1. Design of Multi-Agent Systems Teacher Bart Verheij Student assistants Albert Hankel Elske van der Vaart Web site http://www.ai.rug.nl/~verheij/teaching/dmas/ (Nestor contains a link)

2. Overview Introduction Task sharing The Contract Net Result sharing Blackboard systems Subscribe/notify Handling inconsistency Coordination Partial global planning Joint intentions Mutual modelling Norms and social laws

3. Working Together Why and how do agents work together? Important to make a distinction between: – benevolent agents – self-interested agents

4. Benevolent Agents If we “own” the whole system, we can design agents to help each other whenever asked In this case, we can assume agents are benevolent: our best interest is their best interest Problem-solving in benevolent systems is cooperative distributed problem solving (CDPS) Benevolence simplifies the system design task enormously

5. Self-Interested Agents If agents represent individuals or organizations, (the more general case), then we cannot make the benevolence assumption Agents will be assumed to act to further their own interests, possibly at expense of others Potential for conflict May complicate the design task enormously

6. Task Sharing and Result Sharing Two main modes of cooperative problem solving: – task sharing: components of a task are distributed to component agents – result sharing: information (partial results, etc.) is distributed

7. Overview Introduction Task sharing The Contract Net Result sharing Blackboard systems Subscribe/notify Handling inconsistency Coordination Partial global planning Joint intentions Mutual modelling Norms and social laws

8. The Contract Net A well known task-sharing protocol for task allocation is the Contract Net: 1.Recognition 2.Announcement 3.Bidding 4.Awarding 5.Expediting

9. The Contract Net

10. Recognition In this stage, an agent recognizes it has a problem it wants help with. Agent has a goal, and either… – realizes it cannot achieve the goal in isolation — does not have capability – realizes it would prefer not to achieve the goal in isolation (typically because of solution quality, deadline, etc.)

11. Announcement In this stage, the agent with the task sends out an announcement of the task which includes a specification of the task to be achieved Specification must encode: – description of task itself (maybe executable) – any constraints (e.g., deadlines, quality constraints) – meta-task information (e.g., “bids must be submitted by…”) The announcement is then broadcast

12. Bidding Agents that receive the announcement decide for themselves whether they wish to bid for the task Factors: – agent must decide whether it is capable of expediting task – agent must determine quality constraints & price information (if relevant) If they do choose to bid, then they submit a tender

13. Awarding & Expediting Agent that sent task announcement must choose between bids & decide who to “award the contract” to The result of this process is communicated to agents that submitted a bid The successful contractor then expedites the task May involve generating further manager-contractor relationships: sub-contracting

14. Issues for Implementing Contract Net How to… – …specify tasks? – …specify quality of service? – …select between competing offers? – …differentiate between offers based on multiple criteria?

15. Overview Introduction Task sharing The Contract Net Result sharing Blackboard systems Subscribe/notify Handling inconsistency Coordination Partial global planning Joint intentions Mutual modelling Norms and social laws

16. Result Sharing in Blackboard Systems  The first scheme for cooperative problem solving: the blackboard system  Results shared via shared data structure (the blackboard)  Multiple ‘agents’ can read and write to the blackboard  Agents write partial solutions to the blackboard  The blackboard may be structured into hierarchy  Mutual exclusion over the blackboard required  bottleneck  Not concurrent activity

17. Result Sharing in Subscribe/Notify Pattern  Common design pattern in object-oriented systems: subscribe/notify  An object subscribes to another object, saying “tell me when event e happens”  When event e happens, original object is notified  Information pro-actively shared between objects  Objects required to know about the interests of other objects  inform objects when relevant information arises

18. Overview Introduction Task sharing The Contract Net Result sharing Blackboard systems Subscribe/notify Handling inconsistency Coordination Partial global planning Joint intentions Mutual modelling Norms and social laws

19. Handling inconsistency Inconsistencies about: – Beliefs – Goals/intentions Inevitable except in small systems Approaches: – Do not allow or ignore – Resolve by negotiation – Design for graceful degradation (best choice)

20. Handling inconsistency A design for graceful degradation: Functionally Accurate/ Cooperative (FA/C) Systems (Lesser and Corkill) A network problem solving structure: 1.Functionally accurate: “the generation of acceptably accurate solutions without the requirement that all shared intermediate results be correct and consistent” 2.Cooperative: an “iterative, coroutine style of node interaction in the network”

21. Handling inconsistency Characteristics of FA/C systems: – Problem solving not in a strict predetermined order, but opportunistically – Communication of high-level data, not raw data – Solve inconsistency implicitly by exchanging partial results, not at beginning or end -There exist multiple solution routes

22. Overview Introduction Task sharing The Contract Net Result sharing Blackboard systems Subscribe/notify Handling inconsistency Coordination Partial global planning Joint intentions Mutual modelling Norms and social laws

23. Coordination Partial global planning (Durfee) Partial: no plan for the entire problem Global: exchange of local plans for non-local view Stages: – Each agent decides on goals & generate short-term plans – Information on interacting plans and goals is exchanged – Each agent alters local plans Metalevel dictates information exchange (which agents, under which conditions)

24. Coordination Joint intentions (Levesque et al)  Coordinated non-cooperative vs. coordinated cooperative action Going to the same place to find shelter or to perform a choreography  Commitment: a promise  Convention: a means for commitment monitoring  Commitments are assumed to persist: Joint Persistent Goals Agents have a collective commitment to bring about some goal and a motivation for the goal. Agents inform each other about the status of the goal (satisfied/impossible/disappearance of motivation)

25. Coordination Mutual modelling (e.g., Gasser’s MACE) Agent keeps models of itself and other agents, e.g., with respect to roles, skills, goals, plans Characteristic: no explicit communication required

26. Coordination Norms and social laws Important in everyday life Origin of conventions: – Offline design (straightforward) – Emergence (scalable, flexible) Tee shirt game (Shoham & Tennenholtz) How to update strategy? (e.g., majority or highest cumulative reward)

27. Overview Introduction Task sharing The Contract Net Result sharing Blackboard systems Subscribe/notify Handling inconsistency Coordination Partial global planning Joint intentions Mutual modelling Norms and social laws

28. Student presentations Week 40 J. Cassell and Vilhjalmsson (1999). Fully Embodied Conversational Avatars: Making Communicative Behaviors Autonomous. Jeroen van Dijk D.R. Traum (1999). Speech Acts for Dialogue Agents. Hanno Koeslag P.J. Gmytrasiewicz and E.H. Durfee (2001). Rational Communication in Multi-Agent Environments. Marlies de Koning