🤖 OntoBOT: A Unified Ontology for Personal Service Robotics

📄 Official repository for the paper:
“An Ontology for Unified Modeling of Tasks, Actions, Environments, and Capabilities in Personal Service Robotics”
Martorana, Urgese, Tiddi, Schlobach (K-CAP 2025)

---

Overview

OntoBOT is a modular OWL ontology developed to formally describe the elements involved in task execution by service robots.
It enables reasoning about:

• What a robot can do (obot:Capability)
• Where it operates (obot:Environment)
• What physical objects afford (obot:Affordance)
• How tasks are structured and executed (obot:Task, pko:Action, pplan:Step)

This ontology bridges symbolic task descriptions and physical feasibility — allowing systems to reason about actionability based on context, embodiment, and capabilities.

---

📌 Purpose of the Paper

“Robots should not just know what to do, but whether they can do it.”

Current robotics ontologies often focus on isolated domains (e.g., perception, execution, planning).
OntoBOT addresses the need for interoperable, grounded, and reasoning-capable models by integrating and extending:

• DUL: foundational entities and agents
• SOMA: affordances and spatial reasoning
• PKO & PPlan: procedural task structure
• ROS.owl: robot software/hardware components

The paper evaluates OntoBOT using two household activities, four robots, and a set of competency questions executed via SPARQL.

---

Core Concepts

OntoBOT integrates and extends foundational ontologies to model how robots perceive, act, and reason.
Its design enables symbolic planning that is grounded in execution and environmental feasibility.

Concept	Description
obot:Agent	A robot, modeled with ROS nodes and capabilities
obot:Environment	A physically bounded space (e.g., kitchen), containing obot:Components
obot:Component	A physical object the robot can interact with (e.g., bowl, fridge)
obot:Affordance	An interaction possibility (e.g., grasping, pouring), relational between agent and object
obot:Task	A goal-oriented activity, composed of actions and steps
obot:Capability	An executable functionality (e.g., grasp), linked to affordances via enablesAffordance

📘 Ontology Schema

Figure: Core structure of the OntoBOT ontology, illustrating how agents, environments, tasks, affordances, and capabilities are semantically interlinked.
Based on obot, pko, pplan, soma, and ros ontologies.

---

Case Study: Kitchen Scenario

We model two robotic activities in a household kitchen, as inspired by the EUROBIN Coopetitions:

1. Prepare breakfast: serve cereal and orange juice
2. Reorganize kitchen: tidy items, load dishwasher

These were used to validate OntoBOT through competency questions, such as:

• ❓ What affordances are required by an activity?
• 🤖 Which robots (Tiago, HSR, UR3, Stretch) have those capabilities?
• ⚠️ What prevents a robot from completing a task?

🧪 Results were generated using manually instantiated RDF graphs and SPARQL queries, available below.

---

📂 Repository Contents

OntoBOT/
│
├── ontoBot/ # Ontology files (.ttl, .owl)
├── case-study/ # RDF data for robots, tasks, environment
│ └── cqs.ipynb # Jupyter notebook with SPARQL queries
├── docs/ # Diagrams and visualizations
├── LICENSE
└── README.md

---

🔍 Evaluation via Competency Questions

"Given this ontology, can we answer relevant task planning questions?"

✅ Yes. All questions from the paper were answered via SPARQL, such as:

• CQ1: What objects and affordances are involved in “prepare breakfast”?
• CQ4: Which robot is capable of all required actions?
• CQ6: What steps fail for Robot X due to missing capabilities?

See: case-study/cqs.ipynb