J. A. Hendler, A. Tate and M. Drummond. AI planning: Systems and techniques. AI Magazine 11(2):61--77, Summer 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....validated by applications developed by Boddy and Dean that involve a small number of anytime algorithms. The validity of the claim with respect to large systems composed of anytime algorithms is presented in Chapter 6. For a survey of artificial intelligence planning systems and techniques see [Hendler et al. 1990] . # Several exceptions will be described in Chapter 2. Claim 5. Effectiveness Operational rationality is a powerful model that can effectively simplify the development of complex real time systems. This is an immediate result of the previous claims and the major simplification they introduce ....

J. A. Hendler, A. Tate and M. Drummond. AI planning: Systems and techniques. AI Magazine 11(2):61--77, Summer 1990.

....the modifications made by other specialists and so interactions that occur between specialists go unchecked and may ultimately lead to design errors and failure. In the past, the resolution of such interactions has been one of the stumbling blocks of many planning and automated design systems [15]. Dtjh Vu attempts to overcome this problem by using an efficient scheme of interaction representation and resolution. Using a set of adaptation strategies, Dtjh Vu can detect and repair many conflicts that might arise. Strategies are organised in terms of the interactions they resolve and each ....

Hendler, J., Tate, A., Drummond, M. AI Planning: Systems and Techniques. AI Magazine, 11, 2, pp 61-77, 1990.

....as user specifications, type of sensors used, etc, is stored in the context. Goals, operators, requests and context are represented as frames. This is shown in Fig. 6, along with the relationships between them. Planning The basic planning mechanism provided is hierarchical script based planning [8]. A plan is a set of steps to attain the desired signal processing objective. It is similar to the block schematics often used in the design of SP algorithms, but is different in two ways: a) plan hierarchy: each box in the plan can be a complex SP tasks, with its own plan and (b) plan ....

J.Hendler et al., "AI planning: systems and techniques," AI Magazine, pp. 61--67, 1990.

....the Planner selects an operator to produce a desired effect only if that effect is marked public. 6.3.2 Multi agent planning with primitive operators The Planner utilises classical partial order means end planning in its plan construction process. For a review of the planning literature see [17]) Thus, given a goal of the form of Figure 15(a) the Planner searches its Plan Database for an operator with a public effect that unifies (with unification bindings ) with the desired effect of the goal. If multiple operators are found, they are ranked by cost, and then by duration. Next, the ....

Hendler, J., Tate, A. & Drummond, M. "AI Planning: systems and techniques", AI Magazine, May 1990.

....problem of identifying and selecting the best plan for the given goal. Therefore in the Classical phase, planning is thought of as a search problem: the space of different options must be searched for the best option. Good summaries of the planning as search paradigm are found in: Georgeff, 1990; Hendler et al. 1990; McDermott, 1992] 2.1.1.1 Representation Strategies The early stages of research in planning focused on determining what steps the agent needed to take to accomplish a goal. Outgrowths of this early planning work also investigated temporal ordering of actions in a plan, goal conflict ....

James Hendler, Austin Tate, and Mark Drummond. AI planning: systems and techniques. Proceedings of the National Conference on Artificial Intelligence, pages 61--77, 1990.

....to a specific course of behavior that the agent has designed carefully in advance. Definition 2 suggests a commitment to a more general course of behavior whose details remain to be determined. These two definitions correspond roughly to two prominent models of planning in the AI literature [Hendler, 4 Tate, and Drummond, 1990], which we label plans as programs and plans as not programs. According to the plans as programs model, a plan is an executable computer program that an agent executes in order to act. Two corollaries follow. Planning is automatic programming. Plan following is plan execution. For some ....

Hendler, J., Tate, A., and Drummond, M. AI planning: Systems and Techniques. AI Magazine, 11, Summer, 1990.

....to determine whether or not it is reachable and whether or not is in the set of open nodes. For a path of length r this means that A has a worse case complexity of O(h r ) This exponential growth is characteristic of state space search based planning algorithms (Hendler, Tate and Drummond [19]) There are two factors that strongly influence how efficiently A can perform on a given problem: 1. The way the problem is interpreted as a graph. This has a strong influence on plan generation efficiency because it determines h. h should be kept as small as possible, without sacrificing plan ....

Hendler, James, Tate, Austin and Drummond, Mark. "AI Planning: Systems and Techniques". AI Magazine, pp. 6177, Summer 1990.

