Georgeff, M., and Lansky, A. L. 1986. Procedural knowledge. Proceedings of IEEE 74(10):1383--1398.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....if it extends BDIGoalEvent. 3. plan 2 is considered and executed, printing plan2 (and then failing) 4. plan 3 is considered and executed, printing plan3 5. plan 4 is considered and executed, printing plan4 The BDIGoalEvents provide behaviour which is standard in BDI systems such as PRS [8], JAM [9] dMars [10] and others. InferenceGoalEvents however provide a functionality more directed toward reasoning than acting. All relevant plans (or rules) are executed. This gives a behaviour similar to expert systems and is particularly useful for some of the inferencing needed in this ....

Georgeff, M.P., Lansky, A.L.: Procedural knowledge. Proceedings of the IEEE Special Issue on Knowledge Representation 74 (1986) 1383--1398

.... Parameterized Scripts for Motion Planning Patrick Rowe and Anthony Stentz Robotics Institute Carnegie Mellon University, Pittsburgh, PA 15213 Abstract We present an approach for real time planning and execution of the motions of complicated robotic systems. The approach is motivated by the observation that a robot s task can be described as a series of simple steps, or a script. The script is a general template which encodes ....

....world, however their use of parameters is different from the approach presented here, where parameters are calculated for the purpose of controlling a robot. The idea of using procedures to reason about and construct plans has been presented in planners such as the Pro cedural Reasoning System [5] The approach here, however, does not exist at that level of planning which aims to synthesize entire plans for performing a complex task given a clever representation of the world and world states [6] One level down are entities such as SAUSAGES [7] which is described as existing between ....

[Article contains additional citation context not shown here]

Georgeff M., Lansky A., 1986. Procedural Knowledge. in Proc. IEEE Special Issue on Knowledge Representation, pp. 1383-1398.

....cycles Fig. 6. Dynamic agent behavior in one vs two air combat simulation for 100 cycles. 7 Related Work There are at least three main areas of related work. The first one is agent architectures. In this area, Rao s work on reactive plan recognition[20] based on the PRS dMARS architecture[10] is closest in spirit to the work reported here. The underlying concept is to extend agent architectures, specifically the PRS architecture, to enable it to execute models of other agents in service of reactive recognition. The resulting architecture can execute an agent s own behavior while ....

M. P. Georgeff and A. L. Lansky. Procedural knowledge. Proceedings of the IEEE special issue on knowledge representation, 74:1383--1398, 1986.

.... ta4 child scriptsar not fired concurrently) we need not consider mutua exclusion whenaenJ3( 3 script va4F] fiJ ina pazfi t script (context) A scriptca be viewed a procedura knowledge fora arJ t toa3( a certa[ goaa In this sense, AgenTae ha much in common with Procedura Rea soning System (PRS) [17]. AgenTae describes the protocolalJ3 alJ ts,wherea PRS describes the intera4P) fi betweena anJ ta4 a environment. 4 Verification of AgenTalk Functions In order to verifytha AgenTae ha enough functions, the contra3 net protocol al the multistax negotia4[F protocol al described in 4.1 Contract ....

George#, M. P. and Lansky, A. L.: Procedural Knowledge, Proceedings of the IEEE, Vol. 74, No. 10, pp. 1383--1398 (1986).

....interpretations in 2 D (right hand window) or in 3 D (left hand window) zoom in or out as desired, and rotate the volume to obtain the best view. IV. AGENT ARCHITECTURE All SurfaceMapper agents have the same architecture. This architecture is a belief desire intention (BDI) architecture [18] [19] and is characterized by having the mentalistic notions of beliefs (what the agent knows about its environment) desires (states the agent wishes to achieve) and intentions (selected courses of action) at its core. Such an architecture was chosen because it has a proven track record of working ....

M. P. Georgeff and A. L. Lansky, "Procedural knowledge," Proc. IEEE Special Issue Knowl. Representation, vol. 74, pp. 1383--1398 Dec. 1986.

