Localising end-users


In this scenario the smart room adapts its operation to the number of people inside it; e.g. the more people are inside the room, the more often the ventilation will operate in order to circulate fresh air. Furthermore, the smart room will try to meet the personal preferences of each user under the condition that no-deadlocks or race conditions will occur.

We conduct our experiments in the smart/green building automations test-bed at the University of Patras. We use one of the offices to test the performance of our in-door localisation algorithm in real life experiments. The office is relatively small, with many furniture, and a WiFi router installed.

For our implementation we use the TelosB motes byMEMSIC, that are generally characterised as highly constrained devices. We programmed them using the TinyOS 2.x from the SVN trunk of the TinyOS project. We deploy the motes in a small office with dimensions 3.50m × 3.50m. Four of the motes, that have the role of the backbone WSN of the smart building, are located in fixed and predefined positions at the corners of the room at a height of 1.5m from the floor. These motes are called anchor motes and virtually tessellate the room in quadrants. We note that in the case of smart buildings, the positions of the sensor motes consisting the backbone WSN can be carefully engineered. Therefore, the assumption that the positions of the anchor motes are well known is well justified. We enumerate the quadrants from the ids of the corresponding anchor motes.

We set the radio’s transmit power to 2.9dBm as well as the MAC layer of the TinyOS not to perform any retransmissions in case of unsuccessful communication (the default number of this parameter is set to three retransmissions). The moving mote every 50ms sends out a burst of 20 localisation messages to the anchors. We further tessellate each quadrant of the room in four smaller quadrants, so as the room is divided in sixteen partitions. For the experiments, we place the mobile mote at the center of each partition and allow the algorithm to calculate its current position for the next 25 bursts. This procedure is repeated three times.

The figure above depicts the communication success rate between the moving mote and each one of the anchors across the whole room area. Higher success rates are marked with darker grey shade. As we see in general the closer the moving mote is to an anchor mote, the higher the corresponding success rate. However, furniture topology and ambient electromagnetic noise affect the localisation process in some areas of the room; in our case the quadrant corresponding to anchor 3. Figure 3 depicts the overall percentage of successful localisation for each anchor. We notice that in most cases the algorithm is able to successfully determine the correct position of the moving mote. Quadrants 2 and 3 seem to be most affected by the topology of the room, however their success rate remains very high (more than 95%).


Last month January 2019 Next month
week 1 1 2 3 4 5
week 2 6 7 8 9 10 11 12
week 3 13 14 15 16 17 18 19
week 4 20 21 22 23 24 25 26
week 5 27 28 29 30 31