2019/06/26 by Madhura Thosar, Thosar, Madhura, Christian A. Mueller +17
Computer Science · #Advanced Image and Video Retrieval Techniques #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Multimodal Machine Learning Applications #Natural Language Processing Techniques #Robotics (cs.RO)
paper · pdf · doi:10.48550/arxiv.1906.11114
openalex publication_date 2019/06/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Tool-use applications in robotics require conceptual knowledge about objects\nfor informed decision making and object interactions. State-of-the-art methods\nemploy hand-crafted symbolic knowledge which is defined from a human\nperspective and grounded into sensory data afterwards. However, due to\ndifferent sensing and acting capabilities of robots, their conceptual\nunderstanding of objects must be generated from a robot's perspective entirely,\nwhich asks for robot-centric conceptual knowledge about objects. With this goal\nin mind, this article motivates that such knowledge should be based on physical\nand functional properties of objects. Consequently, a selection of ten\nproperties is defined and corresponding extraction methods are proposed. This\nmulti-modal property extraction forms the basis on which our second\ncontribution, a robot-centric knowledge generation is build on. It employs\nunsupervised clustering methods to transform numerical property data into\nsymbols, and Bivariate Joint Frequency Distributions and Sample Proportion to\ngenerate conceptual knowledge about objects using the robot-centric symbols. A\npreliminary implementation of the proposed framework is employed to acquire a\ndataset comprising physical and functional property data of 110 houshold\nobjects. This Robot-Centric dataSet (RoCS) is used to evaluate the framework\nregarding the property extraction methods, the semantics of the considered\nproperties within the dataset and its usefulness in real-world applications\nsuch as tool substitution.\n