....act in response to their environment, but should be able to exhibit opportunistic, goal directed behavior and take the initiative. Thirdly, they should be able to solve their tasks efficiently and must often satisfy real time constraints. This requires access to a set of hard wired procedures [18] with guaranteed execution properties. Fourthly, they are to cope with the presence of other agents. Whereas certain types of interactions often can be performed by employing local mechanisms (i.e. obstacle or collision avoidance in a robot scenario [30] 34] others (i.e. collaboration) ....

M. P. Georgeff and A. L. Lansky, "Procedural knowledge," in Proc. IEEE Knowledge Representation, vol. 74, pp. 1383--1398, 1986.

....work in this area focuses on the theoretical foundations, without providing practical implementations. Moreover, it is based on rather di erent technical assumptions on actions and on the interaction with the environment. Several reactive planners (e.g. Beetz McDermott. 1994; Firby, 1987; George Lansky, 1986; Simmons, 1990 ] have been designed to deal with non determinism at execution time. They are able to execute reactive plans encoding conditional and iterative trial and error strategies. However, none of these systems can generate automatically reactive plans as we do in the work presented in ....

M. George and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383-1398, 1986.

.... on the standard BDI concepts though we have found it necessary to clarify some of the differences between just what these concepts are in the initial philosophical work of Bratman [2] the logical theories of Rao and Georgeff [17] and others (e.g. 4, 21] and the implementations such as PRS [9, 10], dMars [1, 6] and JACK [3] We have also found it important to place some emphasis on the concepts of percepts and actions which appear in many generic models of agents (e.g. 20] and which are very important in the interfacing of the agent deliberation to the external environment. We have found ....

M. P. Georgeff and A. L. Lansky. Procedural knowledge. Proceedings of the IEEE Special Issue on Knowledge Representation, 74:1383--1398, 1986.

.... probabilistic context sensitive grammars (PCSGs) might overcome this problem, but it is difficult to define a probability distribution for a PCSG [15] Huber, et al. 11] present an approach to keyhole plan recognition for coordinating teams of agents based on the Procedural Reasoning System(PRS)[7, 12]. PRS is a planning architecture that uses hierarchical plan specifications very similar to our plan library and a reactive execution engine to allow the system designer to build agents that follow the specified plans. Huber s algorithm automatically generates plan recognition belief networks ....

M. Georgeff and A. Lansky, "Procedural Knowledge," IEEE Special Issue on Knowledge Representation, vol. 74, pp. 1383--1398, October 1986.

....to an external event. The former case is the realm of classical planning; the latter that of reactive planning. According to its aim, Yang s book focuses on how to achieve goals. In general, a goal is a condition on intermediate and final states of actions, i.e. a set of desired behaviours (e. g [17]) Nevertheless, in classical planning two simplifying assumptions are usually made, namely, that a goal is a condition on final states (see for example [6,15] and that this condition is a conjunction of clauses (see for example [11,15] Both these assumptions are also made in Yang s book 4 . ....

M. Georgeff and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383--1398, 1986.

....internal planning activity, e.g. plan generation or plan execution, orelse can represent recovery from failure by replanning (like in most classical planners, e. g [20] as well as recovery from failure by executing special purpose exception handling routines (like in most embedded planners, e.g. [7, 13]) The construct ; resembles the sequential composition of Dynamic Logic. The second action is executed anyway, independently of the failure success of the first action. repeat is recursively defined. It repeats the execution of ff till ff fails. If ff never fails, execution does not terminate. ....

....the real world applications we are interested in. A lot of recent research in planning is more and more focusing on real world applications where failure is a key aspect. This is the case of the research on embedded planners . Several embedded planners have been proposed so far (see for instance [7, 13, 2, 4, 15]) and have been successfully applied in particular application domains (like mobile robots [6, 14, 3] and fault diagnosis for real time systems [7] Most of them provide different and flexible failure handling mechanisms, see for instance [6, 13] Nevertheless, so far this research has focused on ....

