One of the most important protocols in wireless sensor networks is the data collection or convergecast operation, in which large numbers of nodes continuously report data to a sink node. One important issue involved with convergecast operation is the management of the packet transfer process through the media access control (MAC) layer. One existing approach--time division multiple access (TDMA)--has strong synchronization requirements and less adaptability to changing traffic. Motivated by this, the authors have proposed an alternative lightweight MAC protocol called randomized data gathering (RDG). RDG aims to overcome the limitations of TDMA-based protocols while still trying to maintain the performance of “network lifetime, path delay, throughput, jitter, and scalability.” RDG uses the randomization in sensing and data transmission with the help of a transmission announcements process between adjacent nodes. This work contains both analytic and simulation experiment methods to validate the effectiveness of the RDG protocol, with regard to TDMA. The protocols still essentially apply only to low- and medium-sensing/traffic conditions.
What follows is the basic idea of the proposed mechanism. The RDG mechanism consists of two steps. In the first step, sensor nodes sense and then transmit the sensing data with a random delay (that is internally generated). It randomly allocates all generated packets (due to sensing) to be transmitted in a reporting period and is claimed to avoid significant risk of collisions. The randomization process is performed proactively (before any collision takes place) and is claimed as a collision preventive mechanism, as well as a way to avoid global synchronization (required in TDMA). In the second step, a mechanism follows a transmission data announcements process between adjacent nodes. This mechanism aims to reduce the effective node duty cycle. The core mechanism is as follows. Before a node transmits an internally generated packet, it produces the next instance of sensing/transmission and piggybacks it into that packet. “The piggybacked information can take the form of a ... relative time offset.” This is to satisfy “the goal of avoiding any absolute local time reference.” In this way, the receiving node is pre-informed of the delay until the next data transmission by the sender. Therefore, the receiving node “can switch to sleep mode for a significant time interval.” Next, when the receiving node wakes up to receive the next expected packet, it forwards the packet with the received time offset to its parent node. “This process is repeated until the packet reaches” the final destination (the base station).
Some partial theoretical analyses are done on the expression of network lifetime in this data generation and communication model and on throughput performance with an exponential randomization process. In the simulation experiment on the QNAP2 discrete event simulator, RDG is compared to TDMA in terms of network lifetime, average data reporting time, packet loss rate, and throughput. In addition, the jitter and scalability performance of RDG is evaluated.
Overall, the ideas of controlled randomization in data sending and the pairwise proactive exchange of future data transmission time instances are interesting. This approach can be useful for maintaining large networks and varying traffic conditions. However, it will apply only to a low- and medium-sensing/traffic rate so that randomization in data sensing is tolerable. Although there have been numerous works on MAC protocols, this approach compares well with TDMA methods. However, RDG is not evaluated with the class of transmitter-initiated MAC protocols. Moreover, it remains interesting to see how receiver-initiated MAC protocols are compared with RDG. In addition, the theoretical analysis of network lifetime could be completed in terms of a final expression and compared with that of other MAC protocols (or at least TDMA). Nevertheless, the two key ideas of this paper, as mentioned earlier, will be useful for interested readers.