CompulogNet has set up a new area, called Logic-Based Agents. This area is concerned with the use of declarative and logic-based techniques for the specification, design and implementation of agents. It also encompasses development of applications using agent technologies based on such an approach. The area is coordinated by Fariba Sadri and Francesca Toni, both currently at Imperial College of Science, Technology and Medicine, University of London, UK. The formation of this new area within CompulogNet offers an opportunity for the dissemination of information about the field, and offers a forum for discussion and development of strategies for its encouragement and enhancement.
Agent technology is one of the most important and exciting areas of research and development in computer science today. Agents are computer systems that are capable of independent action in dynamic, unpredictable environments. The work on such systems integrates many technologies and concepts in artificial intelligence and other areas of computing.
Logic, in general, and computational logic, in particular, have important roles to play in the conceptualisation and development of agent-based systems. They can play a part in the representation of individual agents, for example:
- the formalization of the knowledge, beliefs and goals of the agents,
- the perception of the environment,
- the generation of timely and appropriate reactions to the environment,
- the construction of plans to solve goals,
- the reasoning component of agents.
Computational logic, in particular, provides a variety of proof procedures that can form the basis of reasoning engines of agents. Research in computational logic has been ongoing for a number of years, and has resulted in fairly mature viable proof procedures, incorporating features such as abduction, integrity checking, argumentation, induction and defeasible reasoning, which are proving useful for agents.
Moreover, logic and computational logic can play a part in the representation of multi-agent systems, i.e. communities of agents interacting with one another, for example for
- the generation of timely and appropriate reactions to other agents,
- the process of negotiation and communication amongst agents,
- the generation of joint plans.
As a sign of the area achieving rapid prominence one can point to the specialised conferences, workshops and journals that are dedicated to it. CompulogNet has organised a specialised Workshop in the subject. A brief report on the Workshop is given below.
Workshop on the Future of Logic-based Agents
In conjunction with AgentLink (http://www.agentlink.org/), Europe's ESPRIT-funded Network of Excellence for Agent-based Computing, CompulogNet sponsored a Workshop on the Future of Logic-based Agents http://www.cs.unitn.it/~pgiorgio/workshop/, on 8 March 1999, in the Department of Computing, of Imperial College, University of London, UK.
More than 70 delegates attended the workshop, which was a larger number than expected. The purpose of the workshop was to bring together researchers from the two communities to discuss a strategy for the future of logic-based agents and to identify short and long term research and development goals and possible applications to strengthen and future the contributions in this area.
The speakers, all of whom were invited, were:
- Juergen Dix and Thomas Eiter, The IMPACT System for Heterogeneous Agents
Abstract:The IMPACT system has been proposed as a platform for creating and deploying software agents in an open agent system. An agent is built on top of code which manipulates an arbitrary data structure through a well-defined interface, and a semantic wrapper is put around this code to turn it into an IMPACT agent. The agent's behaviour is specified through an Agent Program in a rule-based agent programming language. Different formal semantics for Agent Programs exist, which are in the spirit of related semantics in the domain of logic programming. We briefly present the base language of Agent Programs and discuss extensions to Meta Agent Programs (enabling agents to reason about beliefs) and Temporal Agent Programs (enabling agents to reason about time).
Principles and ideas from the logic programming domain have been fruitfully applied in this context. So far, applications from the logistics and flight control domains have been solved in the IMPACT framework. We feel that a contribution to the effort of building software agent systems whose agent programming language is declarative, has a rigid formal semantics, and is capable of dealing with a multiplicity of heterogeneous data structures. Further research and development issues that emerge from the current project state are also addressed.
- Klaus Fischer, Basic Concepts of a Theory for Hybrid Agent Architectures Abstract: The talk presented an overview of the basic concepts of a theory that has been developed in the Multi-Agent Systems Group at DFKI GmbH in Saarbrücken to investigate and specify the hybrid agent architecture InteRRaP. InteRRaP has been successfully applied to a number of application scenarios, e.g. transportaion, loading dock, and flexible manufacturing. Early work on InteRRaP concentrated on the architectural aspects. More recent research concentrates on the definition of a logical basis for a formal specification of the InteRRaP agent architecture.
- Fausto Giunchiglia, Model Checking Multiagent Systems
Abstract : Model checking is a very successful technique which has been applied in the design and verification of finite state concurrent reactive processes. The goal of this research is to lift model checking technique to be applicable to multiagent systems. Our approach allows us to reuse the technology and tools developed in model checking, to design and verify multiagent systems in a modular and incremental way, and also to have a very efficient model checking algorithm.