[Article contains additional citation context not shown here]

M. Georgeff and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383--1398, 1986. 13

....actions, theory of actions, reactive systems. 1 Introduction A lot of recent research in planning is more and more focusing on reactive planning systems, i.e. systems which are able to plan and control execution of plans in a partially known and unpredictable environment (see for instance [5, 14, 17, 18, 41]) While formalizations of classical planners have been proposed (see for instance [32] 37] and [27] this is not the case for reactive planning systems. There actually seems to be a big gap between the approaches followed and the issues faced in the implementation of reactive planning systems ....

.... plan from scratch by searching through the space of states or partial plans (like in classical planners, e.g. 13, 37] which implements probabilistic reasoning (such as in [14] or which implements a simple look up table which given a goal returns a pre compiled plan immediately (such as in [17]) As a consequence, the notion of plan generation action is general enough for a variety of possibly different planning mechanisms. 7 In principle, we might have different ways to execute plans, i.e we might have different plan execution actions. We limit to the case in which we have only one ....

[Article contains additional citation context not shown here]

M. Georgeff and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383--1398, 1986.

....precompiled plans or reflexes, one for each kind of stimuli from sensors. These systems do not have reasoning capabilities, even if reasoning can be sometimes useful. Finally, several systems (called situated systems) trying to integrate reasoning and reacting capabilities have been proposed [ Georgeff and Lansky, 1986; Simmons, 1990; Firby, 1992; Beetz and McDermott, 1994; Ghallab and Laruelle, 1994 ] Any of the previous solutions can be applied to certain classes of problems. For instance, most real world systems need to generate a plan to anticipate predictable events and situations and then to execute the ....

....Formally, we have a set W which is the (possibly infinite) set of states. A finite sequence of states, repetitions allowed, is called behaviour (see figure 1) It represents a possible evolution of the system, i.e. how the system changes while it is executing an action (see for example [ Georgeff and Lansky, 1986 ] As a matter of fact, during the execution of an action a system transits from a state to another and so on, until the action ends. Formally the set of all the possible behaviours is B = W . The concatenation of two behaviours is defined as follows: if w i 2 W, i = 1; j Gamma 1; ....

[Article contains additional citation context not shown here]

M. Georgeff and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383--1398, 1986.

.... of established paths between points in a rover domain, where thinner edges are harder to traverse, and labeled points have associated observation goals While planning and scheduling efficiency is a major focus of our research, another is the support of flexible plan execution systems such as PRS [6], UMPRS [11] RAPS [5] JAM [7] etc. that exploit hierarchical plan spaces while interleaving task decomposition with execution. By postponing task decomposition, such systems gain flexibility to choose decompositions that best match current circumstances. However, this means that refinement ....

M. Georgeff and A. Lansky. Procedural knowledge. Proc. IEEE, 74(10):1383--1398, Oct. 1986.

.... of established paths between points in a rover domain, where thinner edges are harder to traverse, and labeled points have associated observation goals While planning and scheduling efficiency is a major focus of our research, another is the support of flexible plan execution systems such as PRS [6], UMPRS [11] RAPS [5] JAM [7] etc. that exploit hierarchical plan spaces while interleaving task decomposition with execution. By postponing task decomposition, such systems gain flexibility to choose decompositions that best match current circumstances. However, this means that refinement ....

M. Georgeff and A. Lansky. Procedural knowledge. Proc. IEEE, 74(10):1383--1398, Oct. 1986.

....replaced by blackboard architectures[Hayes Roth, 1985] which accrue data about the world and make immediate decisions based on both variable a priori goals and a changing world. Architectures have been proposed and implemented which allow for seamless changing of goals and error recovery[Georgeff and Lansky, 1986]. There are architectures which learn, i.e. which adapt their internal models in a variety of ways such as chunking [Laird et al. 1987] or case based reasoning[Carbonell and Veloso, 1988] All of these architectures still share a common view expressed explicitly in some of the early papers ....

