V. R. Lesser, "A retrospective view of FA/C distributed problem solving", IEEE Trans. on Sys. Man and Cyb. SMC-21(6): 1347-1362. (1991).  Home/Search   Document Details and Download   Summary   Related Articles   Check

....for measuring worth functions (subgoals set, distance between states, probabilistic distance) but do not commit themselves to any of them, assigning an absolutely arbitrary number to each agent s goal. 6 There are several taxonomies of goal relationships in DAI literature (Lesser et al. 1989; Lesser, 1991; Castelfranchi et al. 1992; Von Martial, 1992) but only those linked to the FA C (Functionally Accurate Cooperative) model are useful. We could adopt instead whatever model which bases the significance of the goals on simple AND OR trees. However appeling this approach is, it is difficult to ....

Lesser, V. (1991). A retrospective view of FA/C Distributed Problem Solving. IEEE Transactions on Systems, Man and Cybernetics, 21:1347--1362.

....social layer of the agents. It ranges from a 2 For additional discussion of the concepts of Nash and subgame perfect equilibrium, and Pareto optimality, see Sections 1.5 and 2.1 below. 6 Structured Unstructured Moses, Shoham Tennenholtz [38, 53] DPS Davis Smith [54] Malone [37] Lesser [33, 6] Durfee [11, 12] Zlotkin Rosenschein [64, 45, 61] Sycara [57, 58] Kraus Lehmann [24, 25] MA Wellman [60] Ephrati Rosenschein [13] Grosz [34, 19] Gasser [17] Kraus, Wilkenfeld Zlotkin Figure 1: The rows indicates Degree of control on the social Layer . The columns indicates degree of ....

V.R. Lesser. A retrospective view of fa/c distributed problem solving. IEEE Transactions on Systems, Man, and Cybernetics, Special Issue on Distributed Artificial Intelligence, 21(6):1347--1362, December 1991.

....autonomous agents. We assume that each agent has its own utility function, and that rational behavior involves maximizing its expected utility. Our work belongs to the DAI class of Multi Agent Systems(MA) e.g. 52, 61, 63, 65] rather than to the Distributed Problem Solving (DPS) class (e.g. [58, 17, 41]) as discussed in Section 2 below. In previous work [31, 32, 34] we have developed a formal strategic model of negotiation that takes into consideration the passage of time during the negotiation process. In this paper, we extend this model to deal with incomplete information in multiple ....

V.R. Lesser. A retrospective view of FA/C distributed problem solving. IEEE Transactions on Systems, Man, and Cybernetics, 21(6):1347--1362, 1991.

....of software agents to demonstrate the contracting techniques introduced above. 2 Related work in DAI Research in DAI is divided into two basic classes: Cooperative Distributed Problem Solving and Multi Agent Systems (MA) 8, 28] Research in Cooperative Distributed Problem Solving (e.g. [61, 59, 12, 101, 18]) considers how the work involved in solving a particular problem can be divided among a number of modules or nodes. The modules in a Cooperative Distributed Problem Solving system are centrally designed to improve the following properties of the system [8] Performance: Concurrency may ....

....other agents, usually on a contractual basis [101] This approach assumes that an agent not otherwise occupied will readily take on the task and do it to the best of its abilities. Similarly, results and information are shared among agents in such environments with no expectation of reciprocation [61, 59, 12]. This benevolence is based on an assumption common to many approaches to coordination: that the system s goal is to solve the problem as best it can, thereby giving the agents shared, often implicit, global goals that they are all unselfishly committed to achieving. One of the techniques that is ....

V. Lesser. A retrospective view of FA/C distributed problem solving. IEEE Transactions on Systems, Man, and Cybernetics, Special Issue on Distributed Artificial Intelligence, 21(6):1347--1362, December 1991. 67

....and infer a set of requirements on information distribution and organizational policies. They point out that in a DPS scenario, different agents may have different capabilities, limited knowledge and resources, and thus differing appropriateness in solving the problem at hand. Lesser et al. [16] describes the FA C (functionally accurate, cooperative) architecture in which agents exchange partial and tentative results in order to converge to a solution. Joshi [13] proposes an epistemic utility based approach which allows each agent to dynamically learn about the capabilities of other ....