-Manuel Hermenegildo, Some Strengths and Weaknesses of (Constraint) Logic Programming Systems for Agent-Based Applications Abstract: We discuss what in our opinion are some of the strengths and weaknesses of current (Constraint) Logic Programming systems for developing Agent-Based Applications, with special emphasis on how some important related issues can be addressed. Issues to be discussed include resource awareness, improved modularity, handling state change, support for multi-threading and distribution, advanced interfaces to the external world, efficient support for higher-level languages, performance of meta-programming, etc. We will also give a brief overview of how some of these issues are addressed in the CIAO system.
- Maurizio Martelli, Combining logical agents with rapid prototyping for engineering distributed applications
Abstract: The realisation of new distributed and heterogeneous applications is a challenge that
Software Engineers have to face. LP and MAS can play a very effective role in the rapid
prototyping of new sw products. The talk will propose a general approach to the prototyping
of complex and distributed applications modelled as MAS. The approach has LP as foundation
and deals with different aspects of the problem:
integration of heterogeneous data and reasoning systems, animation of formal specifications and development of agent based software.
- Luis M. Pereira, Lisbon's views on the future of logic based agents
Abstract: We present our views, how they relate to ongoing work, and to an upcoming IST project proposal on Mindful Agents for Informed Choice -MAGIC. Our present MENTAL project aims at an intra and inter architecture for rational agents. The approach is a bottom-up one: we've been integrating into one framework, and implementation, features that we know how to do in LP such as KR, abduction, updates, contradiction removal, inductive LP, actions, planning. In the light of our experience we will comment on the future of LP.
- Fariba Sadri and Francesca Toni, Abductive logic programming for agent communication and protocols
Abstract: FIPA's latest call for proposal has identified the problem of giving a verifiable semantics to communication actions and protocols in agent communication languages as one of the open issues within the FIPA programme. In this talk we identify the possible role of (abductive) logic programming in giving a solution to this problem. We present an abductive logic programming framework in which (meta-)logic programs represent the beliefs of agents, integrity constraints represents prohibitions, obligations and commitment rules for agents, hypotheses (also called abducibles) represent observations and (physical and communicative) actions, and queries answered by an abductive proof procedure represent goals and bservations. The (interactive) execution of the abductive proof procedure is responsible for the behaviour of agents. Within this framework, preconditions and effects of communication actions as well as communication protocols can be directly represented as logic programs and integrity constraints, and the compliance of agents' behaviour to the semantics for communication is a by-product of the soundness of the abductive proof procedure.
- Carles Sierra, Multicontext argumentative agents
Abstract: Argumentation is found to be a useful mechanism to coordinate agents that have potentially conflicting goals. We propose an agent architecture based on multicontext systems (local theories interacting by means of bridge rules) capable of distributed planning. The agent architecture is based on a set of units (local theories) interconnected bymeans of bridge rules that are grouped on 'modules' that are responsible to solve concrete tasks. These modules in turn interact to cooperatively solvepotential conflicting opinions about how to solve a problem. This cooperation can be modelled as a set of bridge rules interconnecting the 'communication' units of the different modules. In a similar way the communication between the agents can be seen as bridge rules taking care of the ontological aspects of the dialogue. The architecture is exemplified by means of two home-improving agents that share resources to satisfy the goals of both agents.
- Michael Wooldridge, Logic and Agent-based Software Engineering
Abstract: The purpose of this presentation is to consider the problem of engineering a multi-agent
system from the formal methods perspective. The presentation focuses on three issues:
(i) how agents might be specified
(ii) how these specifications might be refined or otherwise transformed into efficient implementations and
(iii) how implemented agents and multi-agent systems might subsequently be verified, in order to show that they are correct with respect to their specifications. These issues are discussed with reference to a number of case-studies, and in particular to work on formal methods in mainstream computer science. The presentation concludes by setting out some issues and open problems for future research.
The workshop also included a Panel on The future of Logic-based agents - Research and applications.
The Panel Moderator was Dov Gabbay, and the Panel members were Keith Clark, Michael Fisher, Bob Kowalski, John-Jules Meyer,Murray Shanahan.
The area co-ordinators are co-organising (with Stephen Rochefort of Simon Fraser University, Canada) a Workshop on Multi-Agent Systems in Logic Programming, to be held in conjunction with the International Conference on Logic Programming, November 29 - December 5, 1999, in Las Cruces, New Mexico (USA).
The aim of this workshop is to explore the application of Logic Programming to Multi-Agent Systems. The focus of the workshop will be on multi-agent rather than single agent issues, with emphasis on co-operative and conflicting agents, where there could be common and joint goals as well as goals for individual agents. In such contexts, reasoning about other agents' goals and beliefs, communication, negotiation, belief revision and conflict resolution could be essential for the achievement of the goals of the agents.
Details of the area co-ordinators:
Department of Computing
Imperial College of Science and Technology and Medicine