....link s next function report status back to an external planner and request that the planner generate the next link. This approach, which totally interleaves planning and execution, can be used to interface a SAUSAGES system with incremental planners such as the Procedural Reasoning System (PRS) [Georgeff and Lansky, 1986] which may use the universal planner approach [Schoppers, 1987] At each step in a plan an incremental planner examines everything it knows about the situation before applying the next action. This planning paradigm lets the system easily respond to errors, system failures, goal changes, and ....

Georgeff, M. P. and Lansky, A. L. (1986). Procedural knowledge. In Proc. IEEE Special Issue on Knowledge Representation, pages 1383--1398.

....of incremental interpretation by using it in the Candide system for knowledge acquisition. Incorporating both a graphical interface and a processor for English discourse, Candide is designed to be used in the construction and maintenance of knowledge bases for the Procedural Reasoning System (PRS) [12]. In Candide, a procedure is described by a network whose nodes represent states in the procedure s execution and whose arcs represent state transitions described by suitable arc annotations. Each procedural network also contains invocation conditions, which encode [pre]conditions that must be ....

M. P. Georgeff and A. L. Lansky. Procedural knowledge. Proceedings of the IEEE, Special Issue on Knowledge Representation, pages 1383--1398, 1986.

....RAPS provides a tighter level of structuring, which simplifies tracing and potentially generation of action networks from planners. However, ESL provides better support for structured objects and language extension, both of which were crucial for our large software engineering project. PRS (George# Lansky 1987) and RPL (McDermott 1991) are similar languages that support general execution and agent construction. Interrap (Muller Pischel 1994) and Golog (Levesque et al. 1997) provide execution languages based on logic programming. Unlike most other general execution systems, TCA (Simmons 1990) ....

George#, M. P., and Lansky, A. L. 1987. Procedural knowledge. Technical Report 411, Artificial Intelligence Center, SRI International.

....the steps of the procedure, and an invocation condition, which specifies under which situations the KA is useful. Together, the invocation condition and body of a KA express a declarative fact about the results and utility of performing certain sequences of actions under certain conditions [10]. Figure 3: A Simple Meta Level KA The set of KAs in a PRS application system not only consists of procedural knowledge about a specific domain, but also includes meta level KAs (See Figure 3) that is, KAs able to manipulate other KAs, beliefs, goals, and 9 intentions of PRS itself. The use ....

M. P. Georgeff and A. L. Lansky. Procedural Knowledge. Proceedings of the IEEE Special Issue on Knowledge Representation, 74:1383--1398, 1986.

....reason, the fundamental assumption underlying classical planning (see e.g. Fikes Nilsson 1971; Penberthy Weld 1992) to consider only deterministic domains appears to be too restrictive in many practical cases. There are several reactive planners (see e.g. Beetz McDermott 1994; Firby 1987; Georgeff Lansky 1986; Simmons 1990) which relax this assumption, and deal with non determinism at execution time, i.e. are able to execute an action and test the many different possible outcomes and are able to execute iterative plans implementing trial and error strategies. However, the ability of generating ....

Georgeff, M., and Lansky, A. L. 1986. Procedural knowledge. Proc. of IEEE 74(10):1383--1398.

....what we refer to as the context hierarchy. A mechanism is required to resolvemultiple and possibly con icting goals. We have developed a planner, GRASP (General Reasoning using AbStract Physics) that does just that (see [3] for details) GRASP is a least commitment partial hierarchical planner [12]. Such planners are particularly well suited to continuous and unpredictable domains such as CtF, where the planning space branching factor can be very high. Partial hierarchical planners rely on a library of plan skeletons. Plan skeletons are plans that are not fully elaborated: they may contain ....

