Robo-CSK-Organizer: Commonsense Knowledge to Organize Detected Objects for Multipurpose Robots

Rafael Hidalgo, Jesse Parron, Aparna S. Varde, Weitian Wang

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In the rapidly evolving field of robotics, integration of commonsense knowledge (CSK) in AI systems is becoming highly crucial to enhance the decision-making capabilities of robots, especially in next-generation multipurpose environments. This paper presents Robo-CSK-Organizer, a pioneering system that employs CSK, via a classical knowledge base, to facilitate sophisticated task-based object organization helpful in multipurpose robots. Unlike systems relying solely on deep learning tools such as ChatGPT, our Robo-CSK-Organizer system stands out in various crucial aspects. This includes (1) its ability to resolve ambiguities and maintain consistency in object placement; (2) its adaptability to diverse task-based classifications; and moreover, (3) its contributions to explainable AI (XAI), consequently helping to foster trust and human–robot collaboration. This system’s efficacy is underlined by DETIC (DEtector with Image Classes), an advanced extension of Detectron2 for object identification; BLIP (Bootstrapping Language-Image Pre-training) for context discernment; and most vitally by the adaptation of ConceptNet, a well-grounded commonsense knowledge base for reasoning based on semantic as well as pragmatic knowledge. While we deploy ConceptNet to extract CSK, the process in Robo-CSK-Organizer is generic enough to be replicated with other state-of-the-art knowledge bases. Controlled experiments and real-world applications, synopsized in this paper, make Robo-CSK-Organizer demonstrate superior performance in placing objects in contextually relevant locations, highlighting its clear capacity for commonsense-guided decision-making closer to the thresholds of human cognition. Hence, Robo-CSK-Organizer makes valuable contributions to Robotics and AI.

Original languageEnglish
Title of host publicationProceedings of IEMTRONICS 2024 - International IoT, Electronics and Mechatronics Conference
EditorsPhillip G. Bradford, S. Andrew Gadsden, Shiban K. Koul, Kamakhya Prasad Ghatak
PublisherSpringer Science and Business Media Deutschland GmbH
Pages65-81
Number of pages17
ISBN (Print)9789819747832
DOIs
StatePublished - 2025
EventInternational IoT, Electronics and Mechatronics Conference, IEMTRONICS 2024 - London, United Kingdom
Duration: 3 Apr 20245 Apr 2024

Publication series

NameLecture Notes in Electrical Engineering
Volume1228
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

ConferenceInternational IoT, Electronics and Mechatronics Conference, IEMTRONICS 2024
Country/TerritoryUnited Kingdom
CityLondon
Period3/04/245/04/24

Keywords

  • AI-robotics bridge
  • Commonsense reasoning
  • Explainable models
  • Multipurpose robots
  • Next-generation AI systems
  • Task classification

Fingerprint

Dive into the research topics of 'Robo-CSK-Organizer: Commonsense Knowledge to Organize Detected Objects for Multipurpose Robots'. Together they form a unique fingerprint.

Cite this