[Paper Review] A ROS multi-ontology references services: OWL reasoners and application prototyping issues

Luca Buoncompagni, Alessio Capitanelli+1 more|arXiv (Cornell University)|Jun 30, 2017

Semantic Web and Ontologies7 references3 citations

TL;DR

This paper proposes ARMOR, a ROS-based multi-ontology service that integrates OWL reasoners into robot architectures via a scalable, thread-safe interface. It enables dynamic knowledge manipulation and reasoning through ROSJava, supporting synchronization, distributed access, and extensible reasoning via injected services, with validation in HRI and planning applications using real-world ontologies and OWL API compatibility.

ABSTRACT

This paper introduces a ROS Multi Ontology References (ARMOR) service, a general-purpose and scalable interface between robot architectures and OWL reasoners. ARMOR addresses synchronisation and communication issues among heterogeneous and distributed software components. As a guiding scenario, we consider a prototyping approach for the use of symbolic reasoning in human-robot interaction applications.

Motivation & Objective

• To address the lack of standardized, general-purpose integration between OWL-based symbolic reasoning and ROS robot architectures.
• To overcome limitations in existing ROS semantic frameworks that rely on ad hoc reasoning solutions and lack support for advanced reasoning paradigms.
• To enable dynamic TBox and ABox reasoning in human-robot interaction (HRI) scenarios, supporting runtime belief updates and learning.
• To provide a flexible, extensible, and scalable interface for managing multiple OWL ontologies with synchronization and consistency guarantees across distributed ROS components.
• To support prototyping of semantic reasoning in robotics by enabling integration with tools like ROSPlan and injected services for scene learning.

Proposed method

• ARMOR is implemented as a ROSJava service that exposes the AMOR library, a thread-safe, multi-ontology manager for OWL API and reasoners.
• The system uses a ROS message-based interface with structured commands (e.g., add, remove, query) to manipulate ontologies, supporting client identification and operation targeting via reference names.
• It supports a mounting/unmounting mechanism to ensure consistency during manipulations, preventing concurrent writes while allowing read queries.
• The service enables procedure injection via injected services, allowing custom logic to be injected into the reasoning pipeline for advanced use cases.
• It integrates with standard ROS components using the ROS Parameter Server for configuration and supports GUI-based ontology state visualization for debugging.
• The system supports multiple OWL reasoners via factory-based instantiation, ensuring compatibility with Pellet and other standard reasoners.

Experimental results

Research questions

• RQ1 How can OWL-based symbolic reasoning be effectively and consistently integrated into ROS-based robot architectures?
• RQ2 What architectural patterns enable scalable, distributed, and synchronized access to multiple OWL ontologies in real-time robotic systems?
• RQ3 How can dynamic TBox reasoning and runtime belief classification be supported in HRI applications?
• RQ4 What mechanisms ensure consistency and avoid race conditions when multiple components access and modify shared ontologies?
• RQ5 To what extent can ARMOR support advanced prototyping of semantic reasoning in robotics, such as in PDDL generation and learning from demonstration?

Key findings

• ARMOR successfully enables dynamic, synchronized manipulation and querying of multiple OWL ontologies within a ROS architecture using a standardized ROS message interface.
• The system supports runtime consistency checks and prevents conflicting modifications through a mounting/unmounting mechanism, ensuring data integrity during concurrent access.
• The integration with ROSPlan replaced its internal semantic data structure with ARMOR, demonstrating feasibility in high-level planning pipelines.
• The use of injected services enabled real-time scene learning and belief classification in a human-robot interaction scenario, where users could correct robot beliefs through dialogue.
• The tutorial at ROS Developers Conference 2018 confirmed ARMOR’s usability in practical robotics applications, including topological mapping and SLAM-based navigation.
• ARMOR’s compatibility with the OWL API and support for multiple reasoners (e.g., Pellet) ensures extensibility and long-term maintainability in diverse robotic systems.