multi-agent systems (chapter 9) adapted with permission from adina magda florea [email protected]

Click here to load reader

Post on 19-Jan-2016

20 views

Category:

Documents

0 download

Embed Size (px)

DESCRIPTION

Multi-Agent Systems (Chapter 9) Adapted with permission from Adina Magda Florea [email protected] Benevolent vs.. self-interested agents. Benevolent: cooperative distributed systems. (CDPS) Simplifies the task enormously. Self-interested- potential for conflict. Distributed problem solving. - PowerPoint PPT Presentation

TRANSCRIPT

  • Multi-Agent Systems(Chapter 9)

    Adapted with permission from Adina Magda [email protected]

    Slides from Sobah Abbas Peterson

  • Benevolent vs.. self-interested agents

    Benevolent: cooperative distributed systems. (CDPS) Simplifies the task enormously.Self-interested- potential for conflict

    Slides from Sobah Abbas Peterson

  • Distributed problem solvingGroup coherence - agents want to work together - cooperative agentsCompetence - agents must find ways to work together - coordinate to cooperateTask and result sharing - an agent has many tasks to do and asks other agents to do some of its tasks; then it should integrate the resultsDistributed planning - the problem to be solved is to design and execute a plan in a distributed manner, by many agents*

    Slides from Sobah Abbas Peterson

  • Distributed Problem SolvingMotivations:Speed up through parallelizationDistribution of expertiseDistribution of Data, features changeProblem is inherently distributedDistribution of ResultsGeneral StepsTask decompositionTask allocationExchange sub problem solutionsTask accomplishmentResults Synthesis (make whole)

    Slides from Sobah Abbas Peterson

  • Task DecompositionPartitioning of a task into sub-tasks for possible allocation to another agentGoal is to make sub-tasks independent Minimize coordination (so communication costs dont outweigh gain)Minimize shared dataMinimize share resourcesTask decomposition is a hard problem and generally performed a priori by system designers.

    Slides from Sobah Abbas Peterson

  • Task AllocationHomogenous SystemsAgents identical, allocation simple since each is equally qualified to work on sub-tasksHeterogeneous SystemsSub-task requirements - matched to agent skillsPotentially difficult problem (perfect match problem)

    Slides from Sobah Abbas Peterson

  • Which kind of system to build?Homogenous systems are simplerOnly one kind of agent to buildDont have to consider agent skills when distributing sub-tasksHomogenous systems considered unsuitable for complex problemsLow overall utilization of skills and resources

    Slides from Sobah Abbas Peterson

  • Agent Roles in Task AllocationAgents can assume two rolesServers: Agents capable of providing a serviceClients: Agents requiring a service Agents can be both I.e. An agent may use the services of other agents to complete a service is to providing to another agentTask allocation systems must provide a way to match clients with servers

    Slides from Sobah Abbas Peterson

  • Centralized Allocation SystemsCentralized 3rd party manages client-server matchingHierarchical SubordinationSuperior agents order subordinates to carry out task.Typically a static, pre-defined agent organizationEgalitarian - all agents considered equal Requires special broker or trader agents to manage client requests and server bidsAllows centralized allocation techniques

    Slides from Sobah Abbas Peterson

  • Egalitarian Allocation SystemServersTraderClientABCDRequestARequestARejectCAcceptDAcceptD

    Slides from Sobah Abbas Peterson

  • Distributed Allocation SystemsEach agent individually attempts to obtain required servicesAcquaintance NetworkDirect AllocationAgents can only use the services of the agents it knows aboutPotentially serious scalability issuesDelegated AllocationAgents can ask other agents to use their acquaintances to find an agent capable of providing a particular serviceRequires strongly connect acquaintance networkBoth methods require accurate knowledge of agent skillsMay use various caching strategies to maintain and age acquaintance information

    Slides from Sobah Abbas Peterson

  • Distributed Allocation Systems (cont)Contract NetMarket Place approachClients issue description of tasks Servers reply with bidsClient chooses the best bidderServer affirms its commitmentProven approach from other disciplines/simpleWell suited for dynamic environmentsConcurrent and many-to-many nature of the protocol creates challenging race conditions

    Slides from Sobah Abbas Peterson

  • Task Allocation System TradeoffsBenefitsCoherenceDrawbacksBottleneckFault IntoleranceBenefitsNo BottleneckFault toleranceDrawbacksCoherenceScalabilityLatencyBenefitsProven/SimpleFlexibilityDrawbacksMessage volumeTemporal & Spatial IgnoranceDistributedAcquaintanceContract NetCentralizedTrader

    Slides from Sobah Abbas Peterson

  • Types of TasksIndependentTasks are self-containedCan be performed in any order and concurrentlyInterdependentThe solutions of some sub-tasks are required for the solution of other sub-tasksCoordination possible if dependencies known before Possible dependencies only become apparent at runtimeA Results Sharing mechanism is needed to solve these dependencies

    Slides from Sobah Abbas Peterson

  • Motivations for Results SharingConfidence: Independent derivations affirm/challenge previous results leading to more confidenceCompleteness:Combination of partial results leads to a larger set of resultsPrecision:Sharing of results allows for iterative refinement (agents come to see interface)Timeliness:Obvious performance benefits via parallel processing

    Slides from Sobah Abbas Peterson

  • Result SharingProblem solving proceeds by agents cooperatively exchanging information as the solution is developed.Results may be shared:proactively - one agent sends another agent some information because it believes that the other will be interested in it.reactively an agent sends information to another in response to a request.

    Slides from Sobah Abbas Peterson

  • Result Sharing BenefitsConfidence (checking solutions)Completeness/precision: share local viewsTimeliness: may get results faster (even if agent could do it himself)

    Slides from Sobah Abbas Peterson

  • What about inconsistency?Ignore it but are you throwing away the true information (the part that doesnt fit the expectation)?Resolve it through negotiationDegrade gracefullyprogress opportunistically (not in strict predetermined order)communicate high level results, not raw datainconsistency resolved as you go (not at end)no single solution route (if one is problematic, try another)

    Slides from Sobah Abbas Peterson

  • The Coordination ProblemManaging the interdependencies between the activities of agents. e.g.You and I both want to leave the room. We independently walk towards the door, which can only fit one of us. I graciously permit you to leave first.

    Slides from Sobah Abbas Peterson

  • Coordination TechniquesOrganisational StructuresMulti-agent Planning Norms and social laws Coordination Models based on human teamwork:Joint commitments (Jennings)Mutual Modelling

    Slides from Sobah Abbas Peterson

  • Organizational StructuringOrganizes agents into an organizationMay be based on how the task was decomposedAgents use knowledge of the organization toDetermine with whom to communicatePrioritize tasksAgents only need to know about the local organizational structure (coherence)Choosing an organization structure can, itself, be a difficult problem!

    Slides from Sobah Abbas Peterson

  • Organizational StructuringGeographically distributedcells

    Slides from Sobah Abbas Peterson

  • Organizational StructuresA pattern of information and control relationships between individuals. Responsible for shaping the types of interactions among the agents.Aids coordination by specifying which actions an agent will undertake.Organizational structures may be:Functional (based on skills)Spatial (based on physical location)Temporal (based on time relationship)

    Slides from Sobah Abbas Peterson

  • Organizational Structure Models A pattern for decision-making and communication among a set of agents who perform tasks in order to achieve goals. e.g.Automobile industryHas a set of goals: To produce different lines of carsHas a set of agents to perform the tasks: designers, engineers, salesmenReference: Malone 1987

    Slides from Sobah Abbas Peterson

  • Alternative Coordination Structures 1Product Hierarchy

    Slides from Sobah Abbas Peterson

  • Product Manager (several products)Alternative Coordination Structures 2 Functional Hierarchy

    Slides from Sobah Abbas Peterson

  • Alternative Coordination Structures 3 Centralised MarketProduct Manager 2Product Manager 1Product Manager 3FunctionalManagers

    Slides from Sobah Abbas Peterson

  • Alternative Coordination Structures 4 Decentralised MarketProduct Manager 2DesignersSalesmenEngineersProduct Manager 1Product Manager 3

    Slides from Sobah Abbas Peterson

  • Comparison of Organization Structures the Issues!

    ProductioncostCoordinationcostVulnerabilitycostProduct hierarchyHLH-Funtional hierarchyLM-H+CentralisedmarketLM+H-DecentralisedmarketLHL

    Slides from Sobah Abbas Peterson

  • Organizational Structures - CritiqueUseful when there are master/slave relationships in the MAS.Control over the slaves actions mitigates against benefits of DAI such as reliability, concurrency.Presumes that atleast one agent has global overview an unrealistic assumption in MAS.

    Slides from Sobah Abbas Peterson

  • Partial Global Planning (PGP) A DAI testbed Distributed Vehicle Monitoring Testbed (DVMT) to successfully track a number of vehicles that pass within the range of a set of distributed sensors (agents).Each agent monitors a dedicated areaThere could be overlapping areas

    Slides from Sobah Abbas Peterson

  • Partial Global Planning (PGP) Main principle: cooperating agents exchange information in order to reach common conclusions about the problem solving process.Why is planning partial?Th