Wireless sensor networks (WSNs) constitute a set of tiny electronic devices, powered by batteries with limited sensing, computation, and communications capabilities. These limitations cause the cross-layer energy efficiency optimization design for wireless power sensor networks, a fundamental issue. The most energy-consuming part of the sensor is the radio, and the amount of its energy largely depends on the power and duration of the transmission/receive states. In order to achieve optimal energy consumption in data collection, the sensed data is transmitted via multihop communication along the specified optimal paths to the data collectors; that is, the transmitting nodes that are distant from the receiver use their neighbors as the relaying nodes. Some other approaches to the optimization of power supply in WSNs include periodically putting the sensors in wake and sleep modes, according to a given fixed time schedule. A sequence of the transmit (T), receive (R) or sleep (S) states of a network node is called a duty cycle. Sometimes, the fixed duty cycle can cause long packet delivery latency, lower the network capacity, or increase the network congestion. One of the methods aimed at improving the performance of the network by maintaining the minimum level of energy consumption is the application of data collection protocols with dynamically changing duty cycles.

The adaptive data collection (ADC) protocol--the first fully implemented data-gathering protocol with uniform node addressing and dynamic duty-cycling that was tested in a real heterogeneous environment--is proposed in this paper. ADC is an extension of the pipelined data collection (PDC) protocol proposed by these authors in earlier works. In the PDC protocol, two new features were implemented (free node addressing and dynamic duty cycling). The ADC protocol assigns a randomly generated identifier (RID) and a grade to each sensor. The grade is a measure of the communication hop distance to the sink node. The pair (RID, grade) obtained this way replaces the MAC and IP, which are normally used in the link and network layers for network node addressing.

To manage the data-gathering process, the ADC protocol uses the request-to-forward/clear-to-forward (RTF/CTF) handshake. It is utilized in all stages of the ADC data-gathering process, including network initialization, topology establishment with free addressing, and data transmission with dynamic duty-cycling stages. The RTF/CTF handshake also allows the management of network access and the collision avoidance mechanism in the data transmission process. For randomly generated identifiers, sometimes the same RID is assigned to the nodes with the same grade. The ADC protocol has implemented a three-stage scheme that allows it to handle RID conflicts. To manage the duration of the transmission, the RTF packets contain a dynamic duty-cycling (DDC) flag. If a node with a given grade obtains from its child node the RTF packet with the DDC flag set, then it adjusts the duty cycle schedule by adding a proper sequence of T, R, S states such that the packets can be forwarded in a pipelined fashion and avoid signal interference. To evaluate the performance of the ADC protocol, the authors build an experimental heterogeneous network with six source nodes and one sink. In the tests and simulations, the packet delivery ratio, the average hop delivery latency, and the average node duty cycle were measured. The analysis of the results indicates better performance and load adaptivity in heterogeneous networks for the ADC protocol (over the PDC).

In the paper, the design and implementation aspects of the ADC protocol are covered in detail. Both the test and the simulation environments are described in a clear and readable way. The evaluation of performed experiments is illustrated with many diagrams, which makes it easier to understand the results of the experiments. A rich and up-to-date bibliography allows readers to learn more about the problems raised. This paper should be of interest to researchers and practitioners who intend to design energy-efficient heterogeneous wireless sensor networks.