A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current trends in AI Planning, pages 266--277. IOS Press, Amsterdam, Netherlands, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....that behavior using the artifact operate with respect to the actual object, rather than with respect to an a priori assumption. We call anchoring the process of (1) matching a feature or object hypothesis to an artifact, and (2) updating the artifact using this perceptual information (see [26] for more on anchoring) In Saphira, the structure of decision making for the anchoring problem takes the following form: a) b) Figure 5: Anchoring a corridor artifact to sensor readings for corridor following. feature = hypothesis artifact = updated artifact F NaN F NaN Delta F NaN ....

A. Saffiotti, Pick-up what?, in: C. Backstrom and E. Sandewall, eds., Current Trends in AI Planning (IOS Press, Amsterdam, Netherlands, 1994).

No context found.

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current trends in AI Planning, pages 266--277. IOS Press, Amsterdam, Netherlands, 1994.

No context found.

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current trends in AI Planning, pages 266--277. IOS Press, Amsterdam, NL, 1994.

....close to ours, the problem they address is different from anchoring. Anchoring involves the connection between the symbolic descriptions of objects in the world and the perceptual data associated to these objects. Anchoring in this sense has been first defined by one of the authors of this paper [24] in the context of instantiating the arguments of plan operators in a mobile robot application. The latter work constitutes the starting point of the study that we propose here. The problem of connecting linguistic descriptions of objects to their physical referents has been largely considered in ....

A. Saffiotti. Pick-up what? In C. Bckstrm and E. Sandewall, editors, Current trends in AI planning, pages 266--277. 1994.

....aspect in integrating the decision making and the computer vision systems is the connection between the abstract representations used by the symbolic decision making system to denote a specific physical object, and the data in the computer vision system that correspond to that object. Following [5], we call anchoring the process of establishing this connection. We assume that the decisionmaking process associates each object to a set of properties that (nonunivocally) describe that object. Anchoring this object then means to use the vision apparatus to find an object whose observed features ....

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current Trends in AI Planning --- Proc. of EWSP '93, pages 166--177. IOS Press, Amsterdam, NL, 1994.

....component implicitly incorporate a solution to the anchoring problem, this solution is typical hidden in the code, and it is developed on a system by system basis on a restricted domain. To the best of our knowledge, the first domain independent definition of the anchoring problem was given in [Saffiotti, 1994] , while the first attempt at a computational theory of anchoring was reported in [Coradeschi and Saffiotti, 2000] The goal of this theory was to specify the functionalities and the representations needed to perform anchoring in a general way, that is applicable to a large number of systems. In ....

....the position of a given suitcase in included in the description state, then A will only be anchored to the specific suitcase at that position. The first setup encodes the action pick up a green suitcase, while the second one encodes the action pick up the green suitcase at the given position [Saffiotti, 1994] . Note that the object of the action is denoted by an indefinite reference in the first case, and by a definite one in the second case. 3.2 Partial Matching Some actions may affect the perceptual information which is gathered by the agent: for instance, moving closer to an object may allow the ....

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current trends in AI Planning, pages 266--277. IOS Press, Amsterdam, NL, 1994.

....and integrates them into an internal state S. In general, S integrates information about the part of the physical world which is relevant to the behavior, and which is to be used by the control law . The pivot of this integration is the notion of an anchor. Anchors were originally introduced in [13] as object descriptors, and have been further investigated in [14, 5] An anchor is a data structure in S which collects information about a specific physical object, such as the net in the above example. This information is used by the control law to generate a control value u. Perceptual ....

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current Trends in AI Planning, pages 166--177. IOS Press, 1994.

....work has been reported in that literature which analyzes this problem in a systematic way. Instead, anchoring is typically solved on a system by system basis, and the solution is hidden in the code. To our knowledge, the first general, although preliminary, definition of anchoring was reported in (Saffiotti 1994). More recently, the notion of anchoring has been extended in order to cope with some subtle issues encountered in real applications (Coradeschi, Karlsson, Nordberg 1999) Coradeschi Saffiotti 1999) and (Saffiotti LeBlanc 2000) Those studies were empirical in nature, leading to a ....

Saffiotti, A. 1994. Pick-up what? In Backstrom, C., and Sandewall, E., eds., Current trends in AI Planning. Amsterdam, Netherlands: IOS Press. 266--277.

....on, the action will take place with respect to the actually perceived corridor. The anchoring mechanism, together with the intrinsic tolerance of fuzzy rules, is essential to allow Flakey to operate by relying only on approximate maps and imprecise sensing. This mechanism is further explored in [13]. 3 Blending of behaviors In general, the agent will be engaged in complex activities involving the activation of several behaviors. Correspondingly, several behavior schemas can be simultaneously active in the fuzzy controller, each aimed at one specific goal (see Figure 1) The task of the ....

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current Trends in AI Planning --- Procs. of EWSP '93, pages 166--177. IOS Press, Amsterdam, Nederlands, 1994.

....processes to denote a specific physical object, and the data in 1 In Qualitative and Quantitative Approaches to Reasoning with Uncertainty, A. Hunter and S. Parsons editors, LNAI. Springer, Berlin, Germany, 1999 (in press) the low level processes that correspond to that object. Following [7], we call anchoring the process of establishing this connection. In general, we assume that the high level process associates each object in its universe of discourse to a unique name, and to a set of properties that (nonunivocally) describe that object. For example, an object named car 3 with ....

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current Trends in AI Planning --- Proc. of EWSP '93, pages 166--177. IOS Press, Amsterdam, NL, 1994.

....since control rules are based solely on the properties of the descriptor while the robot must perform the intended behavior with respect to the real feature. The process employed in our approach to maintain the correspondence between descriptors and environmental features is called anchoring (Saffiotti, MOBILE ROBOT CONTROL 1994). Figure 5.5 shows an example of anchoring during the execution of the Follow behavior. Initially (a) the robot is given a descriptor produced from prior map information of a corridor to follow (double lines) After a while (b) enough sonar readings are gathered to allow the perceptual ....

Saffiotti, A. (1994). Pick-up what? In Backstrom, C. and Sandewall, E., editors, Current Trends in AI Planning --- Proc. of EWSP '93, pages 166--177. IOS Press, Amsterdam, NL.

....precisely, each goal directed behavior in Flakey maintains in the state a descriptor of the object relevant to that behavior. For example, the Follow behavior maintains a descriptor of the lane to follow, and uses the rules in Figure 4 with respect to it. An explicit procedure, called anchoring [35], is employed to maintain the correspondence between descriptors and environmental features detected by the sensors. Which features should be used for anchoring depends on the specific instance of the behavior. For example, to go down a given corridor, we use the Follow behavior and anchor the ....

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current Trends in AI Planning --- Procs. of EWSP '93, pages 166--177. IOS Press, Amsterdam, Nederlands, 1994.

....Later, when the actual cup is perceived, we use the output from the perceptual system to continuously update the descriptor, effectively switching to a closed loop regime. We refer to the process of keeping object descriptors coordinated with physical objects through perception as anchoring [Saf94] Example 15 The Follow control schema introduced in Section 2.4 can be used to build a behavior schema for moving down a given corridor, provided that Flakey is in the corridor and it is not blocked by an obstacle: h [at(Corr1) facing(OG) Follow(Corr1) fCorr1, OGg i ; where we indicate in ....

A. Saffiotti. Pick-up what? In C. Backstrom and E. Sandewall, editors, Current Trends in AI Planning, pages 166--177. IOS Press, Amsterdam, Nederlands, 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