....for being suitable as a formal basis for more modern planners. For example, few interesting modern planners will describe actions just by preconditions and postconditions, but there will be time, a difference between preconditions and subgoals, plots, protection intervals, and others, see [ Hendler et al. 1990 ] for an overview. There are two answers to this question. First, even if many existing action formalisms (like the possible worlds formalism that still lacks incorporating, e.g. numerical time, concurrency, or external events) are severely restricted in their expressivity, this need not stay ....

J. Hendler, A. Tate, and M. Drummond. Ai planning: systems and techniques. AI Magazine, 11(2):61--77, 1990.

.... of many deep and complex issues: for example, how to represent and reason 63 TouringMachines Planning Layer j 64 about plans and actions, how to control the search of the agent s plan space, or how to generate optimal plans for a collection of simultaneous or conjunctive tasks [Geo90, HTD90] TouringMachines are primarily task oriented agents and as such should be capable of generating and executing sequences of actions to achieve the tasks assigned to them. The skills required to do this are many and varied. Among others, these include the ability to decompose large, composite ....

.... domains has been shown to be exponentially hard [Cha90] Sanborn and Hendler have recently shown that the analogous problem in dynamic domains is in fact undecidable [SH88, page 96] 3 Hendler et al. also refer to this as trading precision in decision making for time in responding to events [HTD90, page 71] TouringMachines Planning Layer j 67 certainties in the planner s world model) the plan will nevertheless remain an essential piece of information since it can be used, for instance, to provide the agent with important navigational landmarks for use at execution time [Lat91] All ....

[Article contains additional citation context not shown here]

James Hendler, Austin Tate, and Mark Drummond. AI planning: Systems and techniques. AI Magazine, 11(2):61--77, 1990.

....for being suitable as a formal basis for more modern planners. For example, few interesting modern planners will describe actions just by preconditions and postconditions, but there will be time, a difference between preconditions and subgoals, plots, protection intervals, and others, see [ Hendler et al. 1990 ] for an overview. There are two answers to this question. First, even if many existing action formalisms (like the possible worlds formalism that still lacks incorporating, e.g. numerical time, concurrency, or external events) are severely restricted in their expressivity, this need Turning an ....

J. Hendler, A. Tate, and M. Drummond. Ai planning: Systems and techniques. AI Magazine, 11(2):61--77, 1990.

....(by replacement or insertion of actions) and by successive hierarchical refinement. 2.2 Control of execution for supervision We are interested, here, in defining the type of control of execution which is the best adapted for the supervision of programs. Four main types of control of execution [HTD90] have been developed: i) prediction of failures, and planning of additional actions which wait until the world corresponds to what is expected; ii) prediction of failures, and integration in the plan of tests and correcting actions; iii) prediction of failures, and integration in the plan of tests ....

J. Hendler, A. Tate, and M. Drummond. AI Planning: Systems and Techniques. Ai Magazine, Summer:61--77, 1990.

....request. The degree to which these two agree is one measure of the fidelity of control that has been achieved. It will be useful to distinguish two approaches to dynamic system modeling that are analogous to the duality between actions and states that has been well discussed in the AI literature (Hendler et al. 1990). In a state based representation, system motion is viewed as a series of states that are altered by events. In an event based representation, the state of the system is derived implicitly from its initial state and all events that have occurred up to a particular point. While the choice of one ....

Hendler, J., Tate, A. and Drummond, M. 1990. AI Planning: Systems and Techniques. AI Magazine, Vol 11(2).

....for being suitable as a formal basis for more modern planners. For example, few interesting modern planners will describe actions just by preconditions and postconditions, but there will be time, a difference between preconditions and subgoals, plots, protection intervals, and others, see [ Hendler et al. 1990 ] for an overview. There are two answers to this question. First, even if many existing action formalisms (like the possible worlds formalism that still lacks incorporating, e.g. numerical time, concurrency, or external events) are severely restricted in their expressivity, this need not stay ....

J. Hendler, A. Tate, and M. Drummond. AI planning: Systems and techniques. AI Magazine, 11(2):61--77, 1990.

....contingencies. All of these factors contribute to the feasibility of effective computational vision and action in the interface. 4. VisMap As an integrated system, VisMap is strongly influenced by techniques for image processing [Gonzales and Woods, 1992] and artificial intelligence planning [Hendler et al. 1990] . Figure 1 shows the basic architecture. The bulk of the processing in VisMap is devoted to visual processing and motor behavior, as discussed below. 4.1. VISUAL PERCEPTION IN VISMAP The goal of the visual component of VisMap is as follows. Input to the system is the raw contents of the ....

Hendler, James; Tate, Austin; and Drummond, Mark 1990. AI planning: Systems and techniques. AI Magazine 61--77.

....the same effect, the goal being predictable behavior. Extraneous events. The only changes in the environment are due to actions of the planner. This ideal of user planner control is so well entrenched that in both planning and HCI the terms action and event are used interchangeably [9]. User control [3] From the conventional (i.e. nonagent based) perspective, user control over the interface should be nearly absolute. The interface does not initiate actions, but rather responds like a tool [25] Passage of time. Plans remain appropriate in the interval between their ....

James Hendler, Austin Tate, and Mark Drummond. Ai planning: Systems and techniques. AI Magazine, pages 61--77, Summer 1990.

....the same effect, the goal being predictable behavior. Exogenous events. The only changes in the environment are due to actions of the planner. This ideal of user planner control is so well entrenched that in both planning and HCI the terms action and event are used interchangeably (Hendler et al. 1990). User control (App, From the conventional (i.e. nonagent based) perspective, user control over the interface should be nearly absolute. The interface does not initiate actions, but rather responds like a tool (Woods and Roth, 1988) Passage of time. Plans remain appropriate in the ....

Hendler, James; Tate, Austin; and Drummond, Mark 1990. Ai planning: Systems and techniques. AI Magazine 61--77.

.... that the conventional interface must not initiate actions but rather respond to the user s commands as a tool (Apple 1992; Woods Roth 1988) This ideal of user planner control is so well entrenched that in both HCI and planning the terms action and event are used interchangeably (Apple 1992; Hendler, Tate, Drummond 1990). This is just one of many such correspondences. These relationships suggest that the user interface could play the role of a general purpose testbed for planning agents: planners generally abstract away the continuous, uncertain, dynamic, and unobservable properties of an environment, such that ....

Hendler, J.; Tate, A.; and Drummond, M. 1990. Ai planning: Systems and techniques. AI Magazine 61--77.

....Dependency In classical AI planning, a planning problem is specified by an initial state, one or more goal states, and a set of operators that transform one state to another. A plan is a sequence of operator instantiations (actions) that transforms the world from the initial state to a goal state (Hendler et al. 1990). Most planning algorithms rely mainly on the current state description to select an action; the dependency among the sequential events is implicit. SIBL views sequential events as dependent, rather than independent. In a sequence of n instances (s 1 , s i , s j , s n ) we ....

Hendler, J., Tate, A., and Drummond, M. (1990). AI planning: Systems and techniques.

....Graphplan is far faster in the study than the other algorithms. Lplan is often slower because the heuristic is time consuming, but Lplan shows promise because it often performs a small search. Introduction Planning is the problem of finding a combination of actions that achieves a goal (Allen, Hendler, Tate 1990; Hendler, Tate, Drummond 1990) So far, there is limited success in general purpose planning, in large part due to two factors. One factor is the computational complexity of planning, e.g. propositional STRIPS planning is PSPACEcomplete except for very severe restrictions (Backstrom Nebel ....

....in the study than the other algorithms. Lplan is often slower because the heuristic is time consuming, but Lplan shows promise because it often performs a small search. Introduction Planning is the problem of finding a combination of actions that achieves a goal (Allen, Hendler, Tate 1990; Hendler, Tate, Drummond 1990). So far, there is limited success in general purpose planning, in large part due to two factors. One factor is the computational complexity of planning, e.g. propositional STRIPS planning is PSPACEcomplete except for very severe restrictions (Backstrom Nebel 1995; Bylander 1994; Erol, Nau, ....

Hendler, J.; Tate, A.; and Drummond, M. 1990. AI planning: Systems and techniques. AI Magazine 11(2):61--77.

....to a specific course of behavior that the agent has designed carefully in advance. Definition 2 suggests a commitment to a more general course of behavior whose details remain to be determined. These two definitions correspond roughly to two prominent models of planning in the AI literature [Hendler, Tate, and Drummond, 1990], which we label plans are programs and plans are not programs. According to the plans are programs model, a plan is an executable program an atomic level sequence of actions or end effector commands that an agent executes in order to act. Two corollaries follow. Planning is automatic ....

Hendler, J., Tate, A., and Drummond, M. AI planning: Systems and Techniques. AI Magazine , 11, Summer, 1990.

....means end planning. That is, given a goal, a rational agent would adopt a course of action which reduces the difference between her current state and her desired goal state. While significant progress has been made in planning research (for a review of the planning literature, see Steel (1987) and Hendler et al. 1990)) the state of the art of automated planning still leaves a lot to be desired, particularly in their use as the core reasoning component of industrial multi agent systems. We mention a few of the key problems below. Slow execution times: Arguably, the major practical problem with classical AI ....

Hendler J., Tate A. & Drummond M. (1990) AI planning: systems and techniques, AI magazine 11(2) 61-77.

....agree is one measure of the fidelity of control that has been achieved. 5.2.1. 4 Duality of States and Events It will be useful to distinguish two approaches to dynamic system modeling that are analogous to the duality between actions and states that has been well discussed in the AI literature [25]. In a state based representation, system motion is viewed as a series of states that are altered by events. In an event based representation, the state of the system is derived implicitly from its initial state and all events that have occurred up to a particular point. While the choice of one ....

J. Hendler, A. Tate, and M. Drummond, "AI Planning: Systems and Techniques," AI Magazine, Summer, 1990, Vol 11 (2).

.... : 28 6 Concluding Remarks 29 1 Introduction Traditional approaches to task planning assume that the planner has access to all of the world information needed to develop a complete, correct plan a plan which can then be executed in its entirety by a robot [ Hendler et al. 1990, McDermott, 1992 ] Unfortunately, this information about the world may not always be available at plan time. This is particularly true when we consider autonomous robots that must operate under general goals over extended periods in unpredictable and changing environments. When crucial ....

James Hendler, Austin Tate, and Mark Drummond. AI planning: Systems and techniques. AI Magazine, 11(2):61--77, Summer 1990.

....Syntax The adopted representation formalism defines points in the search space as partially elaborated plans, traversed using plan transformations. Plans are represented as an action ordering in which the actions, described by operators, are strung together with temporal ordering relations [10]. There are two types of action: those which can be undertaken by individuals (primitive actions) and those in which groups of agents work together (social actions) 2 . Throughout this section, let the set of agents in the community be represented by A, the set of primitive actions which can be ....

J.Hendler, A.Tate & M.Drummond (1990) "AI Planning: Systems and Techniques", AI Mag., 61-77

....such as user specifications, types of sensors used, etc, is stored in the context. Goals, operators, requests and context are represented as frames. These frames and their inter relationships are shown in Fig. 2. Planning The basic planning mechanism provided is hierarchical script based planning [6]. A plan is a set of steps to attain the desired IU objective. It is similar to the block schematics often used in the design of IU algorithms, but is different in two ways: a) Plan hierarchy: each box in the plan can be a complex IU task, with its own plan, b) Plan abstraction: the components ....

J.Hendler et al., "AI planning: systems and techniques," AI Magazine, pp. 61--67, 1990.

....of actions, that will transform the world from a given initial state to a desired goal state. Ideally the set of actions so produced is then passed on to a robot or a numeric control machine, or some other form of effector, which can follow the plan and produce the desired result [DARPA 90] Hendler et al. 90] Swartout 88] Wilkins 88] Planning problems have been tackled in two major ways: without using domain specific knowledge and approaches which use domain heuristics directly. There have been a number of attempts to combine techniques available at a given time into prototypes able to cope with ....

....which use domain heuristics directly. There have been a number of attempts to combine techniques available at a given time into prototypes able to cope with increasingly more realistic application domain. Table 1 lists some of these efforts and the domains to which they are applied (in part from [Hendler et al. 90] Planner Domain STRIPS [Fikes and Nilsson 71] Simple Robot control HACKER [Sussman 73] Simple Program Generation NOAH [Sacerdoti 77] Mechanical Engineers Apprentice Supervision NONLIN [Tate 77] Electricity Turbine Overhaul NASL [McDermott 78] Electronic Circuit Design OPM [Hayes Roth and ....

Hendler, J., Tate, A. and Drummond, M. "AI planning: systems and techniques", AI Magazine, Summer, 1990.

....tradeoff [27] I primarily consider the expressiveness of operators and formulas. 1.1 Previous Research The literature on planning is voluminous, and no attempt to properly survey the planning literature is attempted here. Instead, the reader is referred to Allen et al. 1] and Hendler et al. [23]. Despite the sizable literature, results on computational complexity are sparse. Dean and Boddy [12] analyze the complexity of temporal projection given a partial ordering of events and causal rules triggered by events, determine what conditions must be true after each event. Their ....

