Saffiotti, A., Ruspini, E., and Konolige, K. (1993). Blending reactivity and goal-directedness in a fuzzy controller. In Proc of the 2nd IEEE Intl Conf on Fuzzy Systems, pages 134--139, San Francisco, CA. Online at http://www.aass.oru.se/asaffio/.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....policy to implement a rainwater pump management. system that maximizes pump service life during normal operation and transitions to a bsafety first policy during heavy rains. While EH may at first seem limited to supervisory control, several applications the Model Car [46] the Mobile Robot [30] and the Automatic Carder Landing System [35] use Exception Handling to generate control signals directly. In these cases, the control policy is tailored to the state space region which is characterized by the exception, advancing a regional goal rather than simply attempting to return to normal ....

....the exception condition. 414 PROCEEDINGS OF TIqE IEEE, VOL. 83, NO. 3, MARCH 1995 o. 8 : Lioguific Pzoary Conuol t: AT START: Corft Ahgnnt Fig. 5. Contro) po[ms expressed by graded trans,dons on role bases. Adapted from (35] Likewise, the Mobile Robot application [30], employs a strategic reach the designated target position control policy, and switches to a tactical avoid collisions policy when a previously undetected or randomly placed barrier is encountered. The pattern of rule firing in the FLC rulebases changes sharply as avoiding the collision ....

[Article contains additional citation context not shown here]

A. Saffiotti, E. Ruspini, and K. Konolige, "Blending reactivity and goal~directedness in a fuzzy controller," in Proc. 2nd IEEE Conf. on Fuzzy Syst., pp. 134-139, 1993 (see also [23]).

....the reported here. A combination between two separated fuzzy controllers (one for each actuator) and non fuzzy controllers in a navigation system can be found in [9] The major resemblance is with the work of Saffiotti, Ruspini and Konolige, which can be appreciated in several references, such as [1], 17] and [18] This work extends into other problems beyond navigation, as is task planning. 3 Fuzzy Behaviors A behavior is considered here as the reaction of the vehicle in the presence of some condition (environmental, in the typical situation) Why should fuzzy behaviors be used First of ....

....block in the left side of Fig. 2 is composed by a fuzzy predicate and the corresponding membership function whose intention is to measure how evident are the conditions of application of the behavior action in the present state of the environment, by returning a value belonging to the interval [0,1]. In other words, if it is considered that the behavior fuzzy controller codes some competence or know how, then the state variable measures the degree of membership of the environment state to the conditions of application of the fuzzy controller. The environment conditions to satisfy are the ....

[Article contains additional citation context not shown here]

Alessandro Saffiotti, Enrique H. Ruspini, and Kurt Konolige. Blending reactivity and goal-directedness in a fuzzy controller. In Proceedings of the Second IEEE International Conference on Fuzzy Systems, pages 134 139, San Francisco, CA, mar 1993.

....the mechanisms. Differences, that can shortly be stated here are: a) the fusion is done by weighted sum, instead of multiplic ation, which leads to further differences in the handling of priorities (or T norm) b) there is no priority inheritance and no hierarchy. Saphira In this architecture [SaKo93], through the use of fuzzy logic, reactive mechanisms and goaloriented mechanisms are smoothly blended into one sequence of control actions. Further, the concept of context dependent blending of mechanisms is an important feature of this architecture, providing a way to determine the current ....

Saffiotti, A., Ruspini, E. H., and Konolige, K. "Blending reactivity and goaldirectedness in a fuzzy controller", in Proc. IEEE Int. Conference on Fuzzy Systems, pp. 134-139, San Francisco, CA (1993).

.... provides tools for creating concurrent behaviors and, further, enforces a behavior based control structure [1] Similarly, COLBERT Saphira [7] which can also control the Pioneer robots (among others) is concerned mainly with the construction of fuzzily blended behavior based control systems [10]. While such tools are very useful, we believe that implementing them at such a low level imposes unnecessary restrictions on the programmer, who should have the choice to build any kind of control system while still enjoying device abstraction and encapsulation. Thus in Player we make a clear ....

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goal-directedness in a fuzzy controller. In Proceedings of the IEEE Intl. Conf on Fuzzy Systems, pages 134--139, San Francisco, CA, 1993.

....specification of control and of goal achievement by execution of a plan (i.e. a series of actions) are quite separated, therefore not much research has been devoted to this problem yet. Some interesting work has recently been done in this direction by Saffiotti, Konolige and Ruspini [SKR93, SRK93] We will return to this problem later in this paper. 3 Three layered software architecture Our approach to autonomous real time systems builds on the three layered software architecture (shown in Figure 1) developed in our group [HNS89, MS91, MNT OS92] since 1986. The three layers are ....