V. R. Lesser, A Retrospective View of FA/C Distributed Problem Solving, IEEE Transactions on Systems, Man, and Cybernetics 21 (1991), no. 6, 1347--1363.

....and infer a set of requirements on information distribution and organizational policies. They point out that in a DPS scenario, different agents may have different capabilities, limited knowledge and resources, and thus differing appropriateness in solving the problem at hand. Lesser et. al [19] describes the FA C (functionally accurate, cooperative) architecture in which agents exchange partial and tentative results in order to converge to a solution. Most of these coordination and cooperation techniques however, are static in nature. They do not learn as their environment changes, ....

V. R. Lesser, A Retrospective View of FA/C Distributed Problem Solving, IEEE Transactions on Systems, Man, and Cybernetics 21 (1991), no. 6, 1347--1363.

....requirements on information distribution and organizational policies. They point out that in a distributed problem solving scenario, different agents may have different capabilities, limited knowledge and resources, and thus differing appropriateness in solving the problem at hand. Lesser et al. Les91] describes the FA C (functionally accurate, cooperative) architecture in which agents exchange partial and tentative results in order to converge to a solution. Decker and Lesser [DL92] propose to use the generalized partial global planning algorithm as an approach to distributed coordination. ....

V. R. Lesser. A Retrospective View of FA/C Distributed Problem Solving. IEEE Transactions on Systems, Man, and Cybernetics, 21(6):1347-- 1363, 1991.

....then monitors the network s evolving state. This particular scenario was chosen because it is typical of a class of behaviour, called functionally accurate cooperative (in which the cooperating agents do not possess the necessary information to solve all their subproblems completely and accurately [49]) which consistently appeared in the industrial applications which were studied. In their original stand alone state, both the AAA and the BAI had their own very rudimentary CSI which could collate alarm messages but not detect disturbances. This meant the two diagnosis systems were continually ....

V. R. Lesser, A Retrospective View of FA/C Distributed Problem Solving, IEEE Trans. on Systems, Man and Cybernetics 21 (6) (1991) 1347-1362.

....for agent interactions at the consensus level of autonomy. Locally autonomous agents, which plan independently but act as part of a larger system, exhibit the most diverse behavior. They may be fully cooperative and act under a functionally accurate, cooperative distributed system (FA C) 11] 12][13]. Alternatively, they may choose to act more selfishly. Durfee and Lesser s [3] method of communicating partial global plans allows agents to interact within a system in many different ways. However, they leave the method of choosing an interaction style open. Locally autonomous agents are unique ....

Lesser, V. R. A Retrospective View of FA/C Distributed Problem Solving. IEEE Transactions on Systems, Man, and Cybernetics 21(6):1347-1362.

....s o u r c e s require Beliefs Plans support justified by support achieve schedule describe Figure 1: Three layers of representation for the interdependencies among goals, resources, plans, and beliefs. 4.1. Layer 1: The Goal Structure Goals can be represented as classical AND OR goal structures [Lesser, 1991]. These goal structures may be dynamic or static, and need not be fully developed before an agent begins operation. A goal structure may be constructed through social activities such as negotiation, cooperation, and observation of environmental changes. The dependencies among goals are ....

....shared by two or more agents, can be viewed as goals intended by all involved agents and can be included in the same type of goal structure as goals intended by single agents. Additionally, a goal s dependency on particular resources can be represented in the same way through unidirectional links [Lesser, 1991]. In this fashion, resource constraints among goals can be characterized. This type of interdependency is defined indirectly through the availability of system resources and is not represented in the goal layer itself. 4.2. Layer 2: The Plan Space The second layer is the representation of ....

Lesser, V. R. A Retrospective View of FA/C Distributed Problem Solving. IEEE Transactions on Systems, Man, and Cybernetics, 21(6), 1347-1362, 1991.

