Milind Tambe and P. S. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In M. Wooldridge, J. P. Muller, and M. Tambe, editors, Intelligent Agents Volume II, Lecture Notes in Artificial Intelligence, pages 156--170. Springer-Verlag, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....chunking is the creation of data structures with special slots. Since some of the slots can themselves contain chunks, there is no theoretical limit on chunk complexity. There are, as yet, no models of multi agent interaction in ACT R. There are several such models implemented in Soar (e.g. 26] [22]) but these have a small number of agents and, if any hierarchical relations, only two layers. For the reasons given in 2.2, none allow for chunking. Those reasons do not apply to SDML models because the greater speed of execution allows for events to recur. However, in the North West Water model, ....

Tambe, M. and P.S. Rosenbloom, (1996) "Architectures for Agents that Track Other Agents in Multi-agent Worlds", Intelligent Agents, II, Springer Verlag Lecture Notes in Artificial Intelligence (LNAI 1037).

....predicting the future, including the behavior of other humans. Similarly, experiments in game theory showed that anticipating the opponent s future moves leads to better results than just reacting to its recent move [9] A lot of work in MAS concentrates on nding out about the opponent s plans [44, 42]. To predict the opponent s behavior, some kind of model of the opponent is needed. Additionally, using models of collaborating agents can in fact turn out to be useful, too, in order to increase team cooperation while saving communication bandwidth. Both cases of models will be referred to as ....

....in this thesis. In domains that are only partially accessible, agents also need to infer unobserved actions from those observations that they were able to make in order to reason successfully about the plans and goals of the other agents. Tambe et al. provide a framework for event tracking [42] that proves successful in military simulations. In such environments it is important to infer unobservable actions like the remote launching of a missile by observable events like turning and radar guidance maneuvers. Tambe uses an operator hierarchy that is implemented in, but not speci c to, ....

Milind Tambe and Paul S. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In Michael Wooldridge, Jorg P. Muller, and Milind Tambe, editors, Proceedings on the IJCAI Workshop on Intelligent Agents II : Agent Theories, Architectures, and Languages, pages 156-170, Heidelberg, Germany, 1996. Springer.

.... same limitations of hierarchical layered intrusion detection which corresponds to gure 1(a) and (b) and gure 4(a) An example from the agent society category that is most relevant to re ective blindness is social diagnosis (Kaminka and Tambe [15] based on agent tracking (Tambe and Rosenbloom [40]) The idea is that agents observe other agents actions and infer their beliefs to compensate for de ciencies in their own sensors. e.g. if an agent is observed to swerve, it can be inferred that something exists 21 which it wishes to avoid, such as a hole in the road. In particular, an agent ....

M. Tambe and P. S. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In Intelligent Agents, Vol. II, LNAI 1037. Springer-Verlag, 1996.

....on the context, the chronicles for handling the temporal aspects, and the game automaton for processing an actually predictive decision making process. One can see it is rather straightforward to integrate those models and make them work together in a smooth way. Our approach can be compared to [13], which is based on task hierarchies (as in static plan recognition approaches like [8] and considers the opponent actions while making dynamic decisions. There the same mechanism employed by the agent in generating its own behaviors is used for tracking others behaviors. This can be quite ....

M. Tambe and P.S. Rosenbloom, `Architectures for agents that track other agents in multi-agent worlds', in Agents, Theories, Architectures, and Languages (ATAL-95), eds., C. Backstrom and E. Sandewall. Springer Verlag Lecture Notes in Artificial Intelligence (LNAI 1037), (1996).

....of highly reactive monitoring: a decision tree for the context, chronicles for the temporal evolutions, and a game automaton for the predictive decision making process. It is straightforward to integrate those models and make them work together in a smooth way. Our approach can be compared to [ Tambe and Rosenbloom, 1996 ] where opponent actions are considered while making dynamic decisions. There a single interpretation of actions and observations is used. Our approach is more general since actions and observations may be of any kind, making it t the more general area of operator artefact reactive loop. In ....

M. Tambe and P.S. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In C. Backstrom and E. Sandewall, editors, Agents, Theories, Architectures, and Languages (ATAL-95). Springer Verlag Lecture Notes in Articial Intelligence 1037, 1996.

....that a class of automata can be learned in polynomial time. Another interesting work on opponent modeling has been presented by Sen and Arora [9] who propose a scheme for learning opponent action probabilities and a maximum expected utility strategy for exploiting weaker opponents. Tambe et al. [11] have proposed an approach for tracking recursive agent models based on a plan recognition task. Gmytrasiewicz [4] has presented the Recursive Modeling Method (RMM) which uses nested models of other agents, combining game theoretic and decision theoretic mechanisms. Suryadi and Gmytrasiewicz [10] ....

M. Tambe and P. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In Intelligent Agents II, Lecture Notes in Artificial Intelligence (LNAI 1037). Springer Verlag, 1996.

.... perform better against them [2] Strategies can be represented in terms of game theory [16] A decision theoretic approach has been also taken in the Recursive Modeling Method [10] Finite automata has been used to represent rationalbounded strategies [12] Some proposals model the others plans [20]. Concerning the model construction method, the simplest case arrives in cooperative settings, where honest agents tell the others what their characteristics are [4] Obviously, this does not work in competitive situations. The most complex situation is when others models are built entirely ....

