Two tutorials will take place in PRIMA 2019:
- Multi-Agent Oriented Programming to engineer complex intelligent systems
- Automated Negotiation: Challenges and Tools
Multi-Agent Oriented Programming to engineer complex intelligent systems (T1)
October 29th | 9:30-11:00 | Sala Principe d'Acaja
The aim of this tutorial is to teach how to use the basic aspects of Multi-Agent Oriented programming (MAOP) in order to engineer complex intelligent systems. With the increasing embedding of AI techniques in Cyber-Physical Systems, applications have to tackle with large and open eco-systems of interacting autonomous entities. Engineering such complex systems in such a context is a challenging and important issue. Using an holistic/systemic approach, focusing on design and programming dimensions of MAS, i.e. agent, environment, interaction and organisation, we bring to the students the right abstractions and approaches for tackling this problem.
After an introduction presenting the foundational concepts of each programming dimension and their current projections in existing practical and usable development platforms in the multi-agent domain, we use an incremental and pedagogical approach for teaching how MAOP helps to engineer complex systems. The course will discuss and demonstrate how the use of one or several programming dimensions may ease to ensure flexibility, reusability, robustness, etc. The examples will be based on our experience in the development of applications with the JaCaMo platform. However, the course will abstract from this practical platform to introduce the relevant techniques for students so that they acquire hands-on experience of fully-fledged multi-agent oriented programming of complex systems, applicable on other platforms.
The intended audience of this course is composed of researchers working in AI wanting to complement their knowledge of the state of the art on using multi-agent oriented programming concepts and techniques for developing decentralized AI applications in the current context of open and large scale applications. We propose a tutorial for students with some familiarity with multi-agent systems, too.
Introduction In the introduction, we present a comprehensive view of Multi-Agent Oriented Programming models issued from existing platforms, languages and applications of multi-agent systems. We intend to give some historical background to the programming of MAS stressing the difference of multi-agent oriented programming (MAOP) with previous proposals of agent oriented programming (Shoham), interaction oriented programming (Singh) and team oriented programming (Tambe). We also present the integrated scenario that will be used as a case study for the rest of the tutorial.
MAOP In this part, we present a comprehensive view of Multi-Agent Oriented Programming models issued from existing platforms, languages and applications of multi-agent systems. In specific, we will discuss the different abstraction and first class entities that can be used as candidates for programming multi-agent systems. The case study will be used to illustrate the use of these different abstractions. To make examples more concrete, we will use the JaCaMo platform as an illustrative implementation.
MAOP for engineering complex systems In this part, we present different refactoring and enrichment of the case study to demonstrate how MAOP helps to ensure important features to engineer complex systems such as: extensibility, reusability, openness, robustness, flexibility and adaptability.
Conclusion Finally, we discuss the state of the art in MAOP and the challenges for future research in relation with mainstream computer science.
Rafael H. Bordini is an Associate Professor at the School of Technology of PUCRS, located in Porto Alegre, Rio Grande do Sul, Brazil. His main area of research in the last few years has been on agentoriented programming, in particular proposing various extensions to a BDI logic-based programming language called AgentSpeak, and also on model checking for that language. He is one of the developers of Jason, an interpreter for the extended version of AgentSpeak. Rafael Bordini is emeritus member of the boards of directors of IFAAMAS ad EURAMAS. He was Programme Cochair for AAMS-2015 and is in the Steering Committee of EMAS, a workshop dedicated to Agent- Oriented Programming Languages and Software Engineering, as well as having been SPC or PC member of many other events such as AAMAS, IJCAI, ECAI, AAAI, PRIMA, JELIA, CLIMA. He has published over 150 papers, co-edited two books on Programming Multi-Agent Systems published by Springer and authored a book about the AgentSpeak programming language and the Jason platform. Rafael Bordini has given tutorials on Agent Programming Languages both at AAMAS and EASSS, and he has also taught under- and post-graduate courses on MAS including agent programming languages.
O. Boissier is full professor of computer science at Mines Saint-Etienne (France). He is member of the Institut Henri Fayol at Mines Saint-Etienne and of Hubert Curien Laboratory UMR CNRS 5516. He is active in the research and development of multi-agent systems. His main research contributions concern: coordination and control of multi-agent systems, multi-agent oriented programming. MSc and PhD theses that he has advised have contributed to the development of MOISE organizational models, agents' architectures, autonomy and control, JaCaMo platform. They all contribute to the definition of multi-agent models, tools and methodologies for the multi-agent oriented engineering of software applications in the smart industry and smart city domains. Olivier Boissier is serving on the program committees of many conferences (IAT/WI, AAMAS) and workshops. He was the program chair of the IAT/WI conference in 2011. O. Boissier has presented tutorials on Organization Oriented Programming at AAMAS 2008 and 2004, EASSS 2007, 2005 on Multi-Agent Oriented Programming at AAMAS 2015, EASSS 2012, 2011, 2010.
Jomi F. Hübner is an Associate Professor at the University of Santa Catarina (Brazil). His main research interests are multi-agent programming and tools to develop such systems. The main projects he is co-developing are Moise (an organisational model and infrastructure), Jason (an interpreter of a language for belief-desire-intention agents), and JaCaMo (a multidimensional programming platform for multiagent systems). He co-organised theIberagents@IBERAMIA 2006, COIN 2008@AAMAS 08, LADS 2010@MALLOW 2010, ProMAS@AAMAS 2012), and was the Workshops co-chair at IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT'11). He has served in the programme committees of major conferences and workshops in multi-agent systems and more generally in artificial intelligence: AAMAS, EUMAS, ProMAS, COIN, AT2AI, LADS, SBIA, RFIA, APLSA, SEAS.