....may or may not correspond to actual The Laboratory for Intelligent Processes and Systems TR98 UT LIPS AGENTS 02 Copyright The University of Texas at Austin 4 truth in all cases. 3.1. Layer 1: The Goal Structure Potential agent goals can be represented as classical AND OR goal structures [16], also called goal trees. These goal structures may be dynamic or static, and need not be fully developed before an agent begins operation. A goal structure may be constructed through social activities such as negotiation, cooperation, and observation of environmental change. The dependencies ....

....The dependencies among goals are represented as links that connect an agent s goals to its other goals or to goals of other agents. These links can be unidirectional or bidirectional [14] Additionally, a goal s dependency on particular resources can be represented through unidirectional links [16]. In this fashion, resource constraints among goals can be characterized. Resource constraints are defined through the availability of system resources and need not be represented in the goal layer itself. In the current implementation of this representation based on Sensible Agents [1] goals ....

Lesser, V. R. 1991. A Retrospective View of FA/C Distributed Problem Solving. IEEE Transactions on systems, man, and cybernetics 21(6): 1347-1362.

....The agents make these decisions by monitoring resource availability and status, execution, and goal achievement. Figure 5 shows the sensible agent system interface for the system and the definition of a goal tree. Agents plan from AND OR goal trees that represent the agents alternatives [16, 17]. The goal sub goal relationship reflects a task reduction paradigm. Autonomy levels can be assigned at a single level or at each step of task reduction (to each goal, its sub goals, and any further sub goals these sub goals may have) The goal tree displayed in Figure 5 defines how an agent ....

V. R. Lesser, "A Retrospective View of FA/C Distributed Problem Solving," IEEE Transactions on Systems, Man, and Cybernetics, vol. 21, pp. 1347-1362, 1991.

....an assignment to the autonomy constructs. The following discussion of the individual autonomy constructs refers to agent goals and subgoals. The goal subgoal relationship reflects a task reduction planning paradigm. This planning paradigm is common in work on agent based systems (Jennings, 1993; Lesser, 1991). Autonomy levels can be assigned at each step of task reduction (to each goal, its subgoals, and any further subgoals these subgoals may have) However, the nature of the autonomy constructs does not restrict their applicability to the goal subgoal paradigm alone; autonomy levels can also be ....

Lesser, V. R. 1991. A Retrospective View of FA/C Distributed Problem Solving. IEEE Transactions on Systems, Man, and Cybernetics 21(6): 1347-1362.

....will introduce how each phase interacts with each other. For simplicity, let us assume that there are two agents in the system, agent1 and agent2, both planning for a single goal (goal1, to maintain an interference free state, as shown in the goal tree of Figure 4) AND OR goal trees developed in [25] are used to represent potential goal decomposition and combinations as well as their interactions. As for the implementation of the naval radar frequency management problems, we selected the Sensible Agent Architecture [23] and assigned each phase of problem solving activities to certain ....

V. R. Lesser, "A Retrospective View of FA/C Distributed Problem Solving," IEEE Transactions on Systems, Man, and Cybernetics, vol. 21, pp. 1347-1362, 1991.

....of autonomy. Locally autonomous agents, which plan independently but act as part of a larger system, exhibit the most diverse behavior. They may be fully cooperative and act under a functionally accurate, cooperative distributed system (FA C) Lesser and Corkill 1981, Lesser and Corkill 1983, Lesser 1991]. Alternatively, they may choose to act more selfishly. Durfee and Lesser s [1987] method of communicating partial global plans allows agents to interact within a system in many different ways. However, they leave the method of choosing an interaction style open. Locally autonomous agents are ....

Lesser, V. R. 1991. A Retrospective View of FA/C Distributed Problem Solving. IEEE Transactions on Systems, Man, and Cybernetics 21(6):1347-1362.

