Results 1  10
of
10
Impact of problem centralization in distributed constraint optimization algorithms
 In AAMAS
, 2005
"... Recent progress in Distributed Constraint Optimization Problems (DCOP) has led to a range of algorithms now available which differ in their amount of problem centralization. Problem centralization can have a significant impact on the amount of computation required by an agent but unfortunately the d ..."
Abstract

Cited by 47 (4 self)
 Add to MetaCart
(Show Context)
Recent progress in Distributed Constraint Optimization Problems (DCOP) has led to a range of algorithms now available which differ in their amount of problem centralization. Problem centralization can have a significant impact on the amount of computation required by an agent but unfortunately the dominant evaluation metric of “number of cycles ” fails to account for this cost. We analyze the relative performance of two recent algorithms for DCOP: OptAPO, which performs partial centralization, and Adopt, which maintains distribution of the DCOP. Previous comparison of Adopt and OptAPO has found that OptAPO requires fewer cycles than Adopt. We extend the cycles metric to define “CycleBased Runtime (CBR) ” to account for both the amount of computation required in each cycle and the communication latency between cycles. Using the CBR metric, we show that Adopt outperforms OptAPO under a range of communication latencies. We also ask: What level of centralization is most suitable for a given communication latency? We use CBR to create performance curves for three algorithms that vary in degree of centralization, namely Adopt, OptAPO, and centralized Branch and Bound search.
Open constraint programming
 ARTIFICIAL INTELLIGENCE 161 (2005) 181–208
, 2005
"... Traditionally, constraint satisfaction problems (CSP) have assumed closedworld scenarios where all domains and constraints are fixed from the beginning. With the Internet, many of the traditional CSP applications in resource allocation, scheduling and planning pose themselves in openworld settings ..."
Abstract

Cited by 37 (5 self)
 Add to MetaCart
(Show Context)
Traditionally, constraint satisfaction problems (CSP) have assumed closedworld scenarios where all domains and constraints are fixed from the beginning. With the Internet, many of the traditional CSP applications in resource allocation, scheduling and planning pose themselves in openworld settings, where domains and constraints must be discovered from different sources in a network. To model this scenario, we define open constraint satisfaction problems (OCSP) as CSP where domains and constraints are incrementally discovered through a network. We then extend the concept to open constraint optimization (OCOP). OCSP can be solved without complete knowledge of the variable domains, and we give sound and complete algorithms. We show that OCOP require the additional assumption that variable domains and relations are revealed in nondecreasing order of preference. We present a variety of algorithms for solving OCOP in the possibilistic and weighted model. We compare the algorithms through experiments on randomly generated problems. We show that in certain cases, open constraint programming can require significantly less information than traditional methods where gathering information and solving the CSP are separated. This leads to a reduction in network traffic and server load, and improves privacy in distributed problem solving.
Decision tradeoff using examplecritiquing and constraint programming
 Constraints
"... Abstract. We consider constructive preference elicitation for decision aid systems in applications such as configuration or electronic catalogs. We are particularly interested in supporting decision tradeoff, where preferences are revised in response to the available outcomes. In several userinvolv ..."
Abstract

Cited by 24 (8 self)
 Add to MetaCart
(Show Context)
Abstract. We consider constructive preference elicitation for decision aid systems in applications such as configuration or electronic catalogs. We are particularly interested in supporting decision tradeoff, where preferences are revised in response to the available outcomes. In several userinvolved decision aid systems we have designed in the past, we were able to observe three generic tradeoff strategies that people like to use. We show how a preference model based on soft constraints is wellsuited for supporting these strategies. Such a framework provides an agile preference model particularly powerful for preference revision during tradeoff analysis. We further show how to integrate the constraintbased preference model with an interaction model called examplecritiquing. We report on user studies which show that this model offers significant advantages over the commonly used rankedlist model, especially when the decision problem becomes complex.
Dealing with incomplete preferences in soft constraint problems
 CP 2007 (The 13th International Conference on Principles and Practice of Constraint Programming
, 2007
"... Abstract. We consider soft constraint problems where some of the preferences may be unspecified. This models, for example, situations with several agents providing the data, or with possible privacy issues. In this context, we study how to find an optimal solution without having to wait for all the ..."
Abstract

Cited by 18 (10 self)
 Add to MetaCart
(Show Context)
Abstract. We consider soft constraint problems where some of the preferences may be unspecified. This models, for example, situations with several agents providing the data, or with possible privacy issues. In this context, we study how to find an optimal solution without having to wait for all the preferences. In particular, we define an algorithm to find a solution which is necessarily optimal, that is, optimal no matter what the missing data will be, with the aim to ask the user to reveal as few preferences as possible. Experimental results show that in many cases a necessarily optimal solution can be found without eliciting too many preferences. 1
Elicitation Strategies for Soft Constraint Problems with Missing Preferences: Properties, Algorithms and Experimental Studies
"... We consider soft constraint problems where some of the preferences may be unspecified. This models, for example, settings where agents are distributed and have privacy issues, or where there is an ongoing preference elicitation process. In this context, we study how to find an optimal solution witho ..."
Abstract

