Cooperation Without Communication

Genesereth, Michael; Ginsberg, Matthew L.; Rosenschein, Jeffrey S.

doi:10.1016/b978-0-934613-63-7.50026-7

Cited by 87 publications

(41 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We still have no broadly useful definitions of terms such as coordination, cooperation, or interaction. To be sure, we do have the "cooperation without communication," "rational deal-making," and "probabilistic interaction" theories of Rosenschein, Genesereth, Breese, and Ginsberg (Rosenschein and Genesereth 1985;Genesereth, Ginsberg, and Rosenschein 1986;Rosenschein and Breese 1989), but these are bound by highly restrictive assumptions. The promising "metalevel control" and "partial global planning" techniques of Lesser and his colleagues (Corkill and Lesser 1983;Durfee and Lesser 1991) have not yet become full-fledged coordination theories that can guide us to other new practical coordination techniques.…”

Section: Discussionmentioning

confidence: 99%

Analyzing a quantitative coordination relationship

Decker

Lesser

1993

Group Decis Negot

View full text Add to dashboard Cite

Coordination is a crucial behavior in cooperative distributed problem solving (CDPS). Analyzing coordination requires an understanding of the interplay between the agents, their problem, and their environment. The core behaviors of distributed coordination in CDPS systems are the coherent specification and scheduling of tasks over the set of distributed agents working on sets of interrelated problems. The complexity of, and uncertainty about, the problem interrelationships make distributed task coordination difficult. This article describes a causal model of this process that links the interrelationships, called coordination relationships, to the local scheduling constraints of distributed agents. Besides coordination relationships, environmental uncertainty and the lack of infinite computational resources also make distributed coordination difficult.It is not only the presence or absence of a coordination relationship that is important, but its quantitative properties: how likely is it to appear, how significant is its effect, and so on. These aspects determine the usefulness of a particular coordination relationship in the context defined by an environment, a problem to be solved, and an agent architecture. This article discusses the analysis of coordination relationships, using as an example our abstract model for the facilitates relationship. We detail the derivation and assumptions of this model and apply it to the design of a generalized coordination module that is separate from, and interfaces cleanly with, the local scheduler of a CDPS agent. A set of simulation experiments is described that test our assumptions and design process in the coordination of a group of real-time problem-solving agents.

show abstract

Section: Discussionmentioning

confidence: 99%

Analyzing a quantitative coordination relationship

Decker

Lesser

1993

Group Decis Negot

View full text Add to dashboard Cite

show abstract

“…In the first paradigm, a single intelligent agent (the "master") constructs a plan to be carried out by a group of agents (the "slaves"), then hands out pieces of the plan to the relevant individuals (Rosenschein 1982;Konolige 1982;Lesser 1987;Katz and Rosenschein 1993). In the second paradigm, a group of intelligent agents jointly construct the final plan and subsequently cooperate in carrying it out (Genesereth et al 1986;Durfee and Lesser 1987).…”

Section: Multiagent Planningmentioning

confidence: 99%

Planning to please: Following another agent's intended plan

Ephrati

Rosenschein

1993

Group Decis Negot

Self Cite

View full text Add to dashboard Cite

In this article we analyze a particular model of control among intelligent agents, that of nonabsolute control. Non-absolute control involves a "supervisor" agent that issues orders to a "subordinate" agent. An example might be a human agent on Earth directing the activities of a Mars-based semi-autonomous vehicle.Both agents operate with essentially the same goals. The subordinate agent, however, is assumed to have access to some information that the supervisor does not have. The agent is thus expected to exercise its judgment in following orders (i.e., following the true intent of the supervisor, to the best of its ability). After presenting our model, we discuss the planning problem: how would a subordinate agent choose among alternative plans? Our solutions focus on evaluating the distance between candidate plans, and is appropriate to any scenario in which one agent wants to follow (as much as possible) another agent's plan.

show abstract

“…Even in the presence of erroneously calculated state transitions, such an approach might work. For instance, Dolev et al, [1986] and Genesereth et al, [1986] present general approaches to the description of distributed systems with faults. It is not clear however to what extent these approaches are useful for the description of asynchronously parallel Boltzmann machines.…”

Section: Asynchronous Parallelismmentioning

confidence: 99%

Boltzmann machines as a model for parallel annealing

Aarts

Korst

1991

Algorithmica

View full text Add to dashboard Cite

Abstract. The potential of Boltzmann machines to cope with difficult combinatorial optimization problems is investigated. A discussion of various (parallel) models of Boltzmann machines is given based on the theory of Markov chains. A general strategy is presented for solving (approximately) combinatorial optimization problems with a Boltzmann machine. The strategy is illustrated by discussing the details for two different problems, namely MATCHING and GRAPH PARTITION-ING.Key Words. Boltzmann machines, Combinatorial optimization, Connectionist models, Neural networks, Simulated annealing.1. Introduction. Simulated annealing is a generally applicable solution method in the field of combinatorial optimization based on an analogy with the physical annealing process [Kirkpatrick et al., 1983]. Numerous performance evaluations have led to the general opinion that simulated annealing can obtain high-quality solutions but often at the cost of large running times [Aarts and Korst, 1989a; Van Laarhoven and Aarts, 19873. As problems will inevitably increase in size, the situation with respect to computational effort will only worsen.Recent efforts have concentrated on investigating possibilities for speeding up simulated annealing algorithms by executing them on parallel machines. Unfortunately, with only moderate success I-Aarts and Korst, 1989a]. Most parallel algorithms presented in the literature follow rather strictly the basic concept of the annealing algorithm, i.e., generation of a sequence of solutions, which is intrinsically sequential and therefore leads to low efficiencies on parallel machines. Our general feeling is that the design of efficient parallel annealing algorithms requires a computational model that differs substantially from the traditional sequential annealing model. Recent interest focuses on computational models such as connectionist models that support in a natural way the exploitation of massive parallelism as is done in the human brain. Connectionist models [Feldman and Ballard, 1982;Fahlman and Hinton, 1987] can be viewed as generalizations of neural network models. They are based on the assumption that information can be processed by massively parallel networks consisting of cooperating simple neuron-like computing elements that are highly interconnected by weighted links. Furthermore, a typical feature of these models is that the response of an individual computing element to other elements in the network is given by a nonlinear scalar function of the weighted

show abstract

Cooperation Without Communication

Cited by 87 publications

References 11 publications

Analyzing a quantitative coordination relationship

Analyzing a quantitative coordination relationship

Planning to please: Following another agent's intended plan

Boltzmann machines as a model for parallel annealing

Contact Info

Product

Resources

About