Alessandro Saffiotti, Enrique H. Ruspini, and Kurt Konolige. Blending reactivity and goal-directedness in a fuzzy controller. In Proc. 2nd IEEE Conf. on Fuzzy Systems, pages 134--139, 1993. 14

....relative importance [10,11,12] Also, to get more out of the activated behaviors, their actions may be blended to derive different control actions that take into consideration their respective goal and their current priority. Fuzzy logic has been successfully used with a robot by Saffiotti et al. [13,14,15] to select behaviors and to combine their control actions. The blending of the control actions of the behaviors allows smooth transitions between behaviors and can lead to more complex emerging behaviors. In this system, behaviors are activated according to a desirability measure obtained from a ....

....recommending behaviors has been inspired by the hedonic axiom which IEEE Int l Conf. on Fuzzy Systems, Barcelona, Spain, July 1997 indicates that the organisms direct their behaviors to minimize aversions and maximize desirable outcomes [2] This differs from the approach of Saffiotti et al. [13,14,15] which uses the notion of desirability to describe the control function of a behavior. In the architecture proposed, these measures make it possible to prevent possible conflicts when recommending behaviors. These conflicts may occur inside the same recommendation module or from the parallel ....

Saffiotti, A., Konolige, K. and Ruspini, E., "Blending reactivity and goal-directedness in a fuzzy controller", in Proc. IEEE Int'l Conf. on Fuzzy Systems, 1993, pp. 134-139.

.... to that of the DAMN behavior architecture developed by Rosenblatt [14] Similar methods have also been used in fuzzy controllers, where fuzzy control rules correspond roughly to behaviors, and defuzzification is analogous to behavior arbitration (for example, see Saffiotti, Ruspini, and Konolige [22]) Kuipers and Byun [12] have developed a spatial learning system that identifies distinctive places, as defined by a set of pre defined criteria (e.g. equal range readings in three directions) and links these places with edges specifying transit behaviors that take the robot from one place to ....

A. Saffiotti, E. Ruspini, and K. Konolige, "Blending reactivity and goal-directedness in a fuzzy controller," Proceedings of the IEEE International Conference on Fuzzy Systems, pp. 134-139, 1993.

....between behaviours may occur when the influence which one behaviour has on another one is propagated further to the rest of the behaviours via a network of interacting behaviours. The above observations are reflected in a number of other approaches to arbitration (Maes, 1991; Payton, 1990; Saffiotti et al. 1993). A network describing the influences (inhibition or enhancing) between behaviours is constructed in advance and used to propagate change in the activation level of one behaviour to the rest of the behaviours. Such a network of behaviours has basically two uses: i) to compute the activation levels ....

Saffiotti, Alessandro, Enrique H. Ruspini and Kurt Konolige (1993). Blending reactivity and goal-directedness in a fuzzy controller.

....activate competing behaviours simultaneously but to differing degrees, thus providing a smooth transition between them. This fusion has previously been proposed in the case of obstacle avoidance and navigation by (Baxter and Bumby, 1993) However, they suggest a mechanism which lacks generality. Saffiotti et al. 1993) present a more general method based upon an activation scheme for the behaviours. They consider activation levels for behaviours implemented by fuzzy rulebases. The activation levels are determined by fuzzy meta rules which arbitrate the dominance of different behaviours to the control outputs. ....

Saffiotti A. et al. (1993) Blending Reactivity and Goal-Directedness in a Fuzzy Controller. Proceedings of the Second IEEE International Conference on Fuzzy Systems, San Francisco, California, March 28 - April 1, 1993, pp. 134-139.

No context found.

Saffiotti, A., Ruspini, E.H., and Konolige, K. (1993) Blending reactivity and goal-directedness in a fuzzy controller. In: Proc. of the 2nd IEEE Int. Conf. on Fuzzy Systems, San Francisco, CA, 134--139. Online at http://www.aass.oru.se/~asaffio/

No context found.

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goaldirectedness in a fuzzy controller. In Proc. of the IEEE Int. Conf. on Fuzzy Systems, pages 134--139, San Francisco, California, 1993. IEEE Press.

No context found.

A. Saffiotti, E.H. Ruspini, and K. Konolige. Blending reactivity and goal-directedness in a fuzzy controller. In Proc. of the 2nd IEEE Int. Conf. on Fuzzy Systems, pages 134--139, San Francisco, CA, 1993.

No context found.

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goaldirectedness in a fuzzy controller. In Proc. of the 2nd IEEE Int. Conf. on Fuzzy Systems, pages 134--139, San Francisco, California, 1993. IEEE Press.

