The paper introduces an AND/OR search space perspective for graphical models that include probabilistic networks (directed or undirected) and constraint networks. In contrast to the traditional (OR) search space view, the AND/OR search tree displays some of the independencies present in the graphical model explicitly and may sometime reduce the search space exponentially. Indeed, most

In the efficient social choice problem, the goal is to assign values, subject to side constraints, to a set of variables to maximize the total utility across a population of agents, where each agent has private information about its utility function. In this paper we model the social choice problem as a distributed constraint optimization problem (DCOP), in which each agent can communicate with other agents that share an interest in one or more variables. Whereas existing DCOP algorithms can be easily manipulated by an agent, either by misreporting private information or deviating from the algorithm, we introduce M-DPOP, the first DCOP algorithm that provides a faithful distributed implementation for efficient social choice. This provides a concrete example of how the methods of mechanism design can be unified with those of distributed optimization. Faithfulness ensures that no agent can benefit by unilaterally deviating from any aspect of the protocol, neither informationrevelation, computation, nor communication, and whatever the private information of other agents. We allow for payments by agents to a central bank, which is the only central authority that we require. To achieve faithfulness, we carefully integrate the Vickrey-Clarke-Groves (VCG) mechanism with the DPOP algorithm, such that each agent is only asked to perform computation, report

We examine whether and how the Multiagent Plan Coordination Problem, the problem of resolving interactions between the plans of multiple agents, can be cast as a Distributed Constraint Optimization Problem (DCOP). We use ADOPT, a state-of-the-art DCOP solver that can solve DCOPs in an asynchronous, parallel manner using local communication between individual computational agents. We then demonstrate how we can take advantage of novel flaw-assignment strategies and plan coordination algorithms to significantly improve the performance of ADOPT on representative coordination problems. We close with a consideration of possible advances in framing our DCOP representation of the Multiagent

This paper considers the problem of performing decentralised coordination of low-power embedded devices (as is required within many environmental sensing and surveillance applications). Specifically, we address the generic problem of maximising social welfare within a group of interacting agents. We propose a novel representation of the problem, as a cyclic bipartite factor graph, composed of variable and function nodes (representing the agents’ states and utilities respectively). We show that such representation allows us to use an extension of the max-sum algorithm to generate approximate solutions to this global optimisation problem through local decentralised message passing. We empirically evaluate this approach on a canonical coordination problem (graph colouring), and benchmark it against state of the art approximate and complete algorithms (DSA and DPOP). We show that our approach is robust to lossy communication, that it generates solutions closer to those of DPOP than DSA is able to, and that it does so with a communication cost (in terms of total messages size) that scales very well with the number of agents in the system (compared to the exponential increase of DPOP). Finally, we describe a hardware implementation of our algorithm operating on low-power Chipcon CC2431 System-on-Chip sensor nodes.

We introduce the Multiagent Agreement Problem (MAP) to represent a class of multiagent scheduling problems. MAP is based on the Distributed Constraint Reasoning (DCR) paradigm and requires agents to choose values for variables to satisfy not only their own constraints, but also equality constraints with other agents. The goal is to represent problems in which agents must agree on scheduling decisions, for example, to agree on the start time of a meeting. We investigate a challenging class of MAP -- private, incremental MAP (piMAP) in which agents do incremental scheduling of activities and there exist privacy restrictions on information exchange. We investigate a range of strategies for piMAP, called "bumping" strategies. We empirically evaluate these strategies in the domain of calendar management where a personal assistant agent must schedule meetings on behalf of its human user. Our results show that bumping decisions based on scheduling difficulty models of other agents can significantly improve performance over simpler bumping strategies.

