B. Ward. ET-Soar: Toward an ITS for Theory-BasedRepresentations. PhD thesis, School of Computer Science, Carnegie Mellon Univ., 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the interaction to be highly flexible and reactive, as well as real time. Examples of such environments include applications in arenas such as education, entertainment, and training. For instance, in the education arena, intelligent tutoring systems need to interact with students in real time[32]. In the arena of entertainment, recent work has focused on real time, dynamic interactivity among multiple agents within virtual reality environments[5, 12, 17] Similarly, in the arena of training, there is a recent thrust on dynamic, real time interactive simulations[24, 26, 33] In these ....

....in real time, while resolving ambiguities. Section 3 presents an approach to agent tracking that addresses the first item above: tracking flexible and reactive behaviors. The approach is based on the model tracing technique used in intelligent tutoring systems (ITS) for tracking student actions[1, 32]. To track the activities of a student, an ITS executes a model of that student to generate predictions. Tracking proceeds by matching these predictions with actual observations. However, as with plan recognition, previous ITS work has primarily focused on static environments. This article ....

B. Ward. ET-Soar: Toward an ITS for Theory-BasedRepresentations. PhD thesis, School of Computer Science, Carnegie Mellon Univ., 1991.

....chooses arbitrarily between the execution of operator , as it generates its own actions, while also tracking L s actions. The above agent tracking process is related to previous work on model tracing in intelligent tutoring systems (ITS) for tracking student actions [ Anderson et al. 1990; Ward, 1991 ] However, that work has primarily focused on static environments. A recently developed ITS, REACT [ Hill and Johnson, 1994 ] extends model tracing to a more dynamic environment. REACT relies upon a plan driven tracking strategy, and deals with the more dynamic aspects of the domain as special ....

....D s heading, range, geometric relationships etc. Fortunately, many of these quantities are symmetric and can be reused from corresponding quantities in state D . It is difficult to resolve the above ambiguities using methods that have been previously suggested in the model tracing literature. Ward [ Ward, 1991 ] notes that previous model tracing systems have mostly relied on communication with the modeled agent to resolve ambiguities. In air combat simulation, such communication with enemy pilots is clearly ruled out. Ward s solution in the absence of such information is an exhaustive backtrack search ....

B. Ward. ET-Soar: Toward an ITS for Theory-Based Representations. PhD thesis, School of Computer Science, Carnegie Mellon Univ., 1991.

.... environments (Bates, Loyall, Reilly 1992; Hayes Roth, Brownston, Gen 1995) Such real time interaction is also seen in robotic environments (Kuniyoshi et al. 1994) In all these environments, agent tracking is a key capability required for intelligent interaction (Tambe Rosenbloom 1995; Ward 1991; Rao 1994; Anderson et al. 1990) It involves monitoring other agents observable actions and inferring their unobserved actions or high level goals, plans and behaviors. This capability is closely related to plan recognition (Kautz Allen 1986; Song Cohen 1991) which involves recognizing ....

Ward, B. 1991. ET-Soar: Toward an ITS for TheoryBased Representations. Ph.D. Dissertation, School of Computer Science, Carnegie Mellon Univ.

....the interaction to be highly flexible and reactive, as well as real time. Examples of such environments include applications in arenas such as education, entertainment, and training. For instance, in the education arena, intelligent tutoring systems need to interact with students in real time[32]. In the arena of entertainment, recent work has focused on real time, dynamic interactivity among multiple agents within virtual reality environments[5, 12, 17] Similarly, in the arena of training, there is a recent thrust on dynamic, real time interactive simulations[24, 26, 33] In these ....

....in real time, while resolving ambiguities. Section 3 presents an approach to agent tracking that addresses the first item above: tracking flexible and reactive behaviors. The approach is based on the model tracing technique used in intelligent tutoring systems (ITS) for tracking student actions[1, 32]. To track the activities of a student, an ITS executes a model of that student to generate predictions. Tracking proceeds by matching these predictions with actual observations. However, as with plan recognition, previous ITS work has primarily focused on static environments. 2 This article ....

B. Ward. ET-Soar: Toward an ITS for Theory-BasedRepresentations. PhD thesis, School of Computer Science, Carnegie Mellon Univ., 1991.