M. P. George and A. L. Lansky. Procedural knowledge. Proceedings of the IEEE Special Issue on Knowledge Representation, 74(10):1383-1398, 1986.

....of the potential needs and effects of all of its potential refinements. Our motivation for doing this is not simply to make interleaved planning and merging with HTNs more efficient, but also to support another crucial use of HTN concepts specifically, flexible plan execution systems such as PRS (Georgeff Lansky 1986), RAPS (Firby 1989) etc. that similarly exploit hierarchical plan spaces. Rather than refine abstract plan operators into a detailed end to end plan, however, these systems interleave refinement with execution. By postponing refinement until absolutely necessary, such systems leave themselves ....

Georgeff, M. P., and Lansky, A. 1986. Procedural knowledge. Proc. IEEE 74(10):1383--1398.

....for plan search, interactive systems for plan reuse) and execute plans. A planning system has to activate, coordinate and control all these devices and modules. Several systems which control acting, sensing and planning have been proposed so far (see for instance [ Beetz and McDermott, 1994; Georgeff and Lansky, 1986; Simmons, 1990 ] and have been successfully applied in particular application domains (like mobile robots and fault diagnosis for real time systems) In spite of this fact, no principled and theoretical account has been given of the behaviours of these systems. The goal of this paper is to ....

....exec( 1 ) 6) Since plan execution actions may perform actions in the real world, they may fail. Failure in plan execution can be handled in different ways. Most classical planners (e.g. Wilkins, 1985 ] handle failure by replanning, i.e. by searching for a new plan. Reactive planners (e.g. Georgeff and Lansky, 1986; Simmons, 1990 ] sometimes have no time for replanning, and therefore handle failure by executing precompiled special purpose exception handling routines. Our logic is expressive enough to represent these different failure handling mechanisms. For instance, tactic (7) reacts to failure of exec( ....

[Article contains additional citation context not shown here]

M. Georgeff and A. L. Lansky. Procedural knowledge. Proceedings IEEE, 74(10):1383--1398, 1986.

....language. In MRG, failure capturing is explicitly represented by the construct orelse. Recovery from failure is uniformly represented by tactics. This allows the user to define plans and failure handling mechanisms with the same language and to flexibly modify them. In PRS [ Georgeff, 1989; Georgeff and Lansky, 1987a; Georgeff, 1991; Georgeff and Lansky, 1987b ] plans (called KAs) describe how certain sequences of actions and tests may be performed to achieve given goals or to react to particular situations. Metalevel KAs encode various methods for chosing among multiple applicable KAs. They provide a high ....

....explicitly represented by the construct orelse. Recovery from failure is uniformly represented by tactics. This allows the user to define plans and failure handling mechanisms with the same language and to flexibly modify them. In PRS [ Georgeff, 1989; Georgeff and Lansky, 1987a; Georgeff, 1991; Georgeff and Lansky, 1987b ] plans (called KAs) describe how certain sequences of actions and tests may be performed to achieve given goals or to react to particular situations. Metalevel KAs encode various methods for chosing among multiple applicable KAs. They provide a high amount of flexibility in forming plans. The ....

M. P. Georgeff and A. L. Lansky. Procedural Knowledge. Technical Report 411, A.I. Center, SRI International, Menlo Park, California, 1987.

....planning, acting and sensing of the external environment, to build a plan that can be executed according to the perceived changes. Looking at the literature, reactive planners are planning systems which are able to plan by changing promptly focus and goals if the external environment requires it [10] and which are able to perform run time decision making [9] From these definitions, one may think about reactive planners as a kind of intelligent planners, however, what really changes in this systems is their behaviour: planning, sensing and execution are strictly interconnected and ....

Georgeff M and Lansky L. Procedural Knowledge. Proceedings of IEEE, 74(10), 1986, 1383-1398.

