F. Bacchus, J.Y. Halpern, and H. Levesque. Reasoning about noisy sensors and eectors in the situation calculus. Articial Intelligence 111(1-2), 1999.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... the situation calculus [24] to deal with time and continuous change [29, 13] exogenous (natural) actions [30] complex robot actions (plans) 22, 12] using sensing to determine which action to execute next [21, 20] as well as with probabilistic state descriptions and probabilistic action outcomes [2, 14]. 6.1 An Application: Probabilistic Plan Debugging In [4] we have developed a scheduling technique that enables autonomous robot controllers to schedule exible plans, that is plans that allow autonomous robots to exploit unexpected opportunities and to reschedule dynamically which applies the ....

F. Bacchus, J.Y. Halpern, and H. Levesque. Reasoning about noisy sensors and eectors in the situation calculus. Articial Intelligence 111(1-2), 1999.

....of new candidate plans under the updated belief. However, most probabilistic framework for reasoning about noisy sensors and e ectors are only concerned with one or the other of the two tasks. 2 To illustrate the subtle di erences between both tasks, we will use a simple example taken from [1]: a mobile robot moving along a straight line in a 1 dimensional world. Initially, the probability that the robot is at position 10 is 60 , otherwise it is at position 9 resp. 11 with probability 20 . The robot can move by means of the low level navigation routine noisyAdv(d) and can obtain an ....

....activate noisySensePos, if jd sensedj 2, execute noisyAdv(d sensed) 3 2 While probabilistic projection is central to probabilistic planners like C Buridan and MAXPLAN [5, 12] they ignore belief update. On the other hand, it is not clear how to project a plan in a framework like [1] for reasoning about belief update. While the theory of POMDPs are concerned with both tasks, the computational cost is prohibitive already in relatively small domains (e.g. 6] Finally, the recently proposed DTGolog [2] assumes full observability of the domain. 3 Here, sensed refers to the ....

[Article contains additional citation context not shown here]

F. Bacchus, J.Y. Halpern, and H. Levesque, `Reasoning about noisy sensors and eectors in the situation calculus', Articial Intelligence 111(1-2), (1999).

....1. Introduction and Motivation In this article we address the problem of representing and reasoning with theories of action in domains in which actions might have non deterministic probabilistic outcomes. This problem has been addressed in the literature by various researchers, for instance [1,10]. In previous work, reported in [8] we deal with actions that can have nitely many (disjoint) outcomes, and, therefore, a discrete probability distribution. The assumption that the possible outcomes is nite is very strong and often inadequate. For instance, if one is modeling the sensors and ....

....For this de nition, we appeal to the iota function. In this case, iota gives the value of n such that holds[bag[n] s] is true. This value has to be unique, in the implementation it gives the rst solution it encounters, if any. The Mathematica de nition is: iota : Function[z,c,Solve[c,z] [1,1,2]] The following expressions correspond to the speci cations of the axioms about the initial situation (s0) the iposs input preconditions and a de nition state constraint, which establishes whether or not the gambler is considered to be winning: holds[bag[n ] s0] n= ibconts iposs[i ,s ] ....

[Article contains additional citation context not shown here]

F. Bacchus, J. Halpern, and H. Levesque. Reasoning about noisy sensors and eectors in the Situation Calculus. Articial Intelligence, 111(1-2):171-208, 1999.

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