30 research outputs found
Neural competition for motion segmentation
No full textSteffen JF, Pardowitz M, Steil JJ, Ritter H. Neural competition for motion segmentation. In: 18th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN). Bruges (Belgium): d-side; 2010: 59-64
Elements of robot learning: from the skill to the task level
No full textPardowitz M. Elements of robot learning: from the skill to the task level. In: Workshop: Bridging the gap between high-level discrete representations and low-level continuous behaviours. 2009
Using Structured UKR Manifolds for Motion Classification and Segmentation
No full textSteffen JF, Pardowitz M, Ritter H. Using Structured UKR Manifolds for Motion Classification and Segmentation. In: Intelligent Robots and Systems (IROS). 2009: 4785-4790.Task learning from observations of non-expert human users will be a core feature of future cognitive robots. However, the problem of task segmentation has only received minor attention. In this paper, we present a new approach to classifying and segmenting series of observations into a set of candidate motions. As basis for these candidates, we use structured UKR manifolds, a modified version of unsupervised kernel regression which has been introduced in order to easily reproduce and synthesise represented dextrous manipulation tasks. Together with the presented mechanism, it then realises a system that is able both to reproduce and recognise the represented motions
Self-emerging Action Gestalts for Task Segmentation
No full textPardowitz M, Steffen JF, Ritter H. Self-emerging Action Gestalts for Task Segmentation. In: 32nd German Conference on Artificial Intelligence (KI-2009). Berlin, Heidelberg: Springer; 2009: 589-596
A Manifold Representation as Common Basis for Action Production and Recognition
No full textSteffen JF, Pardowitz M, Ritter H. A Manifold Representation as Common Basis for Action Production and Recognition. In: 32nd German Conference on Artificial Intelligence (KI-2009). Berlin, Heidelberg: Springer; 2009: 607-614
Determination of Spatial Scale in Martian Landscape Images Acquired by the Curiosity Rover, and Viewing Image Scale and Target Chemistry Using the ASIC Website
No full textAbstractIn this paper we describe a method to compute spatial scales for images acquired by NASA's Mars Curiosity rover (Mars Science Laboratory, MSL). The method is based on the assumption that the rover stands on an infinite plane that may have any orientation with respect to the local gravity vector. While not new, it is the first time that this method is systematically applied to Martian images acquired by a lander. A continuously run software pipeline processes the images acquired by the rover within a 20 m radius, adds approximate scalebars to the raw images, and generates, whenever possible, rectified (warped) versions of those images. The products of this software pipeline and the chemical compositions of relevant rover science targets from NASA's Planetary Data System archive, are made available to the public via the Approximate Scale for Images and Chemistry website, which has been developed in collaboration with the Planetary Data System Analyst's Notebook for the MSL mission. Hyperlinks connect the two resources.Plain Language Summary: We developed a software pipeline that calculates the spatial scale of images acquired by NASA's Mars Curiosity rover. The software pipeline is linked to a new website: the Approximate Scale for Images and Chemistry, in which the scalebar products are paired with information about the shape, size, color, and chemical composition of the imaged site, obtained by the rover suite of instruments. The images mimic the vantage point of human eyes and are therefore well‐suited to inspire field geologists (including those mainly working on Earth) to interpret Martian geologic features.Key Points: A systematic method to generate approximate scalebars for obliquely acquired Martian landscape images was developed. A newly created Approximate Scale for Images and Chemistry (ASIC) website links images, color, spatial scale, and chemistry, as returned by NASA's Curiosity rover in Gale crater. The ASIC website is complementary and strongly linked to the Analyst's Notebook, the data resource for Martian/lunar landed missions.Deutsche ForschungsgemeinschaftProject DEALhttps://asic.mps.mpg.de/https://an.rsl.wustl.edu/msl/http://pds-geosciences.wustl.edu/msl/msl-m-chemcam-libs-4_5-rdr-v1/mslccm_1xxx/data/moc/http://pds-geosciences.wustl.edu/msl/msl-m-chemcam-libs-4_5-rdr-v1/mslccm_1xxx/extras/rmi_mosaics/http://pds-geosciences.wustl.edu/msl/msl-m-chemcam-libs-4_5-rdr-v1/mslccm_1xxx/extras/rmi_contours_in_mcam_images/http://pds-geosciences.wustl.edu/msl/msl-m-apxs-4_5-rdr-v1/mslapx_1xxx/extras
Integrating Feature Maps and Competitive Layer Architectures For Motion Segmentation
No full textSteffen JF, Pardowitz M, Steil JJ, Ritter H. Integrating Feature Maps and Competitive Layer Architectures For Motion Segmentation. Neurocomputing. 2011;74(9):1372-1381
Identification of High-level Object Manipulation Operations from Multimodal Input
No full textBarchunova A, Franzius M, Pardowitz M, Ritter H. Identification of High-level Object Manipulation Operations from Multimodal Input. Presented at the IASTED International Conferences on Automation, Control, and Information Technology
Evaluation of Tactile Features for Object Categorization
No full textSchöpfer M, Pardowitz M, Ritter H. Evaluation of Tactile Features for Object Categorization. In: Workshop on Tactile Sensing in Humanoids @ HUMANOIDS. Paris; 2009.To integrate robot manipulators in every day life, robots need to be able to handle complex situations and scenarios ideally up to the level of humans. The processing of sensory data is vital for a good performance in unstructured environments. This paper presents an approach to handle spartio-temporal tactile time-series. A set of features is extracted in an off-line experimental setup using a tactile database. The results are then verified using the Schunk Dexterous Hand (SDH-2) in a short experimental setup
Using a Piezo-Resistive Tactile Sensor for Detection of Incipient Slippage
No full textSchöpfer M, Schürmann C, Pardowitz M, Ritter H. Using a Piezo-Resistive Tactile Sensor for Detection of Incipient Slippage. In: International Symposium on Robotics. 2010: 1-7.The detection of incipient slip is an important cornerstone in tactile based grasping. In this paper, we present an approach to detect incipient slip using a fast piezo-resistive, yet static tactile sensor pad. Our approach renders special slip sensors obsolete and therefore enables static and dynamic sensing with one sensing mechanism. For the detection of the slip, a fast fourier transform is used to pre-process the data. In a subsequent step, a standard artificial neural net is trained on the data from the frequency domain to detect slippage, as well as to discriminate different surface textures