....of an agent s problem solving interactions in an application independent fashion. Multi agent researchers use simulation environments to test algorithms and representations in order to accurately measure the impact of new research. Existing simulation environments include: DVMT (now DRESUN) (Lesser, 1991), MACE (Gasser et al. 1989) and MICE (Durfee and Montgomery, 1990) Additionally, a number of single use simulations have been used to measure or compare performance for many research topics including swarm behavior (Beslon et al. 1998) constraint planning (Liu, 1996) and hierarchical ....

....each agent can be displayed, including the internal workings of each of the modules, interactions among modules, and interactions among agents. Figure 5 shows an example of the definition of a goal tree. Agents plan from AND OR goal trees which represent the agents alternatives (Jennings, 1993; Lesser, 1991). The goal subgoal relationship reflects a taskreduction planning paradigm. Autonomy levels can be assigned at a single level or at each step of task reduction (to each goal, its subgoals, and any further subgoals these subgoals may have) The goal tree displayed in Figure 5 defines how an agent ....

Lesser, V. R. 1991. A Retrospective View of FA/C Distributed Problem Solving. IEEE Transactions on Systems, Man, and Cybernetics 21(6): 1347-1362.

....among the set of agents arises. Due to its partial view of the problem solving situation, an agent may have access only to a part of the environment, and communication bandwidth limitations and heterogeneity of representations may limit its view of other agents state. Given these uncertainties[18], an agent may have to do evidential reasoning and focused negotiation to resolve them to the extent that it can make positive contributions to the on going problem solving process. More specifically, in a distributed case based reasoning system (DCBR) each agent s partial view may result in best ....

....the system changes the report type that it is trying to generate. This sort of broadening of allowed strategies upon failure is similar to the broadening of retrieval in our algorithm. The negotiated retrieval algorithm benefits from a long history of thought in DAI and Multiagent systems. Lesser[18] discusses the importance of interactions among subgoals and data distributed across a set of agents in distributed problem solving. Selectively sharing relevant partial results of a local search or data can enhance the global problem solving quality and efficiency without paying a large price for ....

Lesser, V. R., "A retrospective view of FA/C distributed problem solving", IEEE Systems, Man, and Cybernetics, 21(6), pp 1346-1363.

.... when it is not possible to decompose the problem into a set of subproblems such that there is a perfect fit between the computational requirements for effectively solving each subproblem and the location of information, expertise, processing, and communication resources in the agent network [33] [34]. This lack of a perfect fit often leads to a situation where there may be insufficient local information or resources for an agent to completely or accurately solve its assigned subproblems through its own processing. Further, resource contention issues in multiagent systems do not entail simply ....

.... that in the future agents will be required to perform some form of metalevel reasoning so as to balance the level of optimality of their control decisions with the level of resources required to make the decisions, based on the characteristics of their tasks and the environment [9] 12] 15] [34], 76] 3 The second principle relates to the need for flexibility in agent problem solving. Agent flexibility with respect to the availability, completeness and accuracy of its information and the availability and capabilities of external resources is often a key aspect of a multiagent system ....

[Article contains additional citation context not shown here]

# V. Lesser, "A Retrospective View of FA/C Distributed Problem Solving," IEEE Trans. Systems, Man, and Cybernetics, vol. 21, no. 6, pp. 1,347--1,362, Nov.--Dec. 1991.

....CDA 8922572. The content of the information does not necessarily reflect the position or the policy of the Government and no official endorsement should be inferred. 0 1 Introduction The core of the distributed coordination problem in Cooperative Distributed Problem Solving (CDPS) systems [9, 17] is the coherent specification and scheduling of tasks over the set of distributed agents working on sets of interrelated problems. From a purely computational standpoint, coordination allows better use of computational resources by intelligent scheduling, e.g. reducing needless redundancy, ....

....environment Representation of another agent s task Representation of a local task and scheduling order X Y Fig. 1: Conceptual view of the distributed coordination process For example, the abstract domain of distributed search contains several potentially uncertain coordination relationships [17]. The primary uncertainty lies in the subgoal relationship: how a particular task relates to the problem as a whole. Will this task be a part of a final solution Are there multiple paths to the goal How much effort will it take A secondary uncertainty lies in the presence of a facilitates ....

