This paper describes the Agilo RoboCuppers 1 team of the image understanding group (FG BV) at the Technische Universitat Munchen. With a team of four Pioneer 1 robots, equipped with CCD camera and a single board computer each and coordinated by a master PC outside the eld we participate in the Middle Size League of the fourth international RoboCup Tournament in Melbourne 2000. We use a multi-agent based approach to represent dierent robots and to encapsulate concurrent tasks within the robots. A fast feature extraction based on the image processing library HALCON provides the data necessary for the on-board scene interpretation. All robot observations are fused to one single consistent view. Decision making is done on this fused data. 1

. This paper describes the AGILO RoboCuppers 1 the RoboCup team of the image understanding group (FG BV) at the Technische Universitat Munchen. With a team of four Pioneer I robots, all equipped with CCD camera and a single board computer, we've participated in all international middle size league tournaments from 1998 until 2001. We use a modular approach of concurrent subprograms for image processing, self localization, object tracking, action selection, path planning and basic robot control. A fast feature extraction process provides the data necessary for the on-board scene interpretation. All robot observations are fused into a single environmental model, which forms the basis for action selection, path planning and low-level robot control. 1

Introduction Many tasks of pattern recognition require the use and manipulation of probabilistic data. Therefore, it is very often necessary not only to classify patterns, but also to find probability of pattern belonging to a specific class, i.e. it is necessary to obtain the distribution fimction of conditional probability. To solve this problem, a neural network can be used, typical representative of which is multilayer perceptron [1]. Because of the fact that the perceptron training is based on the minimization of the error, its output can be viewed as (1) estimation of probability, which is (2) approximated by the perceptron as a result of training. A good result can be achieved due to the artificial neural network (ANN) approximation abilities. 2 Probability estimation The way of interpretation of ANN outputs as probability estimation can be demonstrated by the following example. Let us assume that there are training vectors in a form: v = {xl, x2, y}, where xl, x2 - input par

©Hindawi, 2010. This is the author’s version of the work. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purpose or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the copyright holder. The definite version is published at the Journal of

Robots reason about abstract actions, such as go to position ‘l’, in order to decide what to do or to generate plans for their intended course of action. The use of abstract actions enables robots to employ small action libraries, which reduces the search space for decision making. When executing the actions, however, the robot must tailor the abstract actions to the specific task and situation context at hand. In this article we propose a novel robot action execution system that learns success and performance models for possible specializations of abstract actions. At execution time, the robot uses these models to optimize the execution of abstract actions to the respective task contexts. The robot can so use abstract actions for efficient reasoning, without compromising the performance of action execution. We show the impact of our action execution model in three robotic domains and on two kinds of action execution problems: (1) the instantiation of free action parameters to optimize the expected performance of action sequences; (2) the automatic introduction of additional subgoals to make action sequences more reliable. 1.