Cited by 9 (1 self)
 Add to MetaCart
(Show Context)
We consider soft constraint problems where some of the preferences may be unspecified. This models, for example, settings where agents are distributed and have privacy issues, or where there is an ongoing preference elicitation process. In this context, we study how to find an optimal solution without having to wait for all the preferences. In particular, we define algorithms, that interleave search and preference elicitation, to find a solution which is necessarily optimal, that is, optimal no matter what the missing data will be, with the aim to ask the user to reveal as few preferences as possible. We define a combined solving and preference elicitation scheme with a large number of different instantiations, each corresponding to a concrete algorithm, which we compare experimentally. We compute both the number of elicited preferences and the user effort, which may be larger, as it contains all the preference values the user has to compute to be able to respond to the elicitation requests. While the number of elicited preferences is important when the concern is to communicate as little information as possible, the user effort measures also the hidden work the user has to do to be able to communicate the elicited preferences. Our experimental results on classical, fuzzy, weighted and temporal incomplete CSPs show that some of our algorithms are very good at finding a necessarily optimal solution while asking the user for only a very small fraction of the missing preferences. The user effort is also very small for the best algorithms.
Retractable Contract Network for Distributed Scheduling
"... This paper is about distributed scheduling where individual agents have potentially conflicting interests. It is motivated by BT’s workforce scheduling problem, where multiple service providers have to serve multiple service buyers. The service providers and buyers all attempt to maximize their own ..."
Abstract

Cited by 2 (2 self)
 Add to MetaCart
(Show Context)
This paper is about distributed scheduling where individual agents have potentially conflicting interests. It is motivated by BT’s workforce scheduling problem, where multiple service providers have to serve multiple service buyers. The service providers and buyers all attempt to maximize their own utility. The overall problem is a multiobjective optimization problem; for example, one has to maximize completion rates and service quality and minimize travelling distances. In this paper, BT’s problem is modelled as an open constraint optimization system. Standard contract net is a practical strategy in distributed scheduling where agents may have conflicting objectives. In this paper, we have introduced a retractable contract net protocol, which we call RECONNET, that supports hillclimbing in the space of solutions. It is built upon a jobrelease and compensation mechanism. RECONNET is a general protocol, which could be used to implement complex metaheuristic algorithms such as Tabu Search and Guided Local Search. A system based on RECONNET has been implemented for BT’s workforce scheduling problem. The software, which we call ASMCR, allows the management to have full control over the company’s multiobjectives. The manager generates a Pareto set of solution by defining, for each buyer and seller, the weights given to each objective. ASMCR gives service buyers and sellers ownership of their problem and freedom to maximize their performance under the criteria defined by the management. ASMCR took 5 to 15 minutes to complete when tested on realsized problems. It has potential to be developed into practical solutions to BT’s workforce scheduling problem.
A Costbased Model and Algorithms for Interleaving Solving and Elicitation of CSPs
"... In Constraint Satisfaction Problems it is usually assumed that the CSP is available before the solving process begins, that is, the elicitation of the problem is completed before we attempt to solve the problem. As discussed in the work on Open Constraints and Interactive CSPs, there are situations ..."
Abstract
 Add to MetaCart
In Constraint Satisfaction Problems it is usually assumed that the CSP is available before the solving process begins, that is, the elicitation of the problem is completed before we attempt to solve the problem. As discussed in the work on Open Constraints and Interactive CSPs, there are situations where it can be advantageous and natural to interleave the elicitation and the solving. In particular, it may be expensive, in terms of time or other costs, to elicit certain constraints or parts of the constraints, and, we may very well not need all the complete constraints to be available in order for us to find a solution. In this paper we consider algorithms which take these costs into account. Constraints may be initially incomplete: it may be unknown whether certain tuples satisfy the constraint or not. We assume that we can determine such an unknown tuple, i.e., find out whether this tuple is in the constraint or not, but doing so incurs a known cost, which may vary between tuples. We also assume that we know the probability of an unknown tuple satisfying a constraint. An optimal algorithm for this situation is defined to be one which incurs minimal expected cost in finding a solution. We define algorithms for this problem, based on backtracking search. Specifically, we consider a simple iterative algorithm based on a cost limit on which unknowns may be determined, and a more complex algorithm which delays determining an unknown in order to estimate better whether doing so is worthwhile. We show experimentally the benefit in terms of cost of using the more sophisticated algorithms.
A Constraint Handling Rules Implementation for KnownArcConsistency in Interactive Constraint Satisfaction Problems
, 2004
"... In classical CLP(FD) systems, domains of variables are completely known at the beginning of the constraint propagation process. However, in systems interacting with an external environment, acquiring the whole domains of variables before the beginning of constraint propagation may cause waste of com ..."
Abstract
 Add to MetaCart
In classical CLP(FD) systems, domains of variables are completely known at the beginning of the constraint propagation process. However, in systems interacting with an external environment, acquiring the whole domains of variables before the beginning of constraint propagation may cause waste of computation time, or even obsolescence of the acquired data at the time of use.
Interleaving Solving and Elicitation of Constraint Satisfaction Problems Based on Expected Cost
"... Abstract. We consider Constraint Satisfaction Problems in which constraints can be initially incomplete, where it is unknown whether certain tuples satisfy the constraint or not. We assume that we can determine such an unknown tuple, i.e., find out whether this tuple is in the constraint or not, but ..."
Abstract
 Add to MetaCart
(Show Context)
Abstract. We consider Constraint Satisfaction Problems in which constraints can be initially incomplete, where it is unknown whether certain tuples satisfy the constraint or not. We assume that we can determine such an unknown tuple, i.e., find out whether this tuple is in the constraint or not, but doing so incurs a known cost, which may vary between tuples. We also assume that we know the probability of an unknown tuple satisfying a constraint. We define algorithms for this problem, based on backtracking search. Specifically, we consider a simple iterative algorithm based on a cost limit on which unknowns may be determined, and a more complex algorithm that delays determining an unknown in order to estimate better whether doing so is worthwhile. We show experimentally that the more sophisticated algorithms can greatly reduce the average cost. 1