....mechanisms and strategies for intention reconsideration which work well when combined with the kind of low level control mechanisms required by agents which operate in complex dynamic environments. This paper describes one approach which combines a robust navigation system based on fuzzy logic [14, 15] and a BDI system for handling intentions. Before presenting the combination, however, we discuss the problem of intention reconsideration with respect to the formal model developed in [17] 2 THE FORMAL MODEL Following [17] our agents have two main data structures: a belief set and an ....

....is more complex. All Milou will have is a degree of belief, based on sensor input, that the door is open or closed. As discussed elsewhere, for example [1, 5, 11] handling this uncertainty requires sophisticated models. To solve these problems we turned to the use of Saffiotti s Thinking Cap [14, 15]. 4 FROM THEORY TO PRACTICE The Thinking Cap (TC) 3 is a system for autonomous robot navigation based on fuzzy logic which has been implemented and validated on several mobile platforms [14, 15] The main ingredients of the TC are: a library of fuzzy behaviours for indoor navigation, like ....

[Article contains additional citation context not shown here]

A. Saffiotti, E. H. Ruspini, and K. Konolige, `Blending reactivity and goal-directedness in a fuzzy controller', in Proceedings of the 2nd IEEE International Conference on Fuzzy Systems, pp. 134--139, San Francisco, CA, (1993).

....of context, a formula in fuzzy logic. When more then one behavior is activated, their outputs will have to be fused as discussed below. Fuzzy context rules have been initially applied by Sugeno et al. [27] to switch between flight modes in a fuzzy controlled unmanned helicopter; and by ourselves [25] in the mobile robot Flakey. Successively, fuzzy context rules have been incorporated in an increasing number of robots [30, 28, 16, 29, 12, 13, 15] It should be noted that the use of fuzzy rules for expressing behavior arbitration policies is independent from the way in which individual ....

....and Pin and Watanabe [18] use symmetric rectangles and COG. Both methods can be shown to be equivalent to a vector summation scheme. The majority of authors, however, have opted for a full fledged form of fuzzy fusion: combination of arbitrary fuzzy sets followed by a defuzzification step [32, 2, 25, 30, 16, 28, 15, 29]. 2 A similar distinction is commonly made in the field of data fusion, especially when fuzzy techniques are considered [6] Many a proposer of fuzzy command fusion have considered the problems that can arise from blindly applying COG or MOM defuzzification to the combined fuzzy set. In ....

[Article contains additional citation context not shown here]

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goal-directedness in a fuzzy controller. In Procs. of the 2nd Fuzzy-IEEE Conf., pages 134--139, San Francisco, California, 1993. IEEE Press.

....The graph at the bottom of Figure 4 plots the activation level of each behavior over time. The above example has shown how we can blend reactive behaviors, whose main task is to promptly react to certain perceived events, and purposeful behaviors, those that take explicit goals into consideration [15]. Context dependent blending of reactive and purposeful behaviors can provide an agent with the ability to perform goal oriented activities in an uncertain, dynamic environment. It may also help in compensating for the imprecision of the prior knowledge. Fig (a) b) c) a) Activation Turn ....

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goaldirectedness in a fuzzy controller. In Procs. of the 2nd IEEE Int. Conf. on Fuzzy Systems, pages 134--139, San Francisco, California, 1993. IEEE Press.

....manage an internal state used to detect and escape from limit cycles. More extensively developed autonomous robots have been equipped with a wider repertoire of different behaviors, covering all the elementary sub tasks that they need to perform. This is the case of the autonomous robots Flakey [106, 108], Marge [44] Moria [117] and Lobot [123] these robots include fuzzy behaviors for going to a given position, for orienting towards a target, for docking to an object, for crossing a door, and so on. Other authors have developed several types of fuzzy behaviors, but have only reported ....

....fuzzy logic. When more then one behavior is activated, their outputs will have to be fused as discussed in the next subsection. Fuzzy context rules have been initially applied by Sugeno et al. 113] to switch between flight modes in a fuzzy controlled unmanned helicopter; and by Saffiotti et al. [106] in the mobile robot Flakey. It should be noted that using fuzzy meta rules for expressing behavior arbitration policies is independent of the way in which individual behaviors are implemented that is, these do not need to be fuzzy. For example, Ghanea Hercock and Barnes [42] and Pan et al. ....

[Article contains additional citation context not shown here]

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goaldirectedness in a fuzzy controller. In Procs. of the IEEE Int. Conf. on Fuzzy Systems, pages 134--139, San Francisco, California, 1993. IEEE Press.