V. R. Lesser. A retrospective view of FA/C distributed problem solving. IEEE Transactions on Systems, Man, and Cybernetics, 21(6):1347--1363, November 1991.

....and communication resources may not always be worthwhile. In general, coordination requires the acquisition of non local information about the state of activities of other agents and requires computations to reason about this information in order to make more informed local control decisions [16]. For example, coordination to avoid redundant activities may be unnecessary if processing resources are not overloaded and if communication is neither expensive nor overloaded. In this case, local problem solving is more efficient since there is no additional overhead for coordination, and ....

....to work on and how to best achieve these goals. This means that quite different decision making may occur at different agents. The lack of information about the problemsolving states of other agents leads to uncertainty in local decision making which can, in some cases, cause serious problems [16]. Since the costs for all agents exchanging all information seem unrealistically high, it is important to understand what information in (specific) problemsolving situations should be exchanged. Furthermore, even if an agent has the appropriate nonlocal information, computation costs to explore to ....

V. R. Lesser, "A Retrospective View of FA/C Distributed Problem Solving," IEEE Trans. on System, Man, and Cybernetics, Vol. 21, No. 6, pp. 1347-1362, 1991.

....may lead to inappropriate decisions about which activity it should do next, what results it should transmit to other agents or what results it should ask other agents to produce. If an agent has a view of the activities (task structures) of other agents, it can make more informed choices [4,6,7,13]. Another difficulty is that even with this type of meta level information, there is still residual uncertainty about the outcomes of tasks and what future tasks will be coming into the system that may result in agents still exhibiting non coherent behavior. These difficulties with achieving ....

.... due to the inherent uncertainty in agents activities and the cost of meta level processing, it may not be worthwhile to acquire a complete view of other agents activities, and thus some level of non coherent activity may be necessary in order to achieve an optimal coordination strategy [13]. For example, coordination to avoid redundant activities may be unnecessary if processing resources are not overloaded and if communication channels are neither expensive nor overloaded. In this case, local problem solving is done more efficiently where there is no additional overhead for ....

Lesser, V. R., "A Retrospective View of FA/C Distributed Problem Solving," IEEE Transactions on Systems, Man, and Cybernetics, 21(6):1347-1362, Nov./Dec. 1991.

.... arise when it is not possible to decompose the problem into a set of subproblems such that there is a perfect fit between the computational requirements for effectively solving each subproblem and the location of information, expertise, processing, and communication resources in the agent network [33, 34]. This lack of a perfect fit often leads to a situation where there may be insufficient local information or resources for an agent to completely or accurately solve its assigned subproblems through its own processing. Further, resource contention issues in multiagent systems do not entail simply ....

.... is my conjecture that in the future agents will be required to perform some form of meta level reasoning so as to balance the level of optimality of their control decisions with the level of resources required to make the decisions, based on the characteristics of their tasks and the environment [9, 12, 15, 34, 76]. 3 The second principle relates to the need for flexibility in agent problem solving. Agent flexibility with respect to the availability, completeness and accuracy of its information and the availability and capabilities of external resources is often a key aspect of a multiagent system design. ....

[Article contains additional citation context not shown here]

V. Lesser, "A Retrospective View of FA/C Distributed Problem Solving," IEEE Trans. Systems, Man, and Cybernetics, Vol. 21, No. 6, Nov.--Dec. 1991, pp. 1347--1362.

No context found.

V. R. Lesser, "A retrospective view of FA/C distributed problem solving", IEEE Trans. on Sys. Man and Cyb. SMC-21(6): 1347-1362. (1991).

No context found.

V. R. Lesser, "A retrospective view of FA/C distributed problem solving", IEEE Trans. on Sys. Man and Cyb. SMC-21(6): 1347-1362. (1991).

No context found.

Victor R. Lesser, 1991, A Retrospective View of FA/C Distributed Problem Solving, IEEE Systems, Man and Cybernetics 21(6), 1347 -- 1362.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC