Philippe Morignot. Embedded Planning. In Proceedings of the Second International Conference on A.I. Planning Systems, pages 128133, Chicago, IL, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....which are not planned for are grouped to form a conjunctive goal. ffl Plan generation. This behavior turns (conjunctive) goals into a partial order of partially instantiated action descriptions, given the current situation perceived so far and the action descriptions that can be carried out [ Morignot, 1994 ] ffl Plan monitoring. Chooses the current plan and sends the current action description to the physical level and notes progress. An agent configured with these components can handle a stream of incoming goals: Goals are generated (i) by transformation of percepts and intentions (e.g. ....

Philippe Morignot. Embedded Planning. In Kristian Hammond, editor, Proceedings of the Second International Conference on A.I. Planning Systems, pages 128--133, Chicago, IL, 1994.

....its environment through a sonar infra red ring, a laser CCD 9 camera pair and a bumper ring; it can act through its motors and through a voice synthesizer. The cognitive level of this real agent includes one or two implemented methods per task (e.g. a case based method and a classical one [13] for the plan generation task) similarly, the physical level includes one or two implemented methods per task (e.g. a sonar infra red based corridorfollowing method, a potential eld one for the navigation task, a landmark detection method [20] The scenarios unfold in indoor domains such as ....

Philippe Morignot. Embedded Planning. In Proceedings of the Second International Conference on A.I. Planning Systems, pages 128133, Chicago, IL, 1994.

....a pour mission d indiquer # chaque instant quelle doit #tre la proc#dure physique # activer, compte tenu de toutes les informations dont l agent dispose. Ce niveau poss#de un module de g#n#ration de plans d actions, ou plani cation [Fikes and Nilsson, 1971] Pednault, 1986] Chapman, 1987] [Morignot, 1994]. Le niveau num#rique comprend un module de navigation. Ce module utilise les capteurs sonar et infra rouge pour naviguer dans un environnemnet compos# de couloirs REF OLIVIER ICI. 3.3 Implantation L implantation particuli#re du plani cateur est d#crite dans [Morignot, 1994] Il s agit ....

....1986] Chapman, 1987] Morignot, 1994] Le niveau num#rique comprend un module de navigation. Ce module utilise les capteurs sonar et infra rouge pour naviguer dans un environnemnet compos# de couloirs REF OLIVIER ICI. 3. 3 Implantation L implantation particuli#re du plani cateur est d#crite dans [Morignot, 1994]. Il s agit essentiellement d un plani cateur non lin#aire, correct et complet [Chapman, 1987] Le domaine des variables peut #tre repr#sent# ; le lieu fait l objet d un traitement particulier ; des contraintes arithm#tiques simples peuvent #tre repr#sent#es. Voici la liste des descriptions des ....

Philippe Morignot. Embedded Planning. In Proceedings of the Second International Conference on A.I. Planning Systems, pages 128133, Chicago, IL, 1994.

....of the right plan to the physical level) and also exponential algorithms such as plan generation. Although no upper bound can be theoretically put on the time it takes to generate a plan, our experiments show that in usual cases a plan can be generated in less than 1 second in real time [Morignot 94] which is reasonable given the actual speed of the robot (10 inches per second) These two levels are architecturally isomorphic, in the sense that they both include a mechanism (meta controller, in figure 1) to choose the next physical cognitive module to run, among the executable ones. The ....

.... of physical actions O (operators) Cognitive is initially called with the arguments (O, S init ) In practice, S init is a list of litterals, O is composed of three lists of litterals classified as prerequisites, effects and preservation conditions in the classical planning sense, see [Morignot 94] 11 Cognitive(O,B) 1. Situation Assessment) Put external or reflective percept p on B. 2. Intention Generator) Given p, generate one severals goal G and put them on B. 3. Goal Merger) Given G, use domain knowledge in B to form conjunctive goals G i 4. Plan Generator) Given G i , ....

Philippe Morignot. Embedded Planning. In Proceedings of the 2nd Int. Conf. on A.I. Planning Systems , AIPS'94. Chicago, IL, 1994. p. 128-133.

....planning agent to constantly have final goals such as ( battery level 10) present in any plan before any environment related goal that might be sent to the agent. For a plan generator, this leads to (1) representing arithmetic relations inside the planner formalism ( Penberthy Weld 93] Morignot 94] to a lesser extent) and (2) considering terms the truth value of which changes over time (or has to be sensed within the planner [Etzioni Lesh 93] Unfortunately, planners using a causal theorem [Pednault 91] or modal truth criterion [Chapman 87] to compute the truth value of terms cannot ....

.... et al. 95] This robot, a Nomadics 200, is capable of sensing its environment through a sonar ring, a laser CCD camera pair and a bumper ring; it can act through its motors and a voice synthesizer [Zhu 92] The controlled symbolic modules involve for example goal formulation, plan generation [Morignot 94] plan monitoring (for the reasoning level of the agent) and path planning, navigation, landmark detection [Tokusai 94] for the fast level of the agent) The tested scenarios unfold in indoor domains such as office surveillance or object delivery. Since requests come on the fly from other ....

Philippe Morignot. Embedded Planning. In Proceedings of the 2nd Int. Conf. on A.I. Planning Systems , AIPS'94. Chicago, IL, 1994. p. 128-133.

....sensing its environment through a sonar infra red ring, a laser CCD camera pair and a bumper ring; it can act through its motors and through a voice synthesizer. The cognitive level of this real agent includes one or two implemented methods per task (e.g. a case based method and a classical one [Morignot, 1994] for the plan generation task) similarly, the physical level includes one or two implemented methods per task (e.g. a sonar infra red based corridor following method, a potential eld one for the navigation task, a landmark detection method [Tokusai, 1994] The scenarios unfold in indoor ....

Philippe Morignot. Embedded Planning. In Proceedings of the Second International Conference on A.I. Planning Systems, pages 128133, Chicago, IL, 1994.

.... mean that they are unordered, an uninstantiated variable means that it can be instantiated to more than one constant (the domain of a variable is pruned during the process) Arithmetic equations can be maintained by representing integers as function symbols in actions pre and postconditions [13]; this enables to represent limitations on the capacity of the robot. Task planning is a search in the space of partiallyordered partially instantiated plans [4] A best first search, for example, ensures completeness. A plan is a solution to a given planning problem (i.e. initial plan plus ....

....the performance of a particular behavior but to demonstrate the usefulness of a multi agent architecture for controlling a unique vehicle. However, in cases actually happening in our scenarios in our environment, task planning is faster than action execution (less than one second for task planning [13], several seconds for action execution [1] But due to the polynomial or exponential complexity of each behavior, opposite observations could be made in close scenarios (e.g. safely moving seven stacked boxes with one intermediate location only (problem known as the Towers of Hanoi) For the ....

Philippe Morignot. Embedded Planning. In Kristian Hammond, editor, Proceedings of the Second International Conference on A.I. Planning Systems, pages 128--133, Chicago, IL, 1994.

.... of the AIbot include situation assessment (i.e. recognizing percepts) intention creation (i.e. turning percepts into intentions) goal merging (i.e. turning intentions into actual committed (possibly conjunctive) intentions) plan generation (i.e. turning conjunctive goals into actions [Morignot 94] plan monitoring (i.e. sending each action to the physical level) The physical capabilities of the Aibot include motion (path planning using a cell decomposition and actual motion using potential field) perception and speech. 6 Behavior Behavior . Meta controller Current plan ....

Philippe Morignot. Embedded Planning. In Proceedings of the Second International Conference on A.I. Planning Systems (AIPS'94). Chicago, IL, 1994, p. 128133. 9

....might be applicable for the same purposes, e.g. a reduced version of ADL [Pednault, 1986] such as the one implemented in UCPOP [Penberthy and Weld, 1992] The planner is extended regarding locations, arithmetic equations and variables domain without change of the action representation. See [Morignot, 1994] for further details. 2.2 Search Algorithm The search algorithm of the planner is the algorithm that uses area dependent information of previous section to nd a goal state given an initial state (see below for the de nition of states, etc) For planning, these aread dependent information can ....

....a longer one thus the second property of f . More important, the third point is based on the following theorem: Theorem 1 For a particular prerequisite of an action, the number of times the MTC needs to be called depends on the number of actions in the future (solution) plan only. Proof. See [Morignot, 1994]. BrieAEy, the MTC proposes the following plan modi cations: addition of a uni cation constraint, addition of nonuni cation constraint (e.g. separation [Chapman, 1987] addition of a time ordering constraint among actions (e.g. promotion, demotion [Chapman, 1987] addition of an action to ....

[Article contains additional citation context not shown here]

Philippe Morignot. Embedded Planning. In Proceedings of the Second International Conference on A.I. Planning Systems, pages 128133, Chicago, IL, 1994.

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