Klingspor, V. #1994#. GRDT: Enhancing model-based learning for its application in robot navigation. In: Proceedings of the 4th International Workshop on Inductive Logic Programming #S. Wrobel, Ed.#. Vol. 237 of GMD-Studien. Gesellschaft f#ur Mathematik und Datenverarbeitung MBH. pp. 107#122.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Muggleton (Oxford University) and Cao Feng (now Univ. of Ottawa) MF90 ] is a very efficient bottom up LGG learning algorithm based on the idea of ij determinacy which ensures polynomial search complexity. GRDT . Grdt, or grammar based Rdt, by Volker Klingspor from University of Dortmund [ Kli93 ] is a variant of Rdt that uses antecedent description grammars to specify search bias instead of rule schemata. The Grdt external tool interface is designed so that grammar rules for Grdt can be entered and manipulated through Mobal s user interface. In addition to these learning tools from ....

Volker Klingspor. Grdt: Enhancing model-based learning for its application in robot navigation. LS-8 Report 5, Universitat Dortmund, FB Informatik, Germany, 1993.

....can perform in this environment are limited. To achieve this level of granularity, e.g. with respect to an assembly robot, would require to severly restrict the actual capabilities of the machine. The individual results presented here have also been discussed throughout a number of publications [1, 31, 64, 63, 62, 65, 46, 71, 72, 74, 75, 73, 77, 76, 98, 99, 69, 54, 61, 60, 35, 56, 33, 93, 96, 94, 97, 95, 125, 124, 126, 153, 154, 156, 81, 90]. As in the other workpackages, several joint publications are among those [54, 61, 60, 33, 64, 63, 62, 65] 4.4.1 Learning from human demonstrations A common characteristics of all learning techniques that have been applied throughout the work done in workpackage 4 is that they are basically ....

.... learning how to calculate qualitativ data from quantitative data [156] ffl The special requirements for learning in robotics yielded in new developed learning algorithms, which are tailor made for the robot learning tasks, but are general enough to be used as usual logic based learners [81, 90, 74]. 5 Conclusions The development of learning systems and particularly learning robots is certainly an exciting field that also promises to become economically relevant in the very near future. B Learn II must therefore be considered as an important project that successfully combined European ....

Volker Klingspor. GRDT: Enhancing model-based learning for its application in robot navigation. In Stefan Wrobel, editor, Proc. of the Fourth International Workshop on Inductive Logic Programming, GMD-Studien Nr. 237, pages 107--122, St. Augustin, Germany, 1994. GMD.

....of perceptions and actions and represent objects and situation in the environment on a level of abstractions above the geometric one. Rules such as standing parallel moving standing through door [Kaiser et al. 1995] are generated directly from sensor data by employing an inductive learner [Klingspor, 1994b] They can be used to locate specific objects in the environment or to trigger specific actions of the robot, such as a door passing without collision avoidance, as soon as a door is recognized. Obviously, operational concepts have to take the actions of the robot into account as well [Klingspor ....

Klingspor, Volker. 1994b. GRDT: Enhancing Model-Based Learning for Its Application in Robot Navigation. Pages 107--122 of: Wrobel, Stefan (ed), Proc. of the Fourth International Workshop on Inductive Logic Programming. GMD-Studien Nr. 237. St. Augustin, Germany: GMD.

....target representation, i.e. programs, the concept definition language generated is not as easily predictable for a naive user, as it may be in scheme based or Clause Set approaches. Moreover, the order in which clauses are generated cannot be controlled as easily. ffl Combined approaches: GRDT [Kli94] and Dlab grammar as shown in section 3.2.4, combine the rule schemata of [EHR83] and Clause Set language. 3.4.2 Comparison of static versus generative approaches Many works on declarative bias concentrated on the study of the most efficient bias in terms of the number of hypotheses remaining ....

V. Klingspor. GRDT: Enhancing model-based learning for its application in robot navigation. LS-8 report 5, Universitat Dortmund, 1994.

.... algorithm, learning how to calculate qualitativ data from quantitative data [83] The special requirements for learning in robotics yielded in new developed learning algorithms, which are tailor made for the robot learning tasks, but are general enough to be used as usual logic based learners [44, 46, 38]. 6.4 Technical and methodological issues In contrast to most applications of machine learning, in B Learn II we did not focus on learn ing only on a specific level of abstraction (like only learning abstract descriptions or only learning reactive behavior) Instead, we integrated learning on ....

Volker Klingspor. GRDT: Enhancing model-based learning for its application in robot naviga- tion. In Stefan Wrobel, editor, Proc. of the Fourth International Workshop on Inductive Logic Programming, GMD-Studien Nr. 237, pages 107-122, St. Augustin, Germany, 1994. GMD.

No context found.

Klingspor, V. (1994a). GRDT: Enhancing model-based learning for its application in robot navigation. LS-8 Report No. 5, University of Dortmund, Lehrstuhl Informatik VIII, D-44221 Dortmund.

....problem, namely the learning of basic features (Section 4.1) Then, we give an overview of the representation hierarchy, based on the basic features. Finally, we show how learning can be applied to higher levels of the hierarchy. The used learning algorithm will not be described in detail (but see (Klingspor, 1994)) The reader may think of any learning algorithm for (restricted) first order logic because, in principle, any first order learning algorithm can solve our learning tasks. Note, that first order learning is necessary in order to handle time intervals and their relations. The classical learning ....

....they cover positive but not negative examples of the target feature. Until no negative (or only few) negative examples are covered, the next special rule schema is instantiated and tested. The most applicable learning algorithms were RDT (Kietz and Wrobel, 1992) and a modification of it, GRDT (Klingspor, 1994). In a nutshell, the results are as follows. Given 1004 examples for the four sensor features, we learned 129 rules, covering 87 of the given examples. For learning sensor group features, we had given 956 examples, from which GRDT learned 136 rules. Using the learned rules for sensor features and ....

Klingspor, V. (1994). GRDT: Enhancing model-based learning for its application in robot navigation. In Wrobel, S., editor, Proc. of the Fourth International Workshop on Inductive Logic Programming, GMDStudien Nr. 237, St. Augustin, Germany.

No context found.

Klingspor, V. #1994#. GRDT: Enhancing model-based learning for its application in robot navigation. In: Proceedings of the 4th International Workshop on Inductive Logic Programming #S. Wrobel, Ed.#. Vol. 237 of GMD-Studien. Gesellschaft f#ur Mathematik und Datenverarbeitung MBH. pp. 107#122.

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