....To say the least, these results in combination with previous analyses are not encouraging for domain independent planning. However, operators must have multiple preconditions, simultaneous positive and negative postconditions, and apparently many more features to implement any interesting domain [9, 23]. While additional features might be good for making a planner more useful as a programming tool, generality has its downside tractability of planning cannot be guaranteed even with moderately expressive operators. All of this again calls into question the restricted language thesis proposed ....

J. Hendler, A. Tate, and M. Drummond. AI planning: Systems and techniques. AI Magazine, 11(2):61--77, 1990.

....noted in [10] 16] 4. Execution in this context corresponds to searching through the space of partial plans if a node is expandable or processing a primitive action if not. 5. is a non commutative, non associative n ary operation (n 2) plan transformations of action ordering plans [9]) Such relationships are an integral component of the solution specification and if they are not satisfied then the desired objective cannot be guaranteed by solution S s . Hence fulfilling an intention means performing the actions and satisfying any relationships which exist with other actions ....

J.Hendler, A.Tate & M.Drummond (1990) "AI Planning: Systems and Techniques", AI Magazine, pp 61-77

....as user specifications, types of sensors used, etc, is stored in the context. Goals, operators, requests and context are represented as frames. These frames and their inter relationships are shown in Figure 2. Planning The basic planning mechanism provided is hierarchical script based planning [6]. A plan is a set of steps to attain the desired IU objective. It is similar to the block schematics often used in the design of IU algorithms, but is different in two ways: a) Plan hierarchy: each box in the plan can be a complex IU task, with its own plan, b) Plan abstraction: the components ....

J. Hendler, A. Tate, and M. Drummond, "AI planning: Systems and techniques," AI Magazine, Vol. 11, No. 2, pp. 61--77, 1990.

....program consisting of primitive actions that an agent executes in order to act. Under this model, planning is a type of automatic programming, and plan following simply consists of direct execution. The majority of planning related AI research to date utilizes this model in various forms (see [20, 29] for overviews) For some researchers, plans impose a sequential structure actions and plan following is predominantly sequential and open loop [10, 33, 7] For others, plans impose less structure but more reactivity to cope with a variety of environmental conditions [1, 28, 25, 30] Others ....

J. Hendler, A. Tate, and M. Drummond. Ai planning: Systems and techniques. AI Magazine, 1990.

....for being suitable as a formal basis for more modern planners. For example, few interesting modern planners will describe actions just by preconditions and postconditions, but there will be time, a difference between preconditions and subgoals, plots, protection intervals, and others, see [12] for an overview. There are two answers to this question. First, even if many existing action formalisms (like the possible worlds formalism that still lacks incorporating, e.g. numerical time, concurrency, or external events) are severely restricted in their expressivity, this need not stay that ....

J. Hendler, A. Tate, and M. Drummond. AI planning: Systems and techniques. AI Magazine, 11(2):61--77, 1990.

....with interactions that arise during the adaptation of a case by the specialists. Specialists are local and therefore ignorant of global interactions between case elements that may lead to problem solving failures; interactions cause problems in many planning and automated design systems (see e.g. Hendler, Tate Drummond, 1990). D j Vu s adaptation strategies detect and repair different classes of interactions that arise. The strategies are organised in terms of the interactions they resolve and each is indexed by a description of the type of failure it can repair. Each strategy also has a set of repair methods for ....

Hendler, J., Tate, A., Drummond, M. (1990) AI Planning: Systems and Techniques. AI Magazine, 11(2), (pp. 61-77).

....solving actions to be completely pre planned. In other words, the problem with classical planners is that they are neither opportunistic nor reactive [36] Recently there has been a significant amount of research on reactive planning [16, 18] and on the interleaving of planning and execution [26]. As we will see, these types of approaches can make planning appropriate for the control of problem solving systems. One of the key motivations for using a planning based control framework is that the goal plan subgoal hierarchy that is instantiated by a planner provides detailed and explicit ....

....through incremental planning and by maintaining only a partial model of the state of the world. Forming complete plans prior to executing them is not possible for control in applications like interpretation where the outcome of actions is uncertain and where external agents can affect the world [26]. Our control planner is a reactive planner, in part, because it interleaves planning with execution and bases further expansion of its plans on the outcome of earlier actions. Likewise, while classical planners maintain a complete model of the state of the world, this is inappropriate for most ....

J. Hendler, A. Tate, and M. Drummond, "AI Planning: Systems and Techniques," AI Magazine, vol. 11, no. 2, pp. 61--77, 1990.