....between the execution of operator D and operator DL , as it generates its own actions, while also tracking L s actions. The above agent tracking process is related to previous work on model tracing in intelligent tutoring systems (ITS) for tracking student actions [ Anderson et al. 1990; Ward, 1991 ] However, that work has primarily focused on static environments. A recently developed ITS, REACT [ Hill and Johnson, 1994 ] extends model tracing to a more dynamic environment. REACT relies upon a plan driven tracking strategy, and deals with the more dynamic aspects of the domain as special ....

....D s heading, range, geometric relationships etc. Fortunately, many of these quantities are symmetric and can be reused from corresponding quantities in state D . It is difficult to resolve the above ambiguities using methods that have been previously suggested in the model tracing literature. Ward [ Ward, 1991 ] notes that previous model tracing systems have mostly relied on communication with the modeled agent to resolve ambiguities. In air combat simulation, such communication with enemy pilots is clearly ruled out. Ward s solution in the absence of such information is an exhaustive backtrack search ....

B. Ward. ET-Soar: Toward an ITS for Theory-Based Representations. PhD thesis, School of Computer Science, Carnegie Mellon Univ., 1991.

....coupled with many additional (unimplemented) mechanisms for how the symbol representation increases its discrimination, how perception maps the symbolic representation into the spatial world, and so on. EFH Soar contains an initial realization of the hypothesis that extends prior work in this area [15]. Hockey is implemented in an object oriented programming language developed at the Center for Design of Educational Computing at Carnegie Mellon University [2, 12] For each object on the screen, Hockey encodes the coordinates of its position and its type. The objects are the puck, obstacles, ....

Ward, B., ET-Soar: Toward an ITS for Theory-based Representations, Ph.D. dissertation, 1991, School of Computer Science, Carnegie Mellon University.

....interact with each other, either collaboratively or non collaboratively, to achieve their goals. Many of these multi agent domains require the interaction to be dynamic and real time. For instance, in education, intelligent tutoring systems interact with students to provide real time feedback[12, 37]. In entertainment, projects such as interactive fiction[3] virtual immersive environments[21] and virtual theatre[11] all involve real time and dynamic multi agent interaction (collaborative and competitive) Similarly, in training, a recent thrust on dynamic, real time simulations e.g. ....

.... by observation[19] RoboCup robotic soccer tournaments(beginning IJCAI 97) 17] In all these environments, agent tracking monitoring other agents observable actions and inferring their high level goals, plans and behaviors is a central capability required for intelligent interaction[1, 26, 35, 37]. While this capability is obviously essential in noncollaborative settings, it is also important in collaborative settings, where communication may be restricted due to cost, risk, lack of a common protocol etc. 13] The key to this capability is tracking an agent s flexible mix of goal driven ....

[Article contains additional citation context not shown here]

B. Ward. ET-Soar: Toward an ITS for Theory-Based Representations. PhD thesis, School of Computer Science, Carnegie Mellon Univ., 1991.

.... recent work has focused on real time, dynamic interactivity among multiple agents within virtual reality environments [6, 9] Such real time interaction is also seen in robotic environments [11] In all these environments, agent tracking is a key capability required for intelligent interaction [38, 36, 40, 22, 4]. It involves monitoring other agents observable actions and inferring their mental state their goals, beliefs, intentions and tracking this over time. This capability is closely related to plan recognition [10, 31] which involves recognizing agents plans based on observations of their ....

....learning. The limitations of the straightforward approach are identified, followed by the augmentations in DEFT that address these limitations. 3. 1 RESC: An Approach to Real time Dynamic Agent Tracking The RESC (REal time Situated Commitments) approach to agent tracking[38] builds on model tracing[4, 40]. Here, a tracker executes a model of the trackee (the agent being tracked) matching the model s predictions with observations of the trackee s actions. One key innovation in RESC is the use of commitments. In particular, due to ambiguity in trackee s actions, there are often multiple matching ....

B. Ward. ET-Soar: Toward an ITS for Theory-Based Representations. PhD thesis, School of Computer Science, Carnegie Mellon Univ., 1991.

....agent is following a known plan step by step, and have difficulty interpreting deviations from the plan. The modeling process can be underconstrained , postulating mental activities that are difficult to infer from the agent s observable actions. An example of this style of modeling can be seen in (Ward, 1991), where the tutor attempts to track the student by generating production paths that could have led to an observed action. Finally, they tend to be unfocused they do not target their analysis on those situations where tutorial intervention is warranted. For instance, intelligent tutors that use ....

