Indoor positioning challenges
With the rapid rise of IoT technology, there is an increased demand for indoor positioning solutions. Existing technologies such as GPS and GLONASS rely on the assumption that the transmitting satellite has a direct line of sight with the receiving device. Obviously this does not hold true for indoor spaces, as the view of the sky is obfuscated by the building in most cases. Therefore, such technologies cannot be reliably used for indoor positioning or scenarios where the line of sight with the satellite is not direct.
Let’s take a step back and briefly explain how GPS works. The technology relies on specialized satellites in Earth's orbit which carry stable clocks. Furthermore, we know their position precisely. Each satellite transmits his location and time via a radio signal. The receiver picks up signals from at least four different satellites and combines the received data (satellite positions and the timestamps of when the data packets were sent) to work out its own location.
The same principles can be applied to the indoor positioning problem. We place transmitting devices on fixed locations inside buildings and mobile receivers will read those emitted signals and use them to work out their own location. We could also take the dual approach and reverse the roles. That is, place receiving devices on fixed locations and make the transmitter mobile. In either case the problem in hand is to localize an agent moving inside an environment.
The mathematics of both cases is analogous but there are more practical considerations to take in account. We will compare those in some future posts.
Now that we have identified the two main approaches, let’s discuss the actual strategy we use to work out the agent's location. We will use the second case, where the agent is carrying the transmitting device and the receivers are fixed, to explain the way this strategy works, but bear in mind that the same holds true for the dual approach, with the roles reversed.
To meet the demand for a large number of agents to be localized in real-time, a sensible approach is to leverage on existing technologies a typical agent might carry, namely Bluetooth, or more specifically Bluetooth Low Energy. Most modern mobile phones have Bluetooth radios, but we can also use a simple BLE beacon.
That is, a low cost, coin-sized device, with a battery, which incorporates a CPU, a Bluetooth radio, and a possible array of sensors, such as accelerometers, magnetometers, gyro, GPS, NFC, RFID, etc. A beacon will periodically broadcast it’s identity, while the receiver listens for those broadcasts.
Once a signal has been scanned, we measure its strength. This measurement is called the RSSI (Radio Signal Strength Indicator). Based on the RSSI, we can estimate the distance between the agent and the receiver, and if the broadcast has been scanned by at least three different receivers, using the estimated distances, we can work out the agent's location via standard multilateration techniques.
However, this approach relies on an assumption that the estimated distance, from the RSSI measurement, approximates the true distance between the agent and receiver accurately. In reality, the RSSI can be highly affected by the indoor environment, which introduces noise and uncertainty in the measurement. In this post, we won't delve deeper into the physics of how and why this happens. Nonetheless, the error from the RSSI measurement propagates through to the estimated agent's location, and can never be completely avoided. However, we can be more clever in the way we estimate the distance between the agent and the receiver from the RSSI measurement.
One way to achieve this is by keeping the error controlled and accounting for it accordingly. Furthermore, we can incorporate a-priori knowledge of the agent’s moving dynamics into the positioning algorithms and achieve a positional accuracy of up to 1 meter.
This is much better than GPS which in most cases achieves a 5-15 meter accuracy. A further challenge lies in the fact that indoor spaces are often dynamic environments with both the agents and obstacles moving and changing constantly. Effectively, this makes the environment geometry dynamic and radiofrequency characteristics non-constant. Thus for the best results, a good positioning strategy must take these considerations into account.