....impossible to achieve. 3 The Thinking Cap The Thinking Cap (TC) is a system for autonomous robot navigation based on fuzzy logic which has been implemented and validated on several mobile platforms. A full description of the TC can be found in [11] Parts of the TC were previously reported in [13, 12, 14]. The main ingredients of TC are: a library of fuzzy behaviours for indoor navigation, like obstacle avoidance, wall following, and door crossing; a context depending blending mechanism that combines the recommendations from different behaviours into a tradeoff control; a set of ....

.... Including this idea in our framework would lead to a tower of meta controllers similar to the one suggested in [15] In closing, we note that the example shown above has only been run in simulation although the navigation system alone has been extensively validated on several real robots [13, 12, 11]. We are aware that the actual verification of the ideas sketched in this paper will only come from intensive testing in real and challenging environments. We are currently in the process of implementing our integrated system on a Nomad 200, and we hope to be able to show the first experimental ....

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goaldirectedness in a fuzzy controller. In Proc. of the 2nd IEEE Int. Conf. on Fuzzy Systems, pages 134--139, San Francisco, California, 1993. IEEE Press.

....at higher levels of abstractions, through the use of symbolic planners or topological maps, and then have to be considered at the control level. The problem of how to integrate high level and low level representations and processes is one of the central issues in mobile robotics. In previous works [4, 3], we have proposed a technique to bring abstract goals into a fuzzy controller based on the use of object descriptors: the relevant properties of the object of an action (e.g. a door to cross) are extracted from the map and wrapped into a data structure the descriptor. This is then ....

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goal-directedness in a fuzzy controller. In Procs. of the 2nd Fuzzy-IEEE Conf., pages 134--139, San Francisco, CA, 1993.

....into a dynamic sequence of control actions. In particular, reactive and goal oriented behaviors are blended into a single sequence of control actions. These techniques have been first implemented on Flakey, the mobile robot platform of the Artificial Intelligence Center of SRI International (Saffiotti et al. 1993; Konolige et al. 1997) The illustrative examples that we present in this paper are drawn from our experience with this robot. The scope of these techniques extends, however, beyond this particular test bed, as they effectively deal with basic problems in the development of intelligent ....

Saffiotti, A., Ruspini, E. H., and Konolige, K. (1993). Blending reactivity and goal-directedness in a fuzzy controller. In Proc. of the 2nd IEEE Int. Conf.

....Sugeno and Nishida to drive a model car along a track [43] and the more recent behavior based fuzzy controllers included in the award winning robots Flakey [6] Marge [30] and Moria [44] In the rest of this section, we detail our use of fuzzy control to implement basic behaviors in Flakey. See [38, 40] for a more complete treatment. In our approach, we express desirable behavioral traits as quantitative preferences , defined over the set of possible control actions, from the perspective of the goal associated with that behavior. Following the formal semantic characterization of Ruspini [33, ....

....at planning time, as the planner does not know about the obstacles in the corridor. However, the context rules just re activate the Sense behavior when the need to do so arises. In this sense, our plans are similar to Schopper s universal plans [41] Following its implementation on Flakey [38], CDB has been used by several researchers in autonomous robotics [13, 26, 44, 45, 46] CDB provides a flexible means to implement complex behavior coordination strategies in a modular way using a logical rule format. This modularity simplifies writing and debugging of complex behaviors: the ....

A. Saffiotti, E. H. Ruspini, and K. Konolige. Blending reactivity and goal-directedness in a fuzzy controller. In Procs. of the 2nd IEEE Int. Conf. on Fuzzy Systems, pages 134--139, San Francisco, California, 1993. IEEE Press.

No context found.

Saffiotti, A., Ruspini, E., and Konolige, K. (1993). Blending reactivity and goal-directedness in a fuzzy controller. In Proc of the 2nd IEEE Intl Conf on Fuzzy Systems, pages 134--139, San Francisco, CA. Online at http://www.aass.oru.se/asaffio/.

No context found.

Alessandro Saotti, Enrique Ruspini and Kurt Konolige, Blending reactivity and goal-directedness in a fuzzy controller, Procs. of second IEEE conference on fuzzy systems, 1993, 134-139.

No context found.

Alessandro Saotti, Enrique Ruspini and Kurt Konolige, Blending reactivity and goal-directedness in a fuzzy controller, Proceedings of the second IEEE conference on fuzzy systems, 1993, 134-139.

No context found.

Saffiotti, A., Ruspini, E. H. and Konolige, K., "Blending Reactivity and GoalDirectedness in a Fuzzy Controller," Proceedings of the Second IEEE Conference on Fuzzy Systems, San Francisco, CA, March 1993, pp. 134-139.

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