Method
GNSS is often unavailable indoors. BLE beacons estimate distance through RSSI or AoA; Wi‑Fi fingerprinting matches scanned AP lists with a pre-collected database. UWB can provide higher accuracy but at a higher cost. Combining map constraints with pedestrian dead reckoning (PDR) can smooth the trajectory.
With cellular
Positioning computation can be done at the edge or in the cloud; when asset tags report the last location over 4G Cat.1, power consumption and reporting frequency need to be considered.
Operations and maintenance
Fingerprint databases need updating after AP changes or renovations; beacon batteries require a bulk replacement plan.
Algorithm and engineering trade-offs
RSSI-based positioning is strongly affected by human blockage, multipath, and transmission power fluctuations, usually requiring Kalman filtering, particle filtering, or fusion with fingerprinting and PDR. AoA solutions require high demands on antenna arrays and calibration but can achieve better accuracy at the same anchor density. Maintaining a Wi-Fi fingerprint database is costly; crowdsourced collection and automatic differential updates can be used, but protection against malicious tampering is needed. UWB provides centimeter-level accuracy, suitable for industrial and warehouse applications, and requires collaborative planning with cellular gateways for upload bandwidth and tag power consumption.
With cellular modules and asset tracking
After the asset tag scans nearby information via BLE, the 4G Cat.1 gateway can report the 'last seen location' when leaving the area, balancing real-time performance and battery life. In scenarios such as underground parking lots where the gateway may have no cellular signal, offline events should be cached and uploaded in batches once connectivity is restored. Multi-tenant campuses should be mindful of AP and beacon ID conflicts, and the cloud must isolate projects accordingly. Regarding privacy, indoor tracking is equally sensitive, and authorization and anonymization strategies should distinguish between employees and visitors.
Engineering Practice Supplement
Focusing on the implementation of 'indoor positioning (BLE/Wi‑Fi)' in IoT and cellular module engineering, it is recommended to define network constraints (target market frequency and decommissioning plan, APN and private network policy, roaming and tariffs, PSM/eDRX power-saving modes) and terminal constraints (antenna and RF layout, power ripple and surge, thermal design and derating, EMC/safety regulations) as verifiable acceptance items during project initiation. During R&D, simulate weak networks, reproduce boundary failures with high/low temperature and voltage deviation tests, and retain logs and waveforms; in mass production, establish batch and calibration traceability, and handle component replacement through change review. After launch, include connection success rate, reconnection frequency, latency, and power consumption in the operations dashboard, with anomalies traceable to versions and batches.
Fingerprint coordinates and beacon calibration
Indoors there is no stable satellite geometry, so a position depends on references that people install. A BLE beacon estimates relative distance from received signal strength or angle of arrival. A Wi-Fi fingerprint matches the scanned access-point identities and the signal-strength vector against a coordinate library collected in advance. Both depend on the references not having moved since collection. The fingerprint coordinate system must match the floor plan, including origin, axis direction, and floor number. Beacon transmit power and advertising interval belong in the calibration record. Pedestrian dead reckoning fills the step length between two observations with an accelerometer and a gyroscope, but it drifts with the step model and can be pulled back only at the next trustworthy observation.
Deployment is worth it when the ceiling or the column grid can hold fixed anchors, the floor plan stays relatively stable during operation, and the service needs room level or shelf level rather than the road level used outdoors. If anchors are too sparse, the algorithm cannot separate distance from multipath. An asset tag that reports “who last saw me” over cellular only when it leaves an area has to accept that there is no continuous track for that interval. Where there is no public network underground, the gateway must be able to cache events.
Floor errors after renovation
Shelves, partitions, and glass curtain walls change the multipath. An old fingerprint still returns a coordinate that looks certain, while the person is already in the next room. After a beacon battery dies, its identity disappears from the library, and the matcher can place the remaining signal on the wrong floor. An access point that is replaced but keeps the old name changes the key of the fingerprint vector, and the whole floor shifts together. Body shadowing makes the strength at one spot differ with facing. Without heading or several samples, one scan is treated as another walk. When beacon identities from different tenants collide, the cloud uses an anchor from area A in area B.
Floor changes should be helped by barometric pressure or by a stair constraint, not by comparing only the strongest signal. The outdoor segment still belongs to GNSS. The indoor-outdoor boundary needs a clear hysteresis so the position does not hop at the door. When there is no anchor observation for a short time, limit the extrapolated distance in the spirit of dead reckoning. After a timeout, mark the position unknown instead of drawing onward at the last speed.
The calibration record should state the floor origin, the axis direction, beacon transmit power, and the collection date. After renovation or an access-point replacement, mark the old fingerprint invalid. It must not keep returning a coordinate that looks certain. When the signal disappears briefly, mark the position unknown, limit extrapolation with the step model, and stop extrapolating after a timeout. Isolate beacon identities of different tenants in the cloud. When a satellite fix and an indoor fix both exist at a doorway, set hysteresis so the two coordinates do not switch back and forth. When the gateway has no public network, cache the “last seen” event and upload it by sequence number after the link returns. Do not treat cache order as the person’s real walking order.
Put anchor batteries and access-point changes on the inspection list. Remove a dead reference from the library instead of leaving it to pull the match. Smooth strength swings caused by body shadowing with several samples. Do not treat one scan as a walk. Bring stairs or barometric pressure into the floor decision so the strongest signal does not land on the wrong floor. A position reported to the cellular gateway should carry the floor and a confidence.