....In cases where the student takes the wrong action, the model tracer must also contain mal rules that would simulate the student s faulty mental model that resulted in the wrong action being taken. Such executable cognitive models can be quite detailed in extreme cases such as the system of (Ward, 1991), internal cognitive processes such as perception are modeled as well. Techniques such as these are inherently underconstrained: they waste effort performing searches that are unproductive, either because they do not lead to a solution, or because the resulting solutions are more detailed than is ....

Ward, B. (1991). ET-Soar: Toward an ITS for theory-based representations. Ph.D. diss., CMUCS -91-146, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA.

.... ] medical diagnosis [ Washington and Rosenbloom, 1989 ] blood analysis [ Johnson et al. 1992 ] production line scheduling [ Hsu et al. 1989 ] chemical process modeling [ Modi and Westerberg, 1989 ] natural language understanding [ Lehman et al. 1991 ] and intelligent tutoring [ Ward, 1991 ] 1986 was also the start of one of the major research directions of Soar. It was during this time that Allen Newell proposed that many of the assumptions embedded in Soar were an appropriate basis for modeling human cognition. Newell s proposal was also based on the assumption that the time ....

.... et al. 1993 ] and video game control [ John et al. 1990 ] Another class of systems where built that interacted with other software systems [ Newell and Steier, 1991 ] including databases, symbolic mathematics packages, chemical process simulators, drawing packages, tutorial environments [ Ward, 1991 ] building design tools [ Steier, 1990 ] and physical world simulators. A full summary of the domains tackled with Soar 5 can be found in the Introduction to [ Rosenbloom et al. 1993 ] 8 Soar 6 (1992) Although Soar 5 broadened the types of domains that Soar could be applied to, it was ....

B. Ward. ET-Soar: Toward an ITS for Theory-Based Representations. PhD thesis, Carnegie Mellon University, 1991. Available as Technical Report CMU-CS-91-146.

....action corroborated. The change in the opponent s heading due to its turn is then noted in state opponent . This technique of event tracking, where an agent models another by pretending to be in that agent s position, has been previously used in automated tutoring systems(Anderson et al. 1990; Ward 1991). These tutoring systems need the ability to model the actions of the students being tutored. For this, these systems use student centered problem spaces in which states and operators model the students under scrutiny. This student modeling technique is referred to as model tracing. The approach ....

....real time tracking. Suggested solutions for addressing ambiguity typically address the representational aspect of this issue, but not the other two. In particular, one method of addressing the ambiguity is to maintain multiple operator hierarchies, so as to track each possibility independently(Ward 1991). Thus, if there is some ambiguity about whether the opponent is actually executing RUN AWAY opponent or EMPLOYMISSILE opponent , then there are two separate operator hierarchies maintained for the two possibilities. If there is further ambiguity in the EMPLOY MISSILE opponent possibility between ....

[Article contains additional citation context not shown here]

Ward, B. 1991. ET-Soar: Toward an ITS for Theory-Based Representations. Ph.D. Dissertation, School of Computer Science, Carnegie Mellon Univ.

....noncollaboratively, to achieve their goals. Many of these multi agent domains are real time and dynamic, requiring the interaction to be highly flexible and reactive. For instance, in the education arena, intelligent tutoring systems need to interact with students while they are solving problems (Ward 1991). In the arena of entertainment, recent work has focused on real time, dynamic interactivity among multiple agents within virtual reality environments (Bates, Loyall, Reilly 1992; Hayes Roth, Brownston, V. 1995) Similarly, in the arena of training, there is a recent thrust on dynamic, ....

....with tens or hundreds of collaborative and non collaborative intelligent agents. Such real time interaction is also seen in robotic environments (Kuniyoshi et al. 1994) In all these environments, agent tracking is a key capability required for intelligent interaction (Tambe Rosenbloom 1995; Ward 1991; Rao 1994) It involves monitoring other agents observable actions and inferring their unobserved actions or high level goals, plans and behaviors. This capability is closely related to plan recognition (Kautz Allen 1986; Azarewicz et al. 1986) which involves recognizing agents plans based ....

Ward, B. 1991. ET-Soar: Toward an ITS for TheoryBased Representations. Ph.D. Dissertation, School of Computer Science, Carnegie Mellon Univ.

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