....may be ruled out in protecting goals that have already been achieved. The major limitation with decomposition, however, is that for most planning problems the goals and subgoals often interact or conflict with each other in non serializable ways 2 . 1 A review of this work is presented in [Hendler et al. 1990]. 2 The most famous example of this is the Sussman anomaly, in which solving one goal undoes the Unfortunately, it appears impossible to achieve both efficiency and generality in handling goal subgoal interactions. Domain independent planners attempt to handle interactions that can occur in ....

J. Hendler, A. Tate, and M. Drummond "AI Planning: Systems and Techniques" AI Magazine, 11(2), May, 1990, 61-77.

....Foundation or ONR. example, strips style planning systems 1 were developed more than twenty years ago (Fikes et al. 1971) and most of the practical work on AI planning systems for the last fifteen years has been based on hierarchical task network (htn) decomposition (Sacerdoti, 1977; Tate, 1990; Wilkins, 1988a; Wilkins, 1988b) In contrast, although the past few years have seen much analysis of planning using strips style operators, Chapman, 1987; Mcallester et al. 1991; Erol et al. 1992b; Erol et al. 1992a) there has been very little analytical work on htn planners. One big ....

....to give an intuitive feel for htn planning. A precise formal description is presented in Section 3. One of the motivations for htn planning was to close the gap between AI planning techniques such as strips style planning, and operations research techniques for project management and scheduling (Tate, 1990). Thus, there are some similarities between htn planning and strips style planning, but also some significant differences. htn planning uses actions and states of the world that are similar to those used in strips style planning. 2 Each state of the world is represented by the set of atoms ....

[Article contains additional citation context not shown here]

:61--77, Summer 1990.

....those planning systems. For example, strips style planning systems 1 were developed more than twenty years ago (Fikes et al. 1971) and most of the practical work on AI planning systems for the last fifteen years has been based on hierarchical task network (htn) decomposition (Sacerdoti 1977; Tate 1990; Wilkins 1988a; Wilkins 1988b) In contrast, although the past few years have seen much analysis of planning using strips style operators, Chapman 1987; McAllester et al. 1991; Erol et al. 1992b; Erol et al. 1992a) there has been very little analytical work on htn planners. One big obstacle to ....

....task of making a round trip to New York. This could not be easily expressed as a single goal task either, since the initial and final states would be the same. 2 Due to space limitations, we cannot include the details of our proofs, nor the details of how our work compares to (Sacerdoti 1977; Tate 1990; Wilkins 1988a; Yang 1990; Kambhampati et al. 1992) These are presented in (Erol et al. 1994a) n 1 : achieve[clear(v 1 ) n 2 : achieve[clear(v 2 ) n 3 : do[move(v 1 ; v 3 ; v 2 ) F NaN F NaN F NaN R F NaN Gamma F NaN Gamma F NaN Gamma clear(v1 ) clear(v2 ) on(v1 ; ....

[Article contains additional citation context not shown here]

Tate, A.; Hendler, J. and Drummond, D. AI planning: Systems and techniques. AI Magazine, (UMIACSTR -90-21, CS-TR-2408):61--77, Summer 1990.

No context found.

Tate, A.; Hendler, J. and Drummond, D. AI planning: Systems and techniques. AI Magazine, (UMIACS-TR-90-21, CS-TR-2408):61-- 77, Summer 1990

No context found.

James Hendler, Austin Tate, and Mark Drummond. AI planning: Systems and techniques. AI Magazine, 11:61--77, 1990.

No context found.

Hendler, J. and Tate, A. and Drummond, M. AI Planning: Systems and Techniques. AI Magazine. 11(2), 1990.

No context found.

Hendler, J., Tate, A., and M. Drummond. AI planning: Systems and techniques. AI Magazine, 11, 61-77, (1990).

No context found.

Hendler, J., Tate, A., & Drummond, M. (1990). AI planning: Systems and techniques. AI Magazine 11(2): 61--77.

No context found.

James Hendler, Austin Tate, and Mark Drummond. AI Planning: Systems and Techniques. AAAI, pages 61--77, Summer 1990.

No context found.

Hendler, J.A., Tate, A. & Drummond, M. (1990) "AI Planning: Systems and Techniques", in AI Magazine, 11, No. 2, 1990, pp. 61--77. Also in [Allen et. al., 1990].

No context found.

Hendler, J., Tate, A., and Drummond, M. (1990). AI planning: Systems and techniques. AI Magazine, pages 61--77.

No context found.

Hendler, James, Tate, Austin, and Drummond, Mark, AI Planning: Systems and techniques, AI Magazine, 11(2) (1990) 61--77.

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