A lot of research has already been performed in the field of energy consumption reduction in wireless sensor networks (WSNs), due to the many cases where network and device lifetime is of the utmost importance (e.g., for implanted devices in Wireless Body Area Networks[4]). Unlike in this article however, optimizations are usually performed at the Medium Access Control (MAC) layer, instead of at the physical layer. Also, optimization is often focused on one specific aspect of the network.

Different approaches have been followed in developing energy-efficient MAC protocols. As is stated in[5], two energy-saving approaches can be found in the literature: duty cycling and in-network aggregation. Adjustable duty cycling schemes for lower energy consumption are presented and simulated in[6, 7] discusses an energy-efficient MAC protocol based on ultra-low duty-cycle frame exchanges and scalable network self-configuration. The aggregation approach is mainly aimed at lowering the number of transmissions by using smart routing techniques, e.g., based on game theory[8]. In[9], an energy-balancing routing scheme is proposed for longer network lifetimes. An analytical study of the deployment of traffic-aware relay nodes is presented in[10]. In[11], an energy model for clustered multi-hop WSNs is derived and optimized with respect to the cluster-head selection process, hereby using a probabilistic method. A cluster-head is a node which collects data from other nodes and transfers it to the sink.

In[12], the authors have developed algorithms to minimize and balance energy consumption in WSNs with uniformly distributed sensor nodes, based on a sector-based multi-hop approach. In[13], schemes are proposed that can lead to energy savings up to 30–70%, by defining a cost function that also takes into account possible retransmissions. Zhang et al.[14] have focused on a realistic nonlinear battery model in a general two-hop relay network and developed the relay selection criterion from a battery energy efficiency perspective by following a theoretical and numerical approach.

Several papers perform a joint optimization on multiple layers: in both[15, 16], network lifetime is optimized on physical, MAC, and routing layers. The optimizations follow a theoretical–mathematical approach for a network with a single sink. An analytical approach for the determination of the optimal (common) transmit power in WSNs is presented in[17]. The influence of the exponent of the path loss model on the optimal transmit power is investigated, but just as in[15, 16], isotropic (one-slope) path loss models are assumed, which are often far from realistic in indoor environments with many different walls and wall types. Nonetheless, the correctness of these predictions is of major importance for the eventual performance of the proposed algorithms. In our research, an advanced indoor path loss model is used as the basis for all calculations. It takes into account the physical building layout and has been tested and validated in different indoor environments. Also, unlike in[15, 16], our optimization considers a topology with multiple sinks.

While cognitive approaches have already widely been applied to WSNs, they are usually used for spectrum-sensing purposes[18, 19], rather than for signal quality feedback, which is done in this article. Spectrum sensing is the process of recording the occupation of the different frequency bands and the variation of this occupation over time. Knowledge of the (un)occupied frequency bands allows choosing the frequency band that is most suited for packet transmission. A holistic approach to cognition and a framework that can help achieve end-to-end goals of application-specific sensor networks is provided in[20]. In[21], models and algorithms for self-optimization are presented, in networks that are interconnected through a broadband wireless mesh backbone network. An architectural approach is followed, where *positions* of backbone and terminal nodes are optimized for very large networks (e.g., 50 × 50 km^{2}). The self-optimization is based on a mathematical model and uses flocking algorithms and particle swarm optimization, a relatively slow-converging optimization method.

Very often, the proposed (mechanism for) power consumption reduction is only considered in a theoretical or analytical way, or is based on simulations. For example, in[22] a new metric for energy-efficient cooperative transmission is introduced and applied to a theoretical case. Simulations of an energy-efficient clustering algorithm are presented in[23]. In this article however, the framework is experimentally tested in an actual network.

Not much research has been conducted on symbiotic networking. In[24], cooperation of heterogeneous networks is investigated at the level of a vertical handoff mechanism, based on a cross-layer polynomial regression predictive received signal strength (RSS) approach. A symbiotic integration of heterogeneous wireless networks at application-level is proposed in[25], while[26] discusses cooperative multicast on the physical layer, applied to heterogeneous networks. To the best of the authors’ knowledge, this article is the first to combine the symbiotic networking aspect with a self-regulating physical-layer optimization.

To summarize, we can state that the existing literature on energy consumption reduction in WSNs mainly focuses on theoretical and analytical approaches and simulations. Also, in most of the articles, the optimization is performed on MAC or routing layers instead of on the physical-layer. In research that does take into account physical-layer aspects, the authors use very simple (and less accurate) isotropic path loss models, which are often unreliable in indoor environments. In this article, a framework for physical-layer optimization on multiple levels is proposed, based on an advanced and reliable network planner. Besides an optimal network planning including adaptable transmit powers, also a symbiotic optimization over different networks and network layers is implemented, a new concept in network cooperation. Moreover, a cognitive loop is added to the system, allowing self-regulation of the network planning process, improving network reliability, and adapting to varying propagation environments. Finally, our research does not only rely on theoretical calculations, but also an actual implementation of the optimization is experimentally tested in a wireless test network, increasing the contribution of this study. In the presented framework, advanced (energy) optimization strategies can be implemented and the framework can be used in cooperation with energy-efficient MAC protocols.