Description
Guiding a user (or robot) through an indoor space, which combines indoor-localization (where am I?) with topological / geometric routing over an indoor map. The map itself — typically expressed in IndoorGML or an equivalent topological model — is the structural input. Navigation is distinct from localization: a perfectly localized user without a map cannot navigate, and a perfectly mapped space without localization cannot route. The thesis touches navigation only insofar as the same indoor topology is used for crowd-flow boundary conditions and for emergency-response egress planning.
Why it's hard
- Indoor maps are rarely available, machine-readable, and current at the same time.
- Multi-storey routing requires correct connectivity modeling (stairs, elevators, escalators).
- Dynamic obstacles (closed doors, crowded corridors) require live-data fusion.
- Accessibility constraints (wheelchair-compatible routes) require richer map semantics.
- Map updates lag building modifications by years in real deployments.
Common approaches
- IndoorGML-based topological routing graphs.
- Combined geometric / topological models for fine-grained guidance.
- Map-matched dead-reckoning for low-update-rate localization.
- Crowd-aware routing that down-weights congested edges.