Abstract: Multi-agent planning comprises planning in an environment with multiple autonomous actors. Techniques for multi-agent planning differ from conventional planning in that planning activities are distributed and the planning autonomy of the agents must be respected. We focus on approaches to coordinate the multi-agent planning process. While usually coordination is intertwined with the planning process, we distinguish a number of separate phases in the planning process to get a clear view on the different role(s) of coordination. In particular, we discuss the pre-planning coordination phase and post-planning coordination phase. In the pre-planning part, we view coordination as the process of managing (sub) task dependencies and we discuss a method that ensures complete planning autonomy by introducing additional (intra-agent) dependencies. In the post-planning part, we will show how agents can improve their This chapter appears in the book, Intellige t Techniques for Planning, edited by Ioamis Vla avas and Dimitris Vrakas. Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. 701 E. Chocolate Avenue, Suite 200, Hershey PA 17033-1240, USA Tel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com IDEA GROUP PUBLISHING Coordination in Multi-Agent Planning 195 Copyright © 2005, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. plans through the exchange of resources. We present a plan merging algorithm that uses these resources to reduce the costs of independently developed plans. This (anytime) algorithm runs in polynomial time. INTRODUCTION Often the actions or elementary tasks in a plan have to be performed by different actors. Especially if these actors have a common interest, the plan itself is usually constructed by a single actor. Examples are production planning in factories, arrival and departure planning on airports, planning for building projects, and the planning of armed forces. If, however, the actors involved require some degree of (planning) autonomy themselves, centralized construction of the plan may be not feasible. Here, “autonomy” refers to the ability to make decisions in an individually rational fashion, such as when to perform which action. Such autonomous actors are called agents, and such planning problems are called multi-agent planning problems. Reading this book, one may wonder why we need to study such multi-agent planning problems and techniques as a separate topic. Isn’t it true that such problems are already dealt with in general discussions of planning? The answer to this question comes in two parts. On the one hand, in real-life problems, we deal with multiple agents having their own goals, and it is often impractical or undesirable to create the plan for all agents centrally. These agents may be people or companies simply demanding to plan their actions themselves, or refusing to make all information necessary for planning available to someone else. Furthermore, the planning problem itself may be simply too complex to be solved by one agent, while planning the parts for each agent individually may be feasible. On the other hand, when agents acting in the same environment create their plans individually, they still need to coordinate their actions for a number of reasons. First of all, coordination is needed to prevent chaos (e.g., collisions, deadlock), which may easily arise if each agent just acts on itself. Secondly, coordination may be required because the agents need to meet global constraints, or because there are dependencies between the actions of the different agents. And even when the agents can function completely independently, coordination may help to improve the efficiency of their plans. Summarizing, in quite a few real-life problems there is a clear need to have each agent construct its plan more or less independently, but there is also a need to coordinate these plans. A planning problem that has these key properties is called a multi-agent planning problem. In this chapter, we first present a more precise definition of this multi-agent planning problem. Next, we give an overview of issues that arise when trying to solve such a problem. Based on this overview we then present a classification of existing research within multi-agent planning, paying special attention to the role of coordination in the planning process. In the third and fourth section we discuss two different techniques to coordinate independently planning agents. Firstly, a distributed algorithm is developed that derives a (minimal) set of restrictions on the agents’ plans that can be given to them before they start planning. These restrictions ensure that their plans do not interfere. Secondly, we address a technique to improve the efficiency of the plans after they have been created individually. These two methods are discussed separately and we illustrate 29 more pages are available in the full version of this document, which may be purchased using the "Add to Cart" button on the product's webpage: www.igi-global.com/chapter/coordination-multi-agentplanning-application/24463?camid=4v1 This title is available in InfoSci-Books, InfoSci-Intelligent Technologies, Science, Engineering, and Information Technology, InfoSci-Computer Science and Information Technology, InfoSci-Select, InfoSci-Select. Recommend this product to your librarian: www.igi-global.com/e-resources/libraryrecommendation/?id=1

Abstract: The two models presented in this paper are suitable for multi-agent planning. The Recursive Petri nets allow the plans modeling (both at the agent and multi-agents levels) and their management when abstraction and dynamic refinement are required. RPN allows. easily. the synchronization of individual agents’ plans. They are, in particular, Interesting for the multi-agent validation thanks to the reachability tree building if combined to reduction technics (in order to avoid the combinatory explosion of the the number of states). The main shortcomoing of this model is the absence of explicit handling of temporal constraints. This is why we developed a model based on Hybrid Automata that model different clocks evolving with different speeds. These Clocks may be the resources of each agent and the time.

Abstract: This paper proposes a multi-agent planning framework. It presents an integral cycle we developed to build feasible multi-agent plans in a dynamic context (i.e. aircraft simulation). This cycle takes into account both functional and computational features of our multi-agent planning framework. From modeling to validation, this cycle allows the management of different constraints (e.g. time, physical resources, etc.) at the single- and multi-agent levels

Abstract: Many day-to-day situations involve decision making: for example, a taxi company has some transportation tasks to be carried out, a large firm has to distribute a lot of complicated tasks among its subdivisions or subcontractors, and an air-traffic controller has to assign time slots to planes that are landing or taking off. Intelligent agents can aid in this decision-making process. Agents are often classified into two categories according to the techniques they employ in their decision making: reactive agents (cf. (Ferber and Drogoul, 1992)) base their next decision solely on their current sensory input; planning agents, on the other hand, take into account anticipated future developments — for instance as a result of their own actions — to decide on the most favourable course of action. When an agent should plan and when it should be reactive depends on the particular situation it finds itself in. Consider the example where an agent has to plan a route from one place to another. A reactive agent might use a compass to plot its course, whereas a planning agent would consult a map. Clearly, the planning agent will come up with the shortest route in most cases, as it won’t be confounded by uncrossable rivers, one-way streets, and labyrinthine city layouts. On the other hand, there are also situations where a reactive agent can at least be equally effective, for instance if there are no maps to consult, for instance in a domain of (Mars) exploration rovers. Nevertheless, the ability to plan ahead is invaluable in many domains, so in this paper we will focus on planning agents. The general structure of a planning problem is easy to explain: (the relevant part of) the world is in a certain state, but managers or directors would like it to be in another state. The (abstract) problem of how one should get from the current state of the world through a sequence of actions to the desired goal state is a planning problem. Ideally, to solve such planning problems, we would like to have a general planning-problem solver. However, such an algorithm solving all planning problems can be proven to be non-existing.1 We therefore start to concentrate on a simplification of the general planning problem called ‘the classical planning problem’. Although not all realistic problems can be modeled as a classical planning problem, they can help to solve more