....the potential needs and effects of all of its potential refinements. Our motivation for doing this is not simply to make interleaved planning and merging with HTNs more efficient, but also to support another crucial use of HTN concepts specifically, flexible plan execution systems such as PRS (Georgeff Lansky 1986), RAPS (Firby 1989) etc. that similarly exploit hierarchical plan spaces. Rather than refine abstract plan operators into a detailed end to end plan, however, these systems interleave refinement with execution. By postponing refinement until absolutely necessary, such systems leave themselves ....

Georgeff, M. P., and Lansky, A. 1986. Procedural knowledge. Proc. IEEE 74(10):1383--1398.

....has to deal continuosly with unexpected external events that make planned actions fail. Furthermore, embedded planners have to perform different failure recovering mechanisms depending on the different situations of the external environment. Several embedded planners have been proposed so far [BM92, Fir92, GL86, GTCS91, Sim90, SCT92] and have been successfully applied in particular application domains (like mobile robots [CTDA92, Gat92, GLS86, Sim91] and fault diagnosis for real time systems [GL86] Nevertheless, so far most of the work on embedded planning has concentrated in domain specific applications and in system ....

....situations of the external environment. Several embedded planners have been proposed so far [BM92, Fir92, GL86, GTCS91, Sim90, SCT92] and have been successfully applied in particular application domains (like mobile robots [CTDA92, Gat92, GLS86, Sim91] and fault diagnosis for real time systems [GL86]) Nevertheless, so far most of the work on embedded planning has concentrated in domain specific applications and in system design and architectures. As a consequence, current architectures are far for providing a general framework where all the required foundamental features are available and ....

[Article contains additional citation context not shown here]

M. Georgeff and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383--1398, 1986.

....When modeling human reasoning it is advantageous to make use of a paradigm and a development environment that has its origins firmly rooted in the research of intentionality and practical reasoning [3] That dMARS (a BDI system) is well suited to this task is no accident. Work by Rao and Georgeff [6], 16] bound the science of intentional systems to the practical application of dMARS in a way that aids the modeling of human reasoning. Pilots and controllers who regularly interact with AOD simulations report that the notion of a single agent mapping to a single human is an intuitive one. To ....

M. P. Georgeff and A. L. Lansky. Procedural Knowledge. In Proceedings of the IEEE Special Issue on Knowledge Representation, volume 74, pages 1383-1398, 1986

....of deriving and using summary information to show that this result also applies to the planning problem. Another advantage of finding abstract solutions is that the agent(s) retain flexibility in their plans by preserving further refinement choices. Robust plan execution systems such as PRS (Georgeff Lansky 1986), RAPS (Firby 1989) and JAM (Huber 1999) can then be used to interleave refinement with execution and provide some amount of recovery from unexpected events. Coordinating abstract plans involves reasoning about the concurrent execution of actions, but even for the planning problem, it is ....

Georgeff, M. P., and Lansky, A. 1986. Procedural knowledge. Proc. IEEE 74(10):1383--1398.

....of different roll pitch yaw states of the aeroplane and different actions on the control yoke. This extends work on behavioural cloning in flight simulators [8] and is currently being used for binding trajectory information to symbolic belief, desires and intentionality planning systems [3]. In hand drawn schematic interpretation, diagrams are defined by different strokes or gestures, drawn at different times at particular orientations. We have developed a software package the Consolidated Learning Algorithm (CLARET) 4] which is based on Relational Evidence Theory [5] The ....

Georgeff, M. P., Lannsky, A.: Procedural Knowledge. Proceedings of the IEEE (Special Issue on Knowledge Representation) 74 (1986 1383--1398

....may be partially specified. For this reason, the fundamental assumption underlying classical planning (see e.g. Fikes Nilsson 1971; Penberthy Weld 1992) to consider only deterministic domains appears to be too restrictive in many practical cases. There are several reactive planners (see e.g. (Georgeff Lansky 1986)) which relax this assumption, and deal with non determinism at execution time, i.e. are able to test the different possible outcomes of the executed action and are able to execute iterative plans implementing trial and error strategies. Copyright 1998, American Association for Artificial ....