....other agents are important to consider when choosing an action in an effective manner. A different approach is used by Tambe and Rosenbloom who have presented an agent architecture that conforms the requirements to provide support for flexible and efficient reasoning about other agents models [20]. This work is closely related to plan recognition [11] because the goal is to discover the other agents plans based on their observed actions and execution of models about the others. Nadella and Sen [13] have reported some mechanisms for learning partners and competitors skills in soccer ....

M. Tambe and P. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In Intelligent Agents, volume II of Lecture Notes in Artificial Intelligence (LNAI 1037). Springer Verlag, 1996.

....et al. 1997) show how agents can 2 learn the capabilities of others via repeated interactions, but these agents do not learn to predict what actions other might take. Most of the work in MAS also fails to recognize the possible gains from using explicit agent models to predict agent actions. (Tambe and Rosenbloom, 1996) is an exception and gives another approach for using nested agent models. However, they do not go so far as to try to quantify the advantages of their nested models or show how these could be learned via observations. We believe that our research will bring to the foreground some of the common ....

Tambe, M. and Rosenbloom, P. S. (1996). Architectures for agents that track other agents in multi-agent worlds. In Wooldridge, M., Muller, J. P., and Tambe, M., editors, Intelligent Agents Volume II, Lecture Notes in Artificial Intelligence, pages 156--170. Springer-Verlag.

....agent specialisation and teamwork (e.g. 18] or resource management and load balancing (e.g. 23] although they also mention fault tolerance in a more general sense. The most interesting approach is called social diagnosis (Kaminka and Tambe [16] based on agent tracking (Tambe and Rosenbloom [33]) The idea is that agents observe other agents actions and infer their beliefs to compensate for deficiencies in their own sensors. e.g. if an agent is observed to swerve, it can be inferred that something exists which it wishes to avoid, such as a hole in the road. In particular, an agent may ....

M. Tambe and P.S. Rosenbloom "Architectures for Agents that Track Other Agents in Multi-agent Worlds" in Intelligent Agents, Vol. II Springer-Verlag, 1996 (LNAI 1037)

....differences between the two techniques follow rather 17 directly from differences in the tutoring domains they focus on and the kinds of tutoring interaction that is required in those domains. Our work is somewhat more loosely related to plan recognition (e.g. 13, 5] and agent tracking (e.g. [15]) Plan recognition involves inferring an agent s unknown plan based on their actions. Our pedagogical agent work differs in several ways. The intent is not to infer a plan. We have knowledge of the abstract plan. Rather the issue is evaluating how well the various student behaviors serve the ....

....procedures. This is inconsistent with the nature of the reactive plans we are considering very abstract goal decompositions represented only indirectly by alternative paths through the situation space. Our concern with reactive components in behavior is shared by the agent tracking work (e.g. [15]) which does not assume plans are step by step procedure but rather a mix of plan based and reactive procedures. In contrast to the intent of our work, agent tracking does share with plan recognition the goal of inferring an agent s plan. As a consequence, it assumes access to the information and ....

Tambe, M. & Rosenbloom, P. Architectures for Agents that Track Other Agents in Multi-Agent Worlds. In Wooldridge, M., Muller, J.P. & Tambe, M. (Eds) Intelligent Agents II, Agent Theories, Architecture, and Languages. Berlin: Springer-Verlag, 1996. Pp. 156--170.

....the classical logic (first order) by introducing modal operators for representing the beliefs, goals, and intentions of the agent. Tambe and Rosenbloom have focused their works on agent tracking in a real time, dynamic environment and have examined the implications for agent architectures [TR95a, TR95b] Agent tracking is considered as a capability required for intelligent interaction. Since we are arguing in a previous section that our architecture is intermediary to blackboard model and autonomous agents, we have judged interesting to focus our discussion on the relationship between the work ....

Milind Tambe and Paul S. Rosenbloom. Architectures for Agents that Track Other Agents in Multi-Agent Worlds. In Proceedings of the 1995 Workshop on Agent Theories, Architectures, and Languages, pages 156--170, MontrealCANADA, 1995.

....of the antiair defense domain, refer to the Web page at http : dali:uta:edu=Air:html. 5 Related Work The approaches related to the modeling and coordination in the multiagent environment are in the area of multiagent plan recognition and plan coordination. In work on plan recognition [9] and [12], the objective is to enable an agent to model or recognize the other agents through observation to anticipate the other agents future action, given prior knowledge about these agents. In these approaches, an agent usually compares a pre calculated plan, or a protocol, with the on going ....

....with the on going situation, and then chooses his next action accordingly. The process of mental state recognition [9] assumes the correct and complete knowledge of the plans of the other agents, and it does not represent uncertainty that might be present in real world domains. As Tambe et al. [12] pointed out, ambiguities may persist when an agent must infer unobserved actions and intentions. The multiagent coordination without communication has been dealt with in [11] and [5] Sen, in [11] used a particular reinforcement learning methodology and concentrated on the learning classifier ....

