Dissertation committee:
Λάζαρος Μεράκος,Ομότιμος Καθηγητής,ΕΚΠΑ
Ευάγγελος Ζέρβας,Καθηγητής,ΠΑΔΑ
Ευστάθιος Χατζηευθυμιάδης,Καθηγητής,ΕΚΠΑ
Γεώργιος Αλεξανδρόπουλος,Αναπληρωτής Καθηγητής,ΕΚΠΑ
Παναγιώτης Μαθιόπουλος,Καθηγητής,ΕΚΠΑ
Ιωάννης Σταυρακάκης,Καθηγητής,ΕΚΠΑ
Κωνσταντίνος Χριστοδουλόπουλος,Επίκουρος Καθηγητής,ΕΚΠΑ
Summary:
This thesis studies a class of wireless networks where charging is carried out wirelessly through one base station acting as an energy source.
Equipping each sensor with the appropriate circuitry enables us to effectively provide a stable power source in a cost-effective and easily implementable manner. However, like any wireless transmission, some factors can degrade the charging process, such as the quality of the transmission channel, the distance of the sensor from the power source, and possibly physical or technical obstacles between the transmitter and receiver.
Therefore, in the following chapters, we focus on problems related to the charging process and to what extent the specific charging method affects the quality of its operation, both as a network for detecting critical events and as a network for monitoring environmental parameters.
In the third chapter, we model the charging process considering that the base station transmits in the millimeter wave frequency band, where 5th-generation base stations expect to operate. The millimeter band ranges between 24GHz and 40GHz. This specific spectrum gives us several advantages since the directionality of the signal optimizes the charging process. In the same chapter, we propose a new communication protocol Based on the proposed protocol when the antenna directs the radiation beam at the sector where the sensor is located. The sensor first charges for a certain period and then communicates with the base station to send some basic measurements or to raise the alarm in case of an event.
Therefore, the first problem that we consider and presented in the 4th chapter focuses on optimizing the charging process from the side of the base station in case we have a slow variation of the transmission channel and rare critical events. In this case, the base station can derive reliable estimates of the charging level of the nodes through report messages from the nodes and proactively adjust the charging mode.
Based on the base station's information about each sensor's charge level, we proposed two algorithms that determine how the radiation beam moves between the sectors. The results prove that both algorithms outperform the classic procedures (round robin, random).
The problem we next introduce and present in Chapter 5 focuses on a network of sensors monitoring an area for potential events in the presence of high measurement noise.
In particular, the sensors should autonomously recognize an event, expressed as a change in the distribution of a measurable parameter, and transmit the relevant information to the base station.
However, in noisy environments, false alarms increase, causing nodes to erroneously transmit more frequently to the base station and exhaust their energy reserves.
We approach the problem as an optimal stopping problem. To overcome the effect of high noise levels on the environment measurements, we imposed, before the detection algorithm, a smoothing phase of the distribution of the measured parameter. We proposed two smoothing methods: a classical technique based on AR filters and a new one based on Beta Particle filters.
The effectiveness of each filter in the detection scheme was analyzed through simulations.
Finally, in the 6th chapter, we examine the case where the network periodically sends measurements from an area of interest.
Our goal is to highlight the features of such a network and propose a scheduling process for sending data from each sensor, considering its charge level.
The reason for this is that wireless power sensor networks cannot effectively deliver their measurements when specific conditions exist, such as a far sensor position or poor channel quality.
Therefore, the proposed solution is based on the Markov decision processes theory. The logic is that we allow the sensor autonomously to choose whether to transmit immediately or wait for the next period to be sufficiently charged.
Keywords:
Optimal Stopping Theory, Multi armed Bandit, Wireless Powered Sensor networks, mmWave