Georgeff, M., and Lansky, A. L. 1986. Procedural knowledge. Proc. of IEEE 74(10):1383--1398.

....act in response to their environment, but also be able to exhibit opportunistic goal directed behavior and take the initiative. Thirdly, they should be able to solve their tasks eciently and often have to satisfy real time constraints. This requires access to a set of hardwired procedures [18] with guaranteed execution properties. Fourthly, they are to cope with the presence of other agents. Whereas certain types of interactions can often be performed by employing local mechanisms (e.g. obstacle or collision avoidance in a robot scenario, see [30] 34] others (e.g. collaboration) ....

M. P. George, and A. L. Lansky, \Procedural knowledge," in Proc. of the IEEE Special Issue on Knowledge Representation, 74, pp. 1383-1398, 1986.

....deciding on one way is the purpose of negotiation. Just as agents can enhance planning efficiency by exploiting the hierarchical structure of planning operations, they can similarly enhance the efficiency and quality of coordinating their plans. Flexible plan execution systems such as PRS [10], interleave plan execution with the refinement of abstract plans into specific actions. By postponing refinement until absolutely necessary, such systems leave themselves flexibility to choose refinements that best match the current circumstances. However, when sharing resources with other ....

Georgeff, M.P. and Lansky, A. Procedural knowledge, Proceedings of IEEE 74(10):1383-1398.

....plans. Introduction A lot of recent research in planning is more and more focusing on reactive planning systems, i.e. planning systems which are able to plan and control execution of plans in a partially known and unpredictable environment (see for instance (Beetz and McDermott 1994; Firby 1987; Georgeff and Lansky 1986; Simmons 1990) While formalizations of classical planners have been proposed (see for instance (Lifschitz 1986) for STRIPS (Fikes and Nilsson 1971) this is not the case for reactive planning systems. There actually seems to be a big gap between the approaches followed and the issues faced in ....

M. Georgeff and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383--1398, 1986.

....assumption about the world being modeled, namely that there is only one performing agent. Hence, only one action may happen at a time. As a result, these formalisms do not support simultaneous actions, a major drawback for representing actions in multi agent domains. Georgeff and Lansky (Georgeff and Lansky, 1986; Lansky, 1987) address the limitations of state based formalisms, in particular with respect to simultaneity, but the alternative representations they provide have other restrictions. In particular, Georgeff and Lansky s Proce11 This view of recipes is similar to that described in more recent ....

....by Pollack (Konolige and Pollack, 1989) In that formalism, recipes, which they call plan fragments , are built using the operators BY and TO, which resemble, respectively, the cgen and result in relations. dural Reasoning System has only been used to model multiple, concurrently active actions (Georgeff and Lansky, 1986), and the simultaneity relation in Lansky s GEM concurrency (Lansky, 1987) model does not distinguish between co temporal actions (e.g. two agents lifting a piano together) and actions forming a generational pair. Furthermore, this model lacks a clear notion of what an agent is, making it ....

Georgeff, Michael P. and Amy L. Lansky. 1986. Procedural knowledge. Proceedings of the IEEE, Special Issue on Knowledge Representation, 74(10):1383--1398.

....cycles Fig. 6. Dynamic agent behavior in one vs two air combat simulation for 100 cycles. 7 Related Work There are at least three main areas of related work. The first one is agent architectures. In this area, Rao s work on reactive plan recognition[20] based on the PRS dMARS architecture[10] is closest in spirit to the work reported here. The underlying concept is to extend agent architectures, specifically the PRS architecture, to enable it to execute models of other agents in service of reactive recognition. The resulting architecture can execute an agent s own behavior while ....

M. P. Georgeff and A. L. Lansky. Procedural knowledge. Proceedings of the IEEE special issue on knowledge representation, 74:1383--1398, 1986.

