Navigating
the indoors.

1. The architecture of indoor location tracking

Indoor positioning relies on a network of anchors (receivers/transmitters) placed throughout a building and tags (the devices being tracked). Unlike GPS, which requires a clear view of the sky, IPS must turn complex physical environments into data-rich platforms. This is achieved by translating radio frequency metrics into spatial coordinates.

1.1 Time of Flight (ToF) and Ultra-Wideband (UWB)

ToF calculates distance based on the absolute travel time of a signal. Because RF waves travel at the speed of light (c ≈ 3 × 10⁸ m/s), the distance d between a transmitter and receiver is strictly defined by:

d = c × Δt

For ToF to achieve centimetre-level accuracy, hardware clocks must be synchronised to the nanosecond. Ultra-Wideband (governed by the IEEE 802.15.4z standard) excels here. By transmitting extremely short nanosecond pulses over a wide spectrum (500 MHz+ bandwidth), UWB allows receivers to pinpoint the leading edge of a signal with extreme temporal resolution, making it highly resistant to indoor reflections.

1.2 RSSI and the log-distance path loss model

Bluetooth Low Energy (BLE) and Wi-Fi systems typically rely on the Received Signal Strength Indicator (RSSI). Instead of measuring time, they measure the decay of signal amplitude. The relationship between RSSI and distance is modelled using the log-distance path loss equation:

RSSI = −10n log₁₀(d) + A

Where:

• RSSI is the measured signal strength in dBm.
• n is the path loss exponent (typically 2.0 in free space, but 3.0 to 4.5 in complex indoor environments).
• d is the distance in metres.
• A is the reference RSSI measured at exactly 1 metre.

The core challenge with RSSI is that the variable n is highly volatile indoors due to physical obstructions, causing significant ranging errors if not filtered properly.

2. The physical challenges of indoor environments

While the theoretical applications of IPS are vast, real-world deployment reveals significant physical and logistical hurdles.

2.1 Multipath interference and Rayleigh fading

When an RF signal is transmitted indoors, it reflects off walls, floors, glass, and metal infrastructure. The receiver captures a superposition of the line-of-sight (LOS) signal and dozens of delayed, attenuated non-line-of-sight (NLOS) reflections. This phenomenon causes Rayleigh fading, where phase shifts in the reflected waves cause constructive and destructive interference.

2.2 Non-line-of-sight (NLOS) propagation

If the LOS path is blocked by a concrete pillar or a human body, the receiver will lock onto the first arriving NLOS reflection. This results in a positive distance bias: the system concludes the tag is further away than it actually is. Overcoming the Cramér–Rao Lower Bound (CRLB) in NLOS conditions remains the primary hurdle for hardware-only systems.

2.3 Infrastructure and maintenance overhead

Traditional BLE and UWB systems require dense grids of hardware anchors. For a large hospital or a 500,000 sq. ft. warehouse, this means deploying, calibrating, and eventually replacing the batteries of thousands of individual beacons, creating a massive operational bottleneck.

3. Future trends: the evolution of IPS

The next generation of indoor location tracking is shifting from hardware-heavy deployments to intelligent, software-defined networks.

• The convergence of 5G and UWB. 5G networks are bringing robust, high-bandwidth coverage indoors, while UWB provides centimetre-level precision. Integrating these technologies will enable seamless indoor–outdoor transitions without handoff delays.
• Infrastructure-light solutions. The industry is moving toward crowdsourcing signal fingerprints and using the ambient RF environment (existing Wi-Fi routers), drastically reducing the need for dedicated proprietary beacons.

4. Our solution: infrastructure-light sensor fusion

To bypass the limitations of RSSI volatility and multipath fading without deploying thousands of expensive UWB anchors, our architecture relies on an advanced sensor-fusion engine. By combining ambient RF signals (Wi-Fi/BLE) with pedestrian dead reckoning (PDR) data from a smartphone’s Inertial Measurement Unit (IMU), we create a highly accurate, self-correcting position estimate.

4.1 The Extended Kalman Filter (EKF) architecture

Our EKF operates in a continuous two-step loop, balancing sensor inputs based on their real-time reliability.

Step 1: Prediction (IMU data). Using the smartphone’s accelerometer and gyroscope, we predict the user’s next position based on current velocity and heading:

x_k|k−1 = F_k x_k−1|k−1 + B_k u_k

Here x is the state vector (position, velocity), F is the state transition matrix, and u is the control input from the pedometer.

Step 2: Update (RF fingerprinting data). Because dead reckoning drifts over time due to sensor noise, we use the ambient Wi-Fi/BLE fingerprint as a bounding measurement to correct the drift:

x_k|k = x_k|k−1 + K_k (z_k − H_k x_k|k−1)

K_k is the Kalman gain, dynamically weighting whether to trust the IMU prediction or the RF measurement based on calculated noise covariances.

4.2 AI-driven self-healing radio maps

Traditionally, RF fingerprinting requires manual site surveys. Our platform uses crowdsourced, anonymous telemetry. A Convolutional Neural Network (CNN) processes the continuous stream of RSSI vectors. If a venue moves a Wi-Fi access point, the CNN detects a localised divergence between the expected fingerprint and the actual readings, isolates the anomaly, and regenerates the spatial weights, creating a self-healing mesh with zero human intervention.

5. The business impact and ROI

By solving the physical limitations of IPS with a software-first approach, we unlock transformative operational capabilities across industries:

6. Conclusion

The future of indoor spatial intelligence does not require drilling holes in the ceiling or managing thousands of batteries. By leveraging existing ambient signals, advanced EKF sensor fusion, and self-healing neural networks, indoor positioning can finally scale globally. This software-first approach democratises location data, turning passive buildings into intelligent, reactive environments.