Abstract not found

We present a new polynomial-space algorithm, called Adopt, for distributed constraint optimization (DCOP). DCOP is able to model a large class of collaboration problems in multi-agent systems where a solution within given quality parameters must be found. Existing methods for DCOP are not able to provide theoretical guarantees on global solution quality while operating both efficiently and asynchronously. Adopt is guaranteed to find an optimal solution, or a solution within a user-specified distance from the optimal, while allowing agents to execute asynchronously and in parallel. Adopt obtains these properties via a distributed search algorithm with several novel characteristics including the ability for each agent to make local decisions based on currently available information and without necessarily having global certainty. Theoretical analysis shows that Adopt provides provable quality guarantees, while experimental results show that Adopt is significanfly more efficient than synchronous methods. The speedups are shown to be partly due to the novel search strategy employed and partly due to the asynchrony of the algorithm.

Hospital Patient Scheduling is an inherently distributed problem because of the way real hospitals are organized. As medical procedures have become more complex, and their associated tests and treatments have become interrelated, the current ad hoc patient scheduling solutions have been observed to break down [15]. We propose a multi-agent solution using the Generalized Partial Global Planning (GPGP) approach that preserves the existing human organization and authority structures, while providing better system-level performance (increased hospital unit throughput and decreased patient stay time). To do this, we extend GPGP with a new coordination mechanism to handle mutually exclusive resource relationships. Like the other GPGP mechanisms, the new mechanism can be applied to any problem with the appropriate resource relationship. We evaluate the this new mechanism in the context of the hospital patient scheduling problem, and examine the effect of increasing interrelations between task...

Formal treatment of collaborative multi-agent systems has been lagging behind the rapid progress in sequential decision making by individual agents. Recent work in the area of decentralized Markov Decision Processes (MDPs) has contributed to closing this gap, but the computational complexity of these models remains a serious obstacle. To overcome this complexity barrier, we identify a specific class of decentralized MDPs in which the agents ’ transitions are independent. The class consists of independent collaborating agents that are tied together through a structured global reward function that depends on all of their histories of states and actions. We present a novel algorithm for solving this class of problems and examine its properties, both as an optimal algorithm and as an anytime algorithm. To the best of our knowledge, this is the first algorithm to optimally solve a non-trivial subclass of decentralized MDPs. It lays the foundation for further work in this area on both exact and approximate algorithms. 1.

We consider an environment where agents' tasks are tightly coupled and require real-time scheduling and execution. In order to complete their tasks, agents need to coordinate their actions both constantly and extensively. We present an approach that consists of a standard operating procedure and a look-ahead coordination. The standard operating procedure regulates task coupling and minimizes communication. The look-ahead coordination increases agents' global visibility and provides indicative information for decision adjustment. The goal of our approach is to prune decision myopia while maintaining system responsiveness in real-time, dynamic environments. Experimental results in job shop scheduling problems show that (1) the look-ahead coordination significantly enhances the performance of the standard operating procedure in solution quality, (2) the approach is capable of producing solutions of very high quality in a real-time environment. Introduction Most research on multiagent sy...

. Hospital Patient Scheduling is an inherently distributed problem because of the way real hospitals are organized. As medical procedures have become more complex, and their associated tests and treatments have become interrelated, the current ad hoc patient scheduling solutions have been observed to break down [23]. We propose a multi-agent solution using the Generalized Partial Global Planning (GPGP) approach that preserves the existing human organization and authority structures, while providing better system-level performance (increased hospital unit throughput and decreased patient stay time). To do this, we extend GPGP with a new coordination mechanism to handle mutually exclusive resource relationships. Like the other GPGP mechanisms, the new mechanism can be applied to any problem with the appropriate resource relationship. We evaluate this new mechanism in the context of the hospital patient scheduling problem, and examine the effect of increasing interrelations between tasks ...

Most existing methods for scheduling are based on centralized or hierarchical decision making using monolithic models. In this study, we investigate a new method based on a distributed and locally autonomous decision structure using the notion of combinatorial auction. In combinatorial auction the bidders demand a combination of dependent objects with a single bid. We show that not only can we use this auction mechanism to handle complex resource scheduling problems, but there exist strong links between combinatorial auction and Lagrangean-based decomposition. Exploring some of these properties, we characterize combinatorial auction using auction protocols and payment functions. This study is a #rst step toward developing a distributed scheduling framework that maintains system-wide performance while accommodating local preferences and objectives. We provide some insights to this framework by demonstrating four di#erent versions of the auction mechanism using job shop scheduling proble...

(For membership information, consult our web page) The material herein is copyrighted material. It may not be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from AAAI.

MarCon (Market-based Constraints) applies market-based control to distributed constraint problems. It offers a new approach to distributing constraint problems that avoids challenges faced by current approaches in some problem domains, and it provides a systematic method for applying markets to a wide array of problems. Constraint agents interact with one another via the variable agents in which they have a common interest, using expressions of their preferences over sets of assignments. Each variable integrates this information from the constraints interested in it and provides feedback that enables the constraints to shrink their sets of assignments until they converge on a solution. MarCon originated in a system for supporting human product designers, and its mechanisms are particularly useful for applications integrating human and machine intelligence to explore implicit constraints. MarCon has been tested in the domain of mechanical design, in which its set-narrowing process is particularly useful. 1.

This article describes an approach to solving distributed constraint optimization problems (DCOP) with n-ary constraints. A key instance of this problem is distributed resource-constrained task scheduling, in which limited resource capacities implicitly imply n-ary relations among the start-times of the tasks. We describe ADOPT-N, an extension of ADOPT [16], a recent successful algorithm for DCOP. ADOPT-N is an optimal asynchronous distributed n-ary constraint optimization algorithm in which specific agents are empowered with n-ary constraint evaluation capabilities. We show how the algorithm’s correctness and optimality relies on (1) the choice of which agents to dedicate to constraint evaluation, and (2) an admissible (partial) variable ordering. Moreover, we demonstrate empirically how ADOPT-N’s performance depends on how much knowledge about the n-ary violations that can arise during resolution can be provided to the algorithm. 1