....items relevant to the receiver s capabilities. This provides a simplified or black box view of possibly quite detailed instructions needed to actually perform the activity (possibly involving iterators and conditionals, etc) Complex execution agent representational and programming languages [11], 13] could be handled by using this abstracted view. For example, reliable task achieving behaviours which included contingencies and safe state paths to deal with unforeseen events could be hidden from the planner by communication in terms of a simplified and more robust model of the execution ....

....system) agents. The actual plan state inside the task assignment and execution system agents is likely to differ from that within the planner. For example, the execution system may be based on more procedural representations as are found in languages like prs (the Procedural Reasoning System [11]) and may allow iteration, conditionals, etc. 3 PLANIT Interactive Planner s Aid The uk Alvey Programme s planit Community Club produced a prototype planners aid called the Interactive Planners Assistant (ipa) during 1986 7 [8] 15] In the planit ipa, rich plan representations were used ....

Georgeff, M.P. and Lansky, A.L., Procedural Knowledge, in Proceedings of the IEEE, Special Issue on Knowledge Representation, Vol. 74, pp 1383-1398, 1986.

....steps of the procedure, and an invocation condition, which specifies in what situations the KA is useful and applicable. Together, the invocation condition and body of a KA express a declarative fact about the results and utility of performing certain sequences of actions under certain conditions [ 3 ] . The body of a KA can be viewed as a plan or plan schema. It is represented as a graph with one distinguished start node and possibly multiple end nodes. The arcs in the graph are labeled with the subgoals to be achieved in carrying out the plan. Successful execution of a KA consists of ....

M. P. Georgeff and A. L. Lansky. Procedural knowledge. In Proceedings of the IEEE Special Issue on Knowledge Representation, volume 74, pages 1383--1398, 1986.

No context found.

Georgeff, M., and Lansky, A. L. 1986. Procedural knowledge. Proceedings of IEEE 74(10):1383--1398.

No context found.

Georgeff, M. P., & Lansky, A. L. (1986). Procedural knowledge. Proceedings of the IEEE special issue on knowledge representation, 74, 1383--1398.

No context found.

M. Georgeff and A. L. Lansky. Procedural knowledge. Proceedings of IEEE, 74(10):1383--1398, 1986.

No context found.

M. P. Georgeff and A. L. Lansky. Procedural knowledge. In Proceedingsof the IEEE Special Issue on Knowledge Representation, volume 74, pages 1383--1398, 1986.

No context found.

Georgeff, M.P., Lansky, A.L.: Procedural knowledge. Proceedings of the IEEE Special Issue on Knowledge Representation 74 (1986) 1383--1398

No context found.

M. Georgeff and A. L. Lansky. Procedural knowledge. Proceedings of IEEE, 74(10):1383--1398, 1986.

No context found.

George#, M. P., and Lansky, A. L. 1987. Procedural knowledge. Technical Report 411, Artificial Intelligence Center, SRI International.

No context found.

M. Georgeff and A. L. Lansky. Procedural knowledge. Proc. of IEEE, 74(10):1383-- 1398, 1986.

No context found.

Georgeff, M. P., & Lansky, A. L. (1986). Procedural Knowledge. Proceedings of the IEEE (Special Issue on Knowledge Representation), 74, 1383-1398.

No context found.

--206. Georgeff, M., and Lansky, A. L. 1986. Procedural knowledge. Proc. of IEEE 74(10):1383--1398. Giunchiglia, E.; Kartha, G. N.; and Lifschitz, V.

No context found.

Georgeff, M. P. & A. L. Lansky (1986) Procedural Knowledge, in Proceedings of the IEEE, Special Issue on Knowledge Representation, Vol. 74, pp. 1383-1398.

No context found.

Georgeff, M. P. and Lansky, A. L. (1986), "Procedural Knowledge", in Proc. of the IEEE, Special Issue on Knowledge Representation, pp. 1383--1398.

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