A. Ricci is an associate professor at DISI (Computer Science and Engineering Department), Alma Mater Studiorum - Università di Bologna. His main research activity concerns multi-agent oriented programming, coordination in Multi-Agent Systems and agent-oriented computing – investigating agents and MAS as a paradigm for general-purpose computer programming and software engineering. A. Ricci is author of several research papers published in conferences and journals concerning agents and multi-agent systems, AAMAS conference and JAAMAS journal in particular, and member of the program committee of international conferences and workshops. He has been coorganiser of several international events, including AGERE! workshop inside the SPLASH conference, the ProMAS (“Programming Multi-Agent Systems”) workshop at AAMAS and ESAW workshop (“Engineering Societies in Agent Worlds”). He has been co-author of seminars and tutorials about Multi-Agent Programming already, in the context of AAMAS and EASSS (European Agent Systems Summer School) . He has served in the programme committees of major conferences and workshops in multi-agent systems and more generally in artificial intelligence: AAMAS, EUMAS, ProMAS, COIN, AT2AI, LADS.
Automated Negotiation: Challenges and Tools (T2)
October 29th | 11:30-13:00 | Sala Principe d'Acaja
Automated negotiation between intelligent agents is attracting more attention from the research community especially with the wider market penetration of intelligent agents and the need to coordinate their behavior. Research in automated negotiations has a long and interesting history extending at least to the eighties. This tutorial will introduce the audience to the field in general providing an overview of established results and focusing on open challenges. Moreover, the tutorial will introduce the audience to two research platforms that are being used for facing these challenges (Genius and NegMAS).
Automated negotiation between intelligent agents is an important research eld in MAS. It is going to become even more important with the wider spread of intelligent agents. It provides both a challenging problem in the principles of multi-agent systems and a potential field of practical applications of multi-agent systems. Both of these are of great interest to PRIMA audience. Last year, one of the three workshops accepted for PRIMA 2018 was about automated negotiations (PRIANAC) exemplifying the growing interest in this area.
The target audience are postgraduate students and researchers in the elds of multi-agent systems, game theory, simulation, and practical applications of MAS. The tutorial will introduce the concepts it needs and is a beginner-level tutorial so the prerequisites are minimal. Knowledge of basic game-theoretic concepts like Pareto-optimality, the concept of an equilibrium, and basics of auction theory are advantageous but not required.
The tutorial will consist of two main parts with a 10-min break.
Theoretical Session (75 min) This part of the tutorial introduces the field of automated negotiation providing both classical results and recent advances:
- The negotiation problem Different definitions of the negotiation problem, negotiation protocol, main differences between negotiations and auctions.
- Utiltity Functions Gives a well-established taxonomy of utility functions for single issue and multi-issue negotiations as well as the different types of relationships between the negotiators' utility functions.
- Classical results Introduces the game-theoretic analysis of negotiations and provides some of the classical results including the prefect equilibrium result of Rubinstein.
- Negotiation Analysis Introduces the main concepts for analyzing negotiation scenarios including the Pareto-front, welfare, fairness, the Nash-point.
- Negotiation Protocols Introduces the most widely used mediated and unmediated negotiation protocols with a focus on the stacked alternating offers protocol.
- Negotiation Strategies Provides the basic structure of a negotiating agent and the simplest strategies (they are in the same time the most widely used). These include time-based aspiration based strategies, tit-for-tat and nice tit-for-tat.
Challenges for Automatic Negotiation Describes the major challenges facing real-world negotiation
agents that provide fruitful directions of research:
- Utility Elicitation During Negotiation Denes the problem of negotiation under uncertainty and provides the motivation for research in utility elicitation during negotiation. Three well-established main approaches to utility elicitation during negotiation will also be discussed. All of these methods were proposed in the past 3 years.
- Situated and Simulataneous Negotiations This part of the tutorial deals with an open area of research in automated negotiation and is less results-focused than the rest of the tutorial. We will introduce the concept of situated and simultaneous negotiations and provide some of the early results (negotiation with outside options, 2006).
Demo Session (30 min) This part of the tutorial will provide a hands-on demo of using two negotiation platforms for tackling the problems introduced in the theoretical session.
- Classical Negotiation Platform (Genius) Describes the de-facto standard platform for negotia- tion research since 2010. The main structure and design objective of Genius are introduced and a demo negotiation session between two of the previously described strategies (time-based aspiration and tit-for-tat) will be conducted using the Genius GUI and the resulting negotiation session will be analyzed using the tools provided by Genius in light of the earlier discussion of negotiation analysis.
- Situated Negotiations Platform (NegMAS) Shows the design and use-cases of the younger NegMAS platform including its tools for utility elicitation and simultaneous negotiations. An interactive python demo will be used to highlight the most important aspects of the platform.
Conclusions (5 min) The tutorial will be wrapped-up by a summary of the information introduced in the first session about automatic negotiation and will provide interesting directions of research inviting the audience to actively participating in pushing forward this exciting domain.
Yasser F. O. Mohammad is a Researcher at AIST, Japan and an Associate Professor at Assiut University, Egypt. He received his PhD from Kyoto University, Japan in 2009 in the area of intelligence science and technology. Recipient of four best paper awards from ICCAS 2012, IEEE/SICE SII 2011, IEA/AIE 2009, and IEA/AIE 2009. Author of Conversational Informatics: A data intensive approach (Springer, 2015) and Data mining for Social Robotics (Springer, 2016). His research focuses on automatic negotiation, multi-agent systems, and applications of data mining techniques to time-series data. Published over 90 international publications in these areas with an h-index of 15. Co founder of Ayonix Inc., Japan and Tebex IT, Egypt and a senior IEEE member.