In this article we describe a set of scalable techniques for learning the behavior of a group of agents in a collaborative multiagent setting. As a basis we use the framework of coordination graphs of Guestrin, Koller, and Parr (2002a) which exploits the dependencies between agents to decompose the global payoff function into a sum of local terms. First, we deal with the single-state case and describe a payoff propagation algorithm that computes the individual actions that approximately maximize the global payoff function. The method can be viewed as the decision-making analogue of belief propagation in Bayesian networks. Second, we focus on learning the behavior of the agents in sequential decision-making tasks. We introduce different model-free reinforcementlearning techniques, unitedly called Sparse Cooperative Q-learning, which approximate the global action-value function based on the topology of a coordination graph, and perform updates using the contribution of the individual agents to the maximal global action value. The combined use of an edge-based decomposition of the action-value function and the payoff propagation algorithm for efficient action selection, result in an approach that scales only linearly in the problem size. We provide experimental evidence that our method outperforms related multiagent reinforcement-learning methods based on temporal differences.

www.elsevier.com/locate/artint We study Coalitional Resource Games (CRGs), a variation of Qualitative Coalitional Games (QCGs) in which each agent is endowed with a set of resources, and the ability of a coalition to bring about a set of goals depends on whether they are collectively endowed with the necessary resources. We investigate and classify the computational complexity of a number of natural decision problems for CRGs, over and above those previously investigated for QCGs in general. For example, we show that the complexity of determining whether conflict is inevitable between two coalitions with respect to some stated resource bound (i.e., a limit value for every resource) is co-NP-complete. We then investigate the relationship between CRGs and QCGs, and in particular the extent to which it is possible to translate between the two models. We first characterise the complexity of determining equivalence between CRGs and QCGs. We then show that it is always possible to translate any given CRG into a succinct equivalent QCG, and that it is not always possible to translate a QCG into an equivalent CRG; we establish some necessary and some sufficient conditions for a translation from QCGs to CRGs to be possible, and show that even where an equivalent CRG exists, it may have size exponential in the number of goals and agents of its source QCG.

A distributed constraint optimization problem (DCOP) is a formalism that captures the rewards and costs of local interactions within a team of agents. Because complete algorithms to solve DCOPs are unsuitable for some dynamic or anytime domains, researchers have explored incomplete DCOP algorithms that result in locally optimal solutions. One type of categorization of such algorithms, and the solutions they produce, is k-optimality; a k-optimal solution is one that cannot be improved by any deviation by k or fewer agents. This paper presents the first known guarantees on solution quality for k-optimal solutions. The guarantees are independent of the costs and rewards in the DCOP, and once computed can be used for any DCOP of a given constraint graph structure. 1

A new search algorithm for solving distributed constraint optimization problems (DisCOPs) is presented. Agents assign variables sequentially and propagate their assignments asynchronously. The asynchronous forward-bounding algorithm (AFB) is a distributed optimization search algorithm that keeps one consistent partial assignment at all times. Forward bounding propagates the bounds on the cost of solutions by sending copies of the partial assignment to all unassigned agents concurrently. The algorithm is described in detail and its correctness proven. Experimental evaluation of AFB on random Max-DisCSPs reveals a phase transition as the tightness of the problem increases. This effect is analogous to the phase transition of Max-CSP when local consistency maintenance is applied [3]. AFB outperforms Synchronous Branch & Bound (SBB) as well as the asynchronous state-of-the-art ADOPT algorithm, for the harder problem instances. Both asynchronous algorithms outperform SBB by a large factor. 1

Abstract.Distributed Constraint Optimization (DCOP) is rapidly emerging as a prominent technique for multiagent coordination. Unfortunately, rigorous quantitative evaluations of privacy loss in DCOP algorithms have been lacking despite the fact that agent privacy is a key motivation for applying DCOPs in many applications. Recently, Maheswaran et al. [1, 2] introduced a framework for quantitative evaluations of privacy in DCOP algorithms, showing that early DCOP algorithms lose more privacy than purely centralized approaches and questioning the motivation for applying DCOPs. Do state-of-the art DCOP algorithms suffer from a similar shortcoming? This paper answers that question by investigating several of the most efficient DCOP algorithms, including both DPOP and ADOPT. Furthermore, while previous work investigated the impact on efficiency of distributed contraint reasoning design decisions, e.g. constraint-graph topology, asynchrony, message-contents, this paper examines the privacy aspect of such decisions, providing an improved understanding of privacy-efficiency tradeoffs. Finally, this paper augments previous work on system-wide privacy loss, by investigating inequities in individual agents ’ privacy loss. 1