OntoNav: A Semantic Indoor Navigation System

Christos Anagnostopoulos, Vassileios Tsetsos, Panayotis Kikiras, Stathes HadjiefthymiadesMCMP@MDM 2005

While outdoor navigation has been largely revolutionized by GPS and Geographic Information Systems (GIS), indoor navigation remains a significant challenge. Most existing indoor routing systems rely exclusively on geometric information, calculating the shortest path between two points. However, this "one-size-fits-all" approach often ignores the semantics of the building and the diverse needs of its users. OntoNav addresses this gap by introducing a user-centric system inspired by "Ambient Intelligence" and "Design for All," computing not just the shortest path, but the most traversable one based on individual physical and cognitive profiles.

System Architecture

OntoNav shifts the paradigm from purely geometric routing to a hybrid geometric-semantic approach. The system architecture is divided into two primary engines that work in tandem to produce personalized routes.

The OntoNav system architecture, illustrating the interaction between geometric and semantic services.

Geometric Path Computation Service (GEO)

This service utilizes a spatial database (such as PostGIS) containing detailed building blueprints to construct a path graph. To manage the complexity of multi-story environments, OntoNav employs a Hierarchical Clustering Graph. In this model, individual floors are treated as clustered sub-graphs connected by vertical passages like elevators and stairways.

Semantic Path Selection Service (SEM)

The core innovation of OntoNav is the Semantic Path Selection Service. This engine uses the Indoor Navigation Ontology (INO) to model users, spatial elements (passages, rooms, obstacles), and "anchors" (landmarks). By reasoning over these semantic layers, the system can filter geometric paths to find those that match a user's specific capabilities.

The Indoor Navigation Ontology (INO)

The INO provides the formal vocabulary required for semantic reasoning. It categorizes spatial elements and their relationships, allowing the system to understand the properties of the environment beyond simple coordinates.

The Indoor Navigation Ontology (INO) hierarchy, showing the classification of path elements and passages.

As shown in the ontology hierarchy, elements like HorizontalPassage and VerticalPassage are further refined into specific types such as Stairway, Elevator, or Ramp. This granularity allows the system to apply physical navigation rules. For example, if a user profile indicates wheelchair reliance, the SEM service can automatically exclude any path containing a Stairway instance.

The Routing Pipeline

The OntoNav routing process operates through four distinct reasoning tasks:

  1. Set-to-Set Mapping: The user's current location and destination are mapped to sets of nearby exits or passages rather than strict points.
  2. Greedy Traversal: The GEO service identifies all geometrically walkable paths between the starting and ending sets.
  3. Semantic Filtering: The SEM service applies physical navigation rules to filter out paths the user cannot traverse (e.g., excluding stairs for mobility-impaired users).
  4. Anchor Selection: Finally, the system applies perceptual navigation rules to select the best landmarks (anchors) for guidance. A blind user might receive auditory descriptions of anchors, while a child might be guided by visual or photographic landmarks.

Key Insights and Results

A crucial distinction made by OntoNav is the difference between a walkable path and a traversable one. A geometrically walkable path is merely a baseline; it only becomes traversable if it satisfies the physical and cognitive constraints defined in a specific user's profile.

Furthermore, the system treats "obstacles" dynamically. Because it utilizes an ontology, obstacles are not limited to physical walls; they can include policy-based or temporary constraints, such as a locked door, a broken elevator, or a crowded hallway that would be difficult for a visually impaired person to navigate.

While the current research focuses on architectural design and theoretical modeling rather than quantitative benchmarks, it qualitatively demonstrates how point-to-point geometric routing can be transformed into a sophisticated semantic reasoning problem. By defining specific rules (e.g., If user cannot walk and path contains stairs, exclude path), the authors show that standard graph traversal can be enriched to support highly specialized accessibility use cases.

Limitations and Future Work

The system currently faces challenges regarding computational complexity, as the greedy graph traversal computes all walkable paths before filtering. Caching these paths is risky due to the dynamic nature of indoor environments (e.g., sudden maintenance or broken facilities).

Future development will focus on a Navigation Assistance Module (NAM) to provide reactive guidance if a user deviates from a path, as well as automated methods for ingesting standard building blueprints into the spatial database for easier deployment.

Summary updated

BibTeX
@inproceedings{anagnostopoulos2005ontonav,
  author = {Christos Anagnostopoulos and Vassileios Tsetsos and Panayotis Kikiras and Stathes Hadjiefthymiades},
  title = {{OntoNav: A Semantic Indoor Navigation System}},
  booktitle = {MCMP@MDM},
  year = {2005},
}
Figure