M. Tambe and P. S. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In Intelligent Agents II: Agent Theories, Architectures, and Languages, pages 156--170. Springer, 1996.

....coordination with the human subjects, and not only among RMM agents. 5 Related Work The approaches related to the modeling and coordination in the multiagent environment are in the area of multiagent plan recognition and plan coordination. In work on plan recognition (Rao and Murray, 1994, and Tambe and Rosenbloom, 1996), the objective is to enable an agent to model or recognize the other agents through observation to anticipate the other agents future action, given prior knowledge about these agents. In these approaches, an agent usually compares a pre calculated plan, or a protocol, with the on going ....

....the on going situation, and then chooses his next action accordingly. The process of mental state recognition (Rao and Murray, 1994) assumes the correct and complete knowledge of the plans of the other agents, and it does not represent uncertainty that might be present in real world domains. As Tambe and Rosenbloom (1996) pointed out, ambiguities may persist when an agent must infer unobserved actions and intentions. Multiagent coordination without communication has been dealt with in Sen and Sekaran (1996) and Mor et al. 1996) Sen and Sekaran (1996) used a particular reinforcement learning methodology and ....

Tambe, M., and Rosenbloom, P. S. (1996). Architectures for agents that track other agents in multi-agent worlds. In Wooldridge, M., Muller, J. P., and Tambe, M., eds., Intelligent Agents II: Agent Theories, Architectures, and Languages, 156--170. Berlin: Springer.

....small number of human participants. We will conduct more exhaustive experiments in future work. 6. Related work The approaches related to the modeling and coordination in the multiagent environment are in the area of multiagent plan recognition and plan coordination. In work on plan recognition [13, 17], the objective is to enable an agent to model or recognize the other agents through observation to anticipate the other agents future action, given prior knowledge about these agents. In these approaches, an agent usually compares a precalculated plan, or a protocol, with the on going situation, ....

....the on going situation, and then chooses his next action accordingly. The process of mental state recognition [13] assumes the correct and complete knowledge of the plans of the other agents, and it does not represent uncertainty that might be present in real world domains. As Tambe and Rosenbloom [17] pointed out, ambiguities may persist when an agent must infer unobserved actions and intentions. Multiagent coordination without communication has been dealt with in [15] and [8] Sen and Sekaran [15] used a particular reinforcement learning methodology and concentrated on the learning ....

M. Tambe and P. S. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In M. Wooldridge, J. P. Muller, and M. Tambe, editors, Intelligent Agents II: Agent Theories, Architectures, and Languages, pages 156--170, Berlin, 1996. Springer.

....the classical logic (first order) by introducing modal operators for representing the beliefs, goals, and intentions of the agent. Tambe and Rosenbloom have focused their works on agent tracking in a real time, dynamic environment and have examined the implications for agent architectures [TR95a] TR95b] Agent tracking is considered as a capability required for intelligent interaction. 3 Glimpse of MAGIQUE Specialist agent is cognitive because it is endowed with capabilities for perception, reasoning and planning. The specialist receive knowledge coming from either its group supervisor or its ....

Milind Tambe and Paul S. Rosenbloom. Architectures for Agents that Track Other Agents in Multi-Agent Worlds. In Proceedings of the 1995 Workshop on Agent Theories, Architectures, and Languages, pages 156--170, Montreal-CANADA, 1995.

....step 1 to step 2 requires generalized methods for blame assignment. In standard discrimination based learning, the next step would be essentially step 8, that is to identify a previously successful agent tracking episode and compare with the current failure episode. 3 For interested readers, [39] provides details; specifically, RESC is based on a modified version of the Soar architecture, where these architectural modifications are included in the form of Soar rules. CRITICAL FAILURE TRIGGERS CORRECTIVE ACTIONS DETERMINE IF AGENT TRACKING INACCURACY CAUSED FAILURE. IF SO, PROCEED. ....

M. Tambe and P. S. Rosenbloom. Architectures for agents that track other agents in multiagent worlds. In Intelligent Agents, Volume II: Lecture Notes in Artificial Intelligence 1037. Springer-Verlag, Heidelberg, Germany, 1996.

No context found.

Milind Tambe and P. S. Rosenbloom. Architectures for agents that track other agents in multi-agent worlds. In M. Wooldridge, J. P. Muller, and M. Tambe, editors, Intelligent Agents Volume II, Lecture Notes in Artificial Intelligence, pages 156--170. Springer-Verlag, 1996.

No context found.

Tambe, M. and P.S. Rosenbloom, (1996) "Architectures for Agents that Track Other Agents in Multi-agent Worlds", Intelligent Agents, II, Springer Verlag Lecture Notes in Artificial Intelligence (LNAI 1037).

No context found.

Tambe, M. & Rosenbloom, P. Architectures for Agents that Track Other Agents in Multi-AgentWorlds. In Wooldridge, M., Muller, J.P.&Tambe, M.#Eds# Intell. Agents II, Agent Theories, Arch., and Lang. Berlin:Springer-Verlag, 1996. 156#170.

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