RSSI Measurements
The goal of this experiment is to attain a near optimal beacon spacing to reduce cost of buying beacons, while preserving the highest possible accuracy. Since accuracy is dependent on many factors, we initially focus on estimating the maximum distance to one beacon. We expect to measure signal error caused by attenuation, which is the gradual loss of quantity which passes through a substance or medium.
Indoor Positioning is a wide area of research with many unsolved problems and different techniques. There is also an annual competition by Microsoft. Some research indicates that "[...] one should assume 1dB of attenuation per metre of an indoor office / residential environment.“ [1] Others claim „[…] that when the RSSI is above the sensitivity threshold (about -87 dBm), the packet reception rate (PRR) is at least 85%.“ [2] On the other hand performing these experiments in an indoor office can show that RSSI is a bad estimator of link quality. [3]
We've measured different combinations of beacons, phones and distances in our office. The beacons have been attached to a flat metal surface, about 0.5 meters above the floor, facing towards the phones (circles 1-3). The phones have been placed on foam cubes, with the screen up, facing towards the beacons (circle 4).
We've advertised iBeacon frames using different beacon models from BlueUp at an interval of 100 ms with the radio Tx power set to +- 0 dBm. Models used were the Maxi (circle 1), the Board (circle 2) and the Mini (circle 3). The key difference between these models is the power source. The Maxi and Mini use a printed meandered planar F-antenna while the board has an external antenna attached.
| Property | Value |
|---|---|
| Radio Version | 4.0 Bluetooth Smart (BLE) |
| Radio Frequency | 2.402 to 2.480 GHz |
| Radio Module | Nordic SoC nRF51822 |
We've scanned using Android devices from different manufacturers:
- BlackBerry
- HTC
- Nokia
- Huawei
- LG
- Motorola
- Samsung
- Xiaomi
All devices were unlocked, purchased in Germany, had WiFi disabled and only the BLE Indoor Positioning App (7c5bb1588e) running. Each measurement had a duration of 60 seconds to make sure that no temporary spikes were affecting the results.
Plotted below is the mean RSSI measured on multiple phones for each beacon model (solid line), as well as the standard deviation (semi-transparent area). All beacons had the same configuration. The used phones were the BlackBerry KEYone, Google Pixel 1, Google Pixel 2, HTC U11 and the Samsung Galaxy S8.
The measured RSSIs from the Board almost match with the expected logarithmic function for signal loss, while the Maxi and Mini don't. That might be related to the more sophisticated antenna that we attached to the Board. Notice the drop at 2 meters as well as the spike at 5 meters.
Plotted below is the RSSI of each advertising package received on different phones, advertised by the Board beacon.
The Galaxy S8 is the only model that produced two groups for each measured distance, resulting in a very high deviation. That makes it almost impossible to estimate distances based on the RSSI alone for this device.
Plotted below is the mean RSSI for different phones (solid line), as well as the standard deviation (semi-transparent area). It's similar to the chart above, just merging the individual RSSIs into a mean value.
Again, we see a drop at 2 meters as well as the spike at 5 meters. We can also observe that the Galaxy S8 had received the by far weakest signals with the highest deviation. If we'd measure an RSSI of -85 and had to estimate a distance, it could be anywhere between 0.5 and 10 meters for this device.
Plotted below is the mean RSSI of all phones (solid line), as well as the standard deviation (semi-transparent area).
Plotted below the the amount of received advertising frames per second for different phones.
A higher frequency allows us to better filter the RSSI values, resulting in a more stable distance estimation.
[1] R. Faragher und R. Harle, „An analysis of the accuracy of bluetooth low energy for indoor positioning applications“, in Proceedings of the 27th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+’14), 2014, S. 201–210.
[2] K. Srinivasan und P. Levis, „RSSI is Under Appreciated“, 2006, S. 5.
[3] K. Benkic, M. Malajner, P. Planinsic, und Z. Cucej, „Using RSSI value for distance estimation in wireless sensor networks based on ZigBee“, 2008, S. 303–306.