceptions from and sends the active behaviour to the reactive level. This level uses a Knowledge Based System (KBS) as automatic reasoning mechanism for plan execution through the selection of reative behaviours. The cognitive level receives symbolic information from the instinctive level to update its logical world model, used for planning. It also sends new local goals to instinctive level. Thus, this presentation will propose a novel SoC whose architecture will fit the computational demands of the aforementioned cognitive model, allowing for fast, low-level, embedded intelligent applications.
· Lecture 3: Ricardo Azambuja Silveira (UFSC, Brasil)
Title: Learning environments based on multiagent systems
Abstract: MAS technologies have been used in different educational applications,such as Intelligent Tutoring Systems (ITS), Interactive Learning Systems (ILS) and Intelligent Learning Environments (ILE). The research in this field aims to improve interactivity between humans and non-human agents, assigning different roles to agents, to promote dialogue between different actors in learning environments, exploring possible ways of extracting information from the application environment, such as The abilities and needs of the students and, finally, promoting the interactivity of the environments. Challenges surrounding innovative teaching and learning environments imply two important aspects. First, these systems must acquire more autonomy, and show skillful behavior, which means they have to take roles dynamically and learn how to behave in a social environment. Second, they must have a comprehensive understanding of the students or teachers who use the system, which means having an adequate representation of their beliefs and goals. These aspects, together, place the research on Informatics and Education clearly in the field of agents and systems multiagents. Clarifying the relationship between multi-agent systems and educational applications involves the discussion around technical issues in specific languages and tools, techniques for modeling agents, formal languages and interaction models of agents suitable for this type of applications.
· (Short Course 1) Prof. Eder Mateus Nunes Gonçalves (FURG, Brasil)
Title: Petri Nets and Multiagent Systems
Abstract: Multiagent Systems (MAS) are seem as computational systems from agents interactions, once these agents act to reach their goals and present autonomy as a differential property. Goals are decomposed considering coordination of agents plans and actions, which can be seen as a discrete state space, oriented by agents actuation. This scenario can be depicted by Discrete Event System (DES), and we can apply tools from this kind of system to specify and validate MAS. The main SED tool applied in MAS is Petri Nets, with approaches used at organizational level, coordination, communication and to specify agents. There are works that apply classical models, as ordinary Petri Nets, Coloured Petri Nets and Object-oriented Petri Nets. Also, there are works that propose extensions to classical models to approach MAS specific issues. This presentation brings an introductory discussion about classical models of Petri Nets applied in MAS and also presents author's main contribution in this area, which include a methodology to specify organizational routines based on Coloured Petri Nets.
· (Short Course 2) Prof. Antonio Carlos da Rocha Costa (UFRGS / FURG, Brasil):
Title: SML - A Society Modeling Language
Abstract: The goal of this short course is to introduce the audience to the first draft of SML, a society modeling language. The general model of agent society underlying the design of the language will be reviewed. The type system, the syntax and the semantics of SML will be presented and discussed. Examples will be examined in details, in order to provide indications of the strength and the scope of applicability of the language. Simple case studies will be proposed to the audience, to allow for a concrete experience with the use of SML. Criticisms and suggestions that the audience be willing to provide will be considered, to be incorporated in the second draft of the language.