Location system for wireless sensor networks

In recent years a great amount of integrated radio circuits have come onto the market. With this type of component available a new kind of application has arisen: wireless sensor networks.

With these systems, sensors, instead of being inside the circuit itself, are installed in autonomous circuits. Thus, work on control and measurement can be distributed, e.g. the measurement of magnitudes over a geographically widespread area. Nevertheless, for the systems described here to be economically viable, each node has to have very low costs, both in its design and in its production. The most important advantage of this type of network is that of duplication: with so many sensors participating in the operation of the network, if one fails, another will fulfil the function until the failed item can be replaced.

However, organising co-operation between so many nodes is no easy task. Given that all nodes have the same hardware and the same software and, moreover, are limited both in energy consumption and in the capacity of the process, the protocols used in these types of networks have to be designed to operate in these very special conditions. In the case in question, the co-operation processes may be greatly simplified if the location of each node is known and how the network is organised geographically. The great number of nodes makes it impossible to fix the position of all these manually; an automated method for each node to calculate its own position needs to be found.

In this PhD thesis, a new algorithm for finding the position of the nodes is put forward and developed. To this end, the distances separating the nodes are utilised. However, given that each sensor has to be very economic, the quality of measurement of these distances is not expected to be high and, consequently, location errors appear. Thus, the algorithm proposed here attempts to calculate the best estimate of the node position, in the knowledge that the distances involved have errors.

Finally, to analyse the results obtained with this algorithm, a simulation platform has been designed which enables a comparison of the performance of the method described here with that of other algorithms put forward in recent years. In this way the computational load imposed at the node can be tested and how the presence of errors affects the measurements in the result obtained.

Moreover, the algorithm was implemented in a real node in order to demonstrate that it can be used in the environment for which it was designed.