C. Baral and S. C. Tran. Relating theories of actions and reactive control. Linkoping Electronic Articles in Computer and Information Science, 3(9), 1998. Available at: http://www.ep.liu.se/ea/cis/1998/009.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....include constructs for sequence, conditionals, loops and recursion. The emphasis is on reactive programs, but their proposal does provide for replanning during execution. Several authors rely on formal theories of actions for the purposes of characterizing appropriate notions of action failures [2, 24], but they do not consider execution monitoring per se. Perhaps the most sophisticated existing plan execution monitor is the XFRM system of Beetz and McDermott [4, 5] This provides for the continual modification of robot plans (programs) during their execution, using a rich collection of ....

C. Baral, T. and Son. Relating theories of actions and reactive control. In Robots, Softbots, Immobots: Theories of Action, Planning and Control, working notes of the workshop held on July 27, 1997 in conjunction with the AAAI-97, Providence, Rhode Island.

....attempted to reconcile the need to carefully plan sequences of actions with the need to interact timely with a rapidly evolving environment. Some researchers have proposed to substitute planning with more lightweighted reasoning techniques, like behaviours ( 9] 8] policies ( 3] 4] 10] [2]) and others. The drawback of these solutions is that, while they solve the problem of a timely interaction with the environment, they pose new problems because of the general reduction in the representation and reasoning power. On the other hand, other researchers have developed solutions which ....

Chitta Baral and Son Cao Tran. Relating theories of actions and reactive control. Transactions on Arti cial Intelligence, 2:211-271, 1998.

....specification aspects. This allows that more complex problems can be managed, including Cognitive Robotics aimed at the construction of robots whose architecture is based on the idea of representing the world by sentences of formal logic and reasoning about it by manipulating those sentences [12, 1]. Most of these formalizations define an entailment relationship (j= between the specifications (of effects of actions and relation among objects of the world) This ability allows to make plans that will take us to particular kind of worlds and explain observations about the 1 Computer Science ....

C. Baral and S.C. Tran, `Relating theories of action and reactive control ', Linkoping Electronic Articles in Computer Science, 3(9), (1998).

....and every occurrence of the Holds predicate in is of the form Holds(F 0 ; s) The universal frame axioms (4) and (5) describe how the world stays the same (as opposed to how it changes) 2 Example 1 (OOEce scenario) Let us consider a mobile robot navigating the oOEce AEoor in Fig. 1 [1]. This example concerns the representation of knowledge about the objects in a robot s environment, and how such knowledge is acquired. We assume the state space has been reduced to a manageable size by careful choice of perception and action primitives. The actions used in this scenario are of ....

.... holds(atroom(A) do(go(forward) B) holds(onrt(wall) B) holds(appcorner, B) holds(atroom(A) B) holds(atroom(301) do(go(forward) A) holds(onrt(wall) A) holds(appcorner, A) holds(atelevator, A) The results of learning can verify the correctness of control modules [1] for a robot, or more interestingly, they can be used to automatically construct control modules given a set of states and a goal. For instance, in the module to pass by a corner, when the robot approaches a corner, it must wait until the corridor is on its right and then turn 90 degrees to ....

C. Baral and S.C. Tran. Relating theories of action and reactive control. Linkoping Electronic Articles in Computer Science, 3(9), 1998.

....control loop and the first one to propose an architecture built on theorem provers that is suitable for realizing complex tasks. Shanahan, 1996 ] describes a map building process using abduction, but then implements his theory in an algorithm that is proved to have his abductive semantics. Baral and Tran, 1998 ] define control modules to be of a form of Stimulus Response (S R) agents (see [ Nilsson, 1998 ] relating them to the family of action languages A (e.g. Gelfond and Lifschitz, 1993 ] Giunchiglia et al. 1997 ] They provide a way to check that an S R module is correct with respect to ....

C. Baral and S.C. Tran. Relating theories of actions and reactive control. Electronic Trans. on Artificial Intelligence, 1998. Under review.

....control loop and the first one to propose an architecture built on theorem provers that is suitable for realizing complex tasks. Shanahan, 1996] describes a map building process using abduction, but then implements his theory in an algorithm that is proved to have his abductive semantics. [Baral and Tran, 1998] define control modules to be of a form of Stimulus Response (S R) agents (see [Nilsson, 1998] relating them to the family of action languages A (e.g. Gelfond and Lifschitz, 1993] Giunchiglia et al. 1997] Baral and Tran provide a way to check that an S R module is correct with respect ....

Chitta Baral and Son Cao Tran. Relating theories of actions and reactive control. Electronic Transactions on Artificial Intelligence, 1998. Under Review.

....a rudimentary approach to the formalisation of epistemic fluents, knowledge goals, and knowledge producing actions. Introduction Contemporary work in cognitive robotics has demonstrated the viability of logicbased high level robot control [Lesprance, et al. 1994] De Giacomo, et al. 1997] [Baral Tran, 1998], Shanahan, 2000b] Building on the progress reported in [Shanahan, 2000b] this paper describes an implemented logic based, high level robot control system in the cognitive robotics style. The controller is programmed directly in logic, specifically in the event calculus, an established ....

C.Baral and S.C.Tran, Relating Theories of Actions and Reactive Control, Linkping Electronic Articles in Computer and Information Science, vol. 3 (1998), no. 9.

....version is proposed as a general architecture for building intelligent agents. Notice that we did not include a speci c action theory above L0, but only showed how can such a theory in uences L0. We compare our work to those of Shanahan ( Shanahan 1996) Shanahan 1998) Baral and Tran ((Baral Tran 1998)) and Reiter et al. ( Reiter 1998) Lesprance et al. 1996) Reiter 1996) Shanahan 1998) describes a map building process using abduction. His abduction is specialized for the speci c scenario of spatial occupancy and noise that one may wish to include. He then implements the theory in an ....

....we can in fact include a version of the original theory presented by Shanahan as an upper layer. Also, since our robot is a more sophisticated mobile robot, any inclusion of such a theory will have to take into account the di erent sensors (sonars instead of switches) The work of Baral and Tran ((Baral Tran 1998)) focuses on the relationship between the members of the family of action languages A ( Gelfond Lifschitz 1993) Giunchiglia, Kartha, Lifschitz 1997) Kartha Lifschitz 1994) and reactive control modules. They de ne control modules to be of a form of StimulusResponse (S R) agents (see ....

Baral, C., and Tran, S. 1998. Relating theories of actions and reactive control. Electronic Trans. on Arti cial Intelligence. Under review.

....G follows. 9 Finally, if there is no plan from the current state to the goal state, the planner will fail to generate even the first operator. In this case we simply output o and strong R completeness follows. 2 Constructions similar to those used in the previous proof have been presented by Baral and Tran [1998]; our research were done independently around the same time. It is crucial that the planner used in the previous theorem generates shortest plans, otherwise, we cannot guarantee A completeness. We illustrate this with a small, contrived example. Example 9 Consider the PSN structure Phi = hP; Oi ....

C. Baral, S. C. Tran, Relating theories of actions and reactive control, Linkoping Electronic Articles in Computer and Information Science 3(9) (1998).

....system and the actual effects of control actions. We show how this can be done for a particular Hybrid Systems example modeled as Hybrid Automata. 1 Introduction There is a large body of work on the problem of hybrid controller synthesis being done in the Artificial Intelligence (AI) e.g. [6]) Hybrid Systems (e.g. 9] and Control Theory (e.g. 10] communities. Typically, a specification of a controller is synthesized from an open loop specification of the plant and a set of desired properties . However, the transformation of such a specification to a real control system has not ....

C. Baral and S. Tran. Relating theories of actions and reactive control. Linkoping Electronic Articles in Computer and Informations Science, 3(009), 1998. URL: http://www.ep.liu.se/ea/cis/1998/009/.

.... process cannot be interrupted, and therefore exogenous events that call for immediate action are ignored by such a planner [1] In dynamic environments however, a planner has to respond rapidly to observations that possibly outdate an already constructed plan (referred to as reactive control [1, 15, 2]) while at the same time keep track of the achievement or maintenance of a high level goal (called classical planning [7, 16] A software agent assisting a user performing routine tasks such as downloading and printing a document from the Web is suggestive for the type of agent under ....

....execution. However, in Aps the conceptual tie between the planning theory and the reactive component is rather loose: the Emergency module is employed upon certain observations, but the module is distinct from the planning theory whereas in [10] reactivity is wired into the theory. Baral and Son [2] attempt to establish a formal connection between theories of reasoning about action (as planning formalisms) and reactive control modules. Lesp erance et al. 11] have recently added a reactive component to ConGolog (a logic based agent programming language) Their framework contains interrupts ....

Chitta Baral and Tran Cao Son. Relating theories of action and reactive control, 1998. Submitted for publication.

....scenarios: 1) single oor planning, 2) lower level elevator planning, and (3) multi oor planning. SD is standard deviation. Shanahan, 1996 ] describes a map building process using abduction, but then implements his theory in an algorithm that is proved to have his abductive semantics. Baral and Tran, 1998 ] de ne control modules to be of a form of Stimulus Response (S R) agents (see [ Nilsson, 1998 ] relating them to the family of action languages A (e.g. Gelfond and Lifschitz, 1993 ] Giunchiglia et al. 1997 ] They provide a way to check that an S R module is correct with respect to ....

C. Baral and S.C. Tran. Relating theories of actions and reactive control. Electronic Trans. on Articial Intelligence, 1998. Under review.

....version is proposed as a general architecture for building intelligent agents. Notice that we did not include a specific action theory above L0, but only showed how can such a theory influences L0. We compare our work to those of Shanahan ( Shanahan 1996) Shanahan 1998) Baral and Tran ((Baral Tran 1998)) and Reiter et al. ( Reiter 1998) Lesprance et al. 1996) Reiter 1996) Shanahan 1998) describes a map building process using abduction. His abduction is specialized for the specific scenario of spatial occupancy and noise that one may wish to include. He then implements the theory in an ....

....we can in fact include a version of the original theory presented by Shanahan as an upper layer. Also, since our robot is a more sophisticated mobile robot, any inclusion of such a theory will have to take into account the different sensors (sonars instead of switches) The work of Baral and Tran ((Baral Tran 1998)) focuses on the relationship between the members of the family of action languages A ( Gelfond Lifschitz 1993) Giunchiglia, Kartha, Lifschitz 1997) Kartha Lifschitz 1994) and reactive control modules. They define control modules to be of a form of StimulusResponse (S R) agents (see ....

Baral, C., and Tran, S. C. 1998. Relating theories of actions and reactive control. Electronic Transactions on Artificial Intelligence. Under Review.

.... environment (e.g. 6] and the agent s internal state or knowledge [10, 13] Since a control program can be seen a mapping from knowledge to actions, an action model can be used to predict whether a control program is executable and whether it will lead the agent to its intended goal (see also [9, 2]) An action model is a good standpoint for the evaluation of the executability of a program, determining if the constraints associated with the actions are met, the conditions for the execution of the actions are achieved, and if such conditions can be evaluated. In some agent programming ....

....form p[i] where p is an action symbol and i is a time point (a non negative integer) An observation is an expression of the form F [i] where F is a non action formula and i is a time point. For example, the rules above about hitting a nail together with the actions A = fhit[0] hit[1] hit[2]g and the observation h[0] 5 constitute an action theory. The semantics of such theories is given below. Provided that the value of the fixed symbol ffi is 2, this action theory for instance yields the conclusion h[3] 0. 3 Semantics The semantics maps an action theory into a set of state ....

[Article contains additional citation context not shown here]

C. Baral. Relating theories of action and reactive control. In Proceedings AAAI-96 Workshop on Theories of action, Planning and Robot Control: Bridging the gap, 1996.

....control loop and the rst one to propose an architecture built on theorem provers that is suitable for realizing complex tasks. Shanahan, 1996 ] describes a map building process using abduction, but then implements his theory in an algorithm that is proved to have his abductive semantics. Baral and Tran, 1998 ] de ne control modules to be of a form of Stimulus Response (S R) agents (see [ Nilsson, 1998 ] relating them to the family of action languages A (e.g. Gelfond and Lifschitz, 1993 ] Giunchiglia et al. 1997 ] They provide a way to check that an S R module is correct with respect to ....

C. Baral and S.C. Tran. Relating theories of actions and reactive control. Electronic Trans. on Articial Intelligence, 1998. Under review.

No context found.

C. Baral and T. Son. Relating theories of actions and reactive control. Electronic transactions on Artificial Intelligence, 2(3-4), July-Dec 1998.

....known. However, the value can not be predicted and it will be known only at run time. The activities in k proposition s are what we have referred to as sensing activities (we also call them knowledge producing activities) and they have been studied formally in the context of planning problems [12, 28]. However, as we will discuss shortly, we need to slightly modify the k propositions if they are to be used in workflow specification. Example 2. Observe a simple workflow which automates a course registration process of a graduate student. Some of the activities description are: advisor(S; C) ....

....have the property that all the rules whose LHS are satisfied at a given state, have the same RHS. Informally, sequential workflows rule out the AND joins (OR rules do not result in concurrent execution of activities) For the class of sequential workflows we can directly apply the results in [12] to actually generate the control module from a given domain description and a given goal 8 . in [12] control modules are referred as plans) 8 The conditions of soundness and completeness for sequential workflows are given in [8] 5 Concluding Remarks, Related Literature and Future Work ....

[Article contains additional citation context not shown here]

C. Baral and T. Son. Relating theories of action and reactive control. In Linkoping Electronic Articles in computer and Information Science, volume 3. 1998.

....be the set of states, where a formula p holds. S is stabilizable w.r.t. E p iff S is alive and there exists an m such that for all trajectories oe whose x 0 2 S, and for all j, oe; j) j= 2m;m p. Conclusion and related work In this paper we formalized the notion of maintenance often mentioned (Baral Son 1998) in the context of robots and agents, as a property of a discrete event dynamic system (DEDS) and compared it with the notion of stability and stabilizability that are most popular in DEDS. We argued why maintainability may be a more preferred notion for certain systems and discussed active ....

....of supportability that generalizes both maintainability and stabilizability. Finally, we developed an automata theory that distinguishes between exogenous and control actions, and developed a temporal logic based on it. Our basic formulation of maintainability is related to the work in (Baral Son 1998). Among the other related works, there has been some work on defining stability of continuous systems in the presence of discontinuities and disturbances; for example (Sontag 1999) In the planning literature there has been some work on planning for temporal goals (Bacchus Kabanza 1998; Weld ....

Baral, C., and Son, T. 1998. Relating theories of actions and reactive control. Electronic transactions on Artificial Intelligence 2(3-4).

.... Phi(a; hs; Si) ffl In the presence of incomplete information and knowledge producing actions, there may not exist simple plans consisting of sequence of actions and we may need to extend the notion of a plan to allow conditional statements. We refer to such plans as conditional plans (e.g. [Lev96, BS97, BS98]) described below. ffl In order to query the system, we specify a query language LQS . A query in LQS has the form after P at s (10) where is a fluent formula and P is a conditional plan as formally defined below. Definition 6 (Conditional Plan) 1. An empty sequence of action, denoted by ....

C. Baral and T. Son. Relating theories of actions and reactive control. Electronic Transactions on Artificial Intelligence, 2(3-4), 1998.

.... f luent(has jack(O) object(O) Actions : agent action(drive to(L) location(L) agent action(get(O) object(O) exogenous action(lose(O) object(O) 2 More details on the combination of reactive and deliberative reasoning in the context of action based languages can be found in (Baral and Son, 1998). It also contains more detailed comparison to the reactive aspects of ConGolog (De Giacomo et al. 1997) Reasoning Agents in Dynamic Domains 277 Causal Laws : impossible if(drive to(L) jack at(L) causes(drive to(L) jack at(L) impossible if(get(ticket) has jack(money) ....

Baral, C. and Son, T. (1998). Relating theories of actions and reactive control. Electronic transactions on Artificial Intelligence, 2(3-4).

.... (a; hs; Si) In the presence of incomplete information and knowledge producing actions, there may not exist simple plans consisting of sequence of actions and we may need to extend the notion of a plan to allow conditional statements. We refer to such plans as conditional plans (e.g. [Lev96, BS97, BS98]) described below. In order to query the system, we specify a query language LQS . A query in LQS has the form after P at s (10) where is a uent formula and P is a conditional plan as formally de ned below. De nition 6 (Conditional Plan) 1. An empty sequence of action, denoted by [ ....

C. Baral and T. Son. Relating theories of actions and reactive control. Electronic Transactions on Articial Intelligence, 2(3-4), 1998.

....Notes 1. To the best of our knowledge the idea that observations should be added together with possible explanations was rst published in (Reiter, 1995) 2. More details on the combination of reactive and deliberative reasoning in the context of action based languages can be found in (Baral and Son, 1998). It also contains more detailed comparison to the reactive aspects of ConGolog (De Giacomo et al. 1997) ....

Baral, C. and Son, T. (1998). Relating theories of actions and reactive control. Electronic transactions on Articial Intelligence, 2(3-4).

....exogenous actions then it will reach the goal. Although this is a drawback for a robot that is being continuously harassed, there is no easy way out. One approach would be to avoid getting into such a situation altogether, and another would be 7 A detailed computation of the closure is given in [BS97b] 17 to be able to recognize it as a failure and trigger a recovery routine. The first approach is taken in [KBSD97,DKKN95] But their drawback is that the robot becomes too conservative and avoids too many situations and may consider certain goals unachievable which will be considered ....

....in S, there is only one rule in the control module Go clockwise 1room which is applicable in that state. This guarantees that the unfolding function unfolds to sequences of simple actions (no compound actions) for all states in S, and also that UM (s) does not end with a M F for any s in S. In [BS97b] we compute UM (s) for all possible states s, and show that the goal is true in each of those states. 3.3 Correctness of maintainance control modules So far we have discussed the correctness of achievement control modules. Recall that maintainance control modules are supposed to maintain a ....

C. Baral and T. Son. Relating theories of actions and reactive control. Technical report, Dept of Computer Sc, University of Texas at El Paso (http://cs.utep.edu/chitta/papers/relating.ps), 1997.

....9 R 8 R 3 R 5 R 4 C 6 C 11 C 17 C 12 C 13 C 14 C 15 C 16 C 10 F 1 Foyer Figure 1: A Diagram Showing the Topological Map for Event 1 Participants at the AAAI 96 Robotics Competition. 2 Our theory of agents : a brief description Our theory of agents is detailed in our earlier work [BS96, BS97a, BS97b] In this theory we are concerned with the control of one agent in a dynamic environment, where exogenous events beyond the control of our agent may occur. Our agent has an action theory (such as a theory in the language [GL93, KL94, San94, BGP96] where actions and their effects are specified ....

C. Baral and T. Son. Relating theories of actions and reactive control. Technical Report, CS Dept., University of Texas at El Paso (http://cs.utep.edu/chitta/chitta.html), 1997.

No context found.

C. Baral and S. C. Tran. Relating theories of actions and reactive control. Linkoping Electronic Articles in Computer and Information Science, 3(9), 1998. Available at: http://www.ep.liu.se/ea/cis/1998/009.

First 50 documents

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