Yannis Dimopoulos

Deterministic planning in the fifth international planning competition: PDDL3 and experimental evaluation of the planners

The international planning competition (IPC) is an important driver for planning research. The general goals of the IPC include pushing the state of the art in planning technology by posing new scientific challenges, encouraging direct comparison of planning systems and techniques, developing and improving a common planning domain definition language, and designing new planning domains and problems for the research community. This paper focuses on the deterministic part of the fifth international planning competition (IPC5), presenting the language and benchmark domains that we developed for the competition, as well as a detailed experimental evaluation of the deterministic planners that entered IPC5, which helps to understand the state of the art in the field. We present an extension of pddl, called pddl3, allowing the user to express strong and soft constraints about the structure of the desired plans, as well as strong and soft problem goals. We discuss the expressive power of the new language focusing on the restricted version that was used in IPC5, for which we give some basic results about its compilability into pddl2. Moreover, we study the relative performance of the IPC5 planners in terms of solved problems, CPU time, and plan quality; we analyse their behaviour with respect to the winners of the previous competition; and we evaluate them in terms of their capability of dealing with soft goals and constraints, and of finding good quality plans in general. Overall, the results indicate significant progress in the field, but they also reveal that some important issues remain open and require further research, such as dealing with strong constraints and computing high quality plans in metric-time domains and domains involving soft goals or constraints.

Ignoring Irrelevant Facts and Operators in Plan Generation

It is traditional wisdom that one should start from the goals when generating a plan in order to focus the plan generation process on potentially relevant actions. The GRAPHPLAN system, however, which is the most efficient planning system nowadays, builds a “planning graph” in a forward-chaining manner. Although this strategy seems to work well, it may possibly lead to problems if the planning task description contains irrelevant information. Although some irrelevant information can be filtered out by GRAPHPLAN, most cases of irrelevance are not noticed.

A unified and general framework for argumentation-based negotiation

This paper proposes a unified and general framework for argumentation-based negotiation, in which the role of argumentation is formally analyzed. The framework makes it possible to study the outcomes of an argumentation-based negotiation. It shows what an agreement is, how it is related to the theories of the agents, when it is possible, and how this can be attained by the negotiating agents in this case. It defines also the notion of concession, and shows in which situation an agent will make one, as well as how it influences the evolution of the dialogue.

On the computational complexity of assumption-based argumentation for default reasoning

Bondarenko et al. have recently proposed an abstract framework for default reasoning. Besides capturing most existing formalisms and proving that their standard semantics all coincide, the framework extends these formalisms by generalising the semantics of admissible and preferred arguments, originally proposed for logic programming only.In this paper we analyse the computational complexity of credulous and sceptical reasoning under the semantics of admissible and preferred arguments for (the propositional variant of) the instances of the abstract framework capturing theorist, circumscription, logic programming, default logic, and autoepistemic logic. Although the new semantics have been tacitly assumed to mitigate the computational hardness of default reasoning under the standard semantics of stable extensions, we show that in many cases reasoning under the admissibility and preferability semantics is computationally harder than under the standard semantics. In particular, in the case of autoepistemic logic, sceptical reasoning under preferred arguments is located at the fourth level of the polynomial hierarchy, whereas the same form of reasoning under stable extensions is located at the second level.

Extended Semantics and Optimization Algorithms for CP-Networks

Preference elicitation is a serious bottleneck in many decision support applications and agent specification tasks. Ceteris paribus (CP)‐nets were designed to make the process of preference elicitation simpler and more intuitive for lay users by graphically structuring a set of CP preference statements—preference statements that most people find natural and intuitive. Beside their usefulness in the process of preference elicitation, CP‐nets support efficient optimization algorithms that are crucial in most applications (e.g., the selection of the best action to execute or the best product configuration). In various contexts, CP‐nets with an underlying cyclic structure emerge naturally. Often, they are inconsistent according to the current semantics, and the user is required to revise them. In this paper, we show how optimization queries can be meaningfully answered in many “inconsistent” networks without troubling the user with requests for revisions. In addition, we describe a method for focusing the users revision process when revisions are truly needed. In the process, we provide a formal semantics that justifies our approach and new techniques for computing optimal outcomes. Some of the methods we use are based on a reduction to the problem of computing stable models for nonmonotonic logic programs, and we explore this relationship closely.

Learning Non-Monotonic Logic Programs: Learning Exceptions

In this paper we present a framework for learning non-monotonic logic programs. The method is parametric on a classical learning algorithm whose generated rules are to be understood as default rules. This means that these rules must be tolerant to the negative information by allowing for the possibility of exceptions. The same classical algorithm is then used to learn recursively these exceptions. We prove that the non-monotonic learning algorithm that realizes these ideas converges asymptotically to the concept to be learned. We also discuss various general issues concerning the problem of learning nonmonotonic theories in the proposed framework.

Multi-Agent Coordination and Cooperation through Classical Planning

Multi-agent planning is a fundamental problem in multi-agent systems that has acquired a variety of meanings in the relative literature. In this paper we focus on a setting where multiple agents with complementary capabilities cooperate in order to generate non-conflicting plans that achieve their respective goals. We study two situations. In the first, the agents are able to achieve their subgoals by themselves, but they need to find a coordinated course of action that avoids harmful interactions. In the second situation, some agents may ask the assistance of others in order to achieve their goals. We formalize the two problems and present algorithms for their solution. These algorithms are based on an underlying classical planner which is used by the agents to generate their individual plans, but also to find plans that are consistent with those of the other agents. The procedures generate optimal plans under the plan length criterion. The central role that has been given to the classical planning algorithm, can be seen as an attempt to establish a stronger link between classical and multi-agent planning.

Reasoning about actions and change in Answer Set Programming

This paper studies computational issues related to the problem of reasoning about actions and change (RAC) by exploiting its link with the Answer Set Programming paradigm. It investigates how increasing the expressiveness of a RAC formalism so that it can capture the three major problems of frame, ramification and qualification, affects its computational complexity, and how a solution to these problems can be implemented within Answer Set Programming. Our study is carried out within the particular language e. It establishes a link between e and Answer Set Programming by presenting encodings of different versions of this language into logic programs under the answer set semantics. This provides a computational realization of solutions to problems related to reasoning about actions and change, that can make use of the recent development of effective systems for Answer Set Programming.

Extending Argumentation to Make Good Decisions

Argumentation has been acknowledged as a powerful mechanism for automated decision making. In this context several recent works have studied the problem of accommodating preference information in argumentation. The majority of these studies rely on Dungs abstract argumentation framework and its underlying acceptability semantics. In this paper we show that Dungs acceptability semantics, when applied to a preference-based argumentation framework for decision making purposes, may lead to counter intuitive results, as it does not take appropriately into account the preference information. To remedy this we propose a new acceptability semantics, called super-stable extension semantics, and present some of its properties. Moreover, we show that argumentation can be understood as a multiple criteria decision problem, making in this way results from decision theory applicable to argumentation.

A general framework for argumentation-based negotiation

This paper proposes a unified and general framework for argumentation-based negotiation, in which the role of argumentation is formally analyzed. The framework makes it possible to study the outcomes of an argumentation-based negotiation. It shows what an agreement is, how it is related to the theories of the agents, when it is possible, and how this can be attained by the negotiating agents in this case. It defines also the notion of concession, and shows in which situation an agent will make one, as well as how it influences the evolution of the dialogue.