• Citer

Experimental Comparison of Radio Duty Cycling Protocols for Wireless Sensor Networks

par Uwase, Marie-Paule
Président du jury Van Biesen, Leo
Promoteur Dricot, Jean-Michel ;Steenhaut, Kris
Co-Promoteur Tiberghien, Jacques
Publication Non publié, 2018-10-30

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Wireless sensor networks are often battery powered and therefore their power consumption is of critical importance. Power requirements can be reduced by switching off radios when they are not needed and by using multi-hop communications to reduce the length of the radio links. Multi-hop communications however require message routing through the network. The Routing Protocol for lossy networks (RPL) has been designed by the Internet Engineering Task Force (IETF) for seamless integration of wireless sensor networks in the Internet. For switching on and off radios, radio duty cycling (RDC) protocols have been added to the traditional medium access control (MAC) protocols. Despite the fact they belong to different layers in the communications stack, it is intuitively clear that the choice of a specific RDC protocol for saving energy can influence the performances of RPL. Exploring experimentally this influence was the initial goal of this research. A 25 nodes wireless sensor network using Zolertia Z1 motes and the Contiki software was used for this investigation. Performance measurements without RDC protocol and with the three different RDC protocols readily available in Contiki were organized and the results of the experiments were compared. Unfortunately, with all three RDC protocols, serious malfunctions obscured the experimental results. Those malfunctions did not show up in absence of a RDC protocol and they could not be reproduced by our simulation studies. To tackle this issue, the behavior of the RDC protocols was scrutinized by means of experimental set-ups that eliminated as much as possible all non RDC related issues. Many, quite varied, malfunctions were discovered which all could have caused the observed RPL issues. Further research and better experimental set-ups made clear that all the discovered RDC malfunctions could be attributed to two real-world facts that were not considered by the implementers of the Contiki RDC protocols. The first cause is the small frequency difference between hardware real time clocks in stand-alone motes. The second is that the threshold built in the receiver to detect radio activity is much higher than the minimum level of signal that the same receiver can decode. Work-arounds have been designed for the observed malfunctions and they have been tested by means of a systematic comparison of the performance of the three modified RDC protocols.