Green Communication for Next-Generation Wireless Systems: Optimization Strategies, Challenges, Solutions, and Future Aspects

Wireless sensor networks (WSNs) have emerged as a backbone technology for the wireless communication era. The demand for WSN is rapidly increasing due to their major role in various applications with a wider deployment and omnipresent nature. The WSN is rapidly integrated into a large number of applications such as industrial, security, monitoring, tracking, and applications in home automation. The widespread use in many different areas attracts research interest in WSNs. Therefore, researchers are taking initiatives in exploring innovation day by day particularly towards the Internet of Things (IoT). But, WSN is having lots of challenging issues that need to be addressed, and the inherent characteristics of WSN severely affect the performance. Energy constraints are one of the primary issues that require urgent attention from the research community. Optimal energy optimization strategies are needed to counter the issue of energy constraints. Although one of the most appropriate schemes for handling energy constraints issues is the appropriate energy harvesting technique, the optimal energy optimization strategies should be coupled together for effectively utilizing the harvested energy. In this high-level systematic and taxonomical survey, we have organized the energy optimization strategies for EH-WSNs into eleven factors, namely, radio optimization schemes, optimizing the energy harvesting process, data reduction schemes, schemes based on cross-layer optimization, schemes based on cross-layer optimization, sleep/wake-up policies, schemes based on load balancing, schemes based on optimization of power requirement, optimization of communication mechanism, schemes based on optimization of battery operations, mobility-based schemes, and finally energy balancing schemes. We have also prepared the summarized view of various protocols/algorithms with their remarkable details. This systematic and taxonomy survey also provides a progressive detailed overview and classification of various optimization challenges for the EH-WSNs that require attention from the researcher followed by a survey of corresponding solutions for corresponding optimization issues. Further, this systematic and taxonomical survey also provides a deep analysis of various emerging energy harvesting technologies in the last twenty years of the era.


Introduction
The current era is the witness of many emerging technological advancements including the Internet of Things (IoT), cloud computing along with wireless sensor networks, and further integration of these emerging technologies for collecting and deep analysis of monitored information [1][2][3].
Also, this analyzed factual information is needed for enhancing the efficiency of the particular industrial system by ensuring optimal resource consumption. Moreover, for monitoring various events pertaining to home/office, human activities, health, defense, agriculture, and industries, etc., wireless sensor networks are utilized extensively [4,5]. In extreme circumstances, sensor nodes are remotely deployed in harsh environmental conditions with the limited capacity of the battery for particular monitoring applications in which long operational periods are needed even years or decades and battery power depletes regularly with the course of time, and in these undesirable situations, it is impossible to provide battery replacement facility or recharging activity; therefore, energy efficiency is considered a major challenging issue for sensor nodes [6,7].
To amicably handle the energy scarcity issue of sensor nodes, there is a need for an alternative energy source to compensate for the energy requirement in case of battery drain out [8,9]. One of the best resolutions to this issue is the use of an energy harvesting technique in which energy from natural sources available in the environment is transformed into power to resuscitate the batteries [10,11]. The new advancements in energy harvesting techniques for WSNs have become a prominent research field. But, in the case of energy harvesting, the energy output may be insufficient due to streaky and spatial variations, and some mechanisms are needed for effective utilization of the harvested energy; further, these mechanisms are collectively termed as energy optimization strategies [12,13].
The energy harvesting wireless sensor networks (EH-WSNs) are having a large number of challenging issues based on their inherent characteristics, and these issues have a direct impact on the performance [14]. In this article, we will discuss the various challenging issues and corresponding suggested solutions in the subsequent sections, although our main focus is on energy efficiency. It is a noteworthy fact that although a large number of research have been completed and others are still going on with a proposal for enhancing energy efficiency, energy optimization has become an evergreen research challenge for EH-WSNs [15]. For EH-WSNs, there is a need for some planned mechanism known as energy optimization strategies for optimum energy consumption in a sensor node by managing energy efficiently. There exist a large number of literature focusing on energy management schemes in the last two decades; the reason is obvious; for optimal performance of WSN, there is a need for energy optimization. The ultimate aim of energy optimization strategies is to manage the energy in the network for increasing the lifetime of tiny sensor nodes, and the network remains operational perennially. Special attention is needed for managing the energy optimization issue for sensor nodes [16].
This systematic and taxonomical survey focused on discussing the energy optimization strategies for EH-WSNs. There are certain specific reasons behind this. The energy optimization strategies in EH-WSN help in achieving the desired balance between processing and transmission of data. By effectively utilizing the optimization strategies, the optimum performance can be achieved in EH-WSNs with the existence of prevailing various limitations. The term optimization in mathematical form can be defined as discovering the minimum or maximum for a particular defined function based on multiple constraints. For an optimization problem, a feasible solution is developed with the help of multiple values that satisfies all the prevailing constraints. The ultimate aim of a well-defined optimization technique is to care-fully observe the feasible solutions and then proposing the optimum solution. Considering the EH-WSNs scenario with multisource energy harvesters, the design of power management circuits can be optimized for enhancing the overall efficiency. One of the most appropriate schemes for handling energy constraints issues of WSNs is the energy harvesting technique, and the proper optimization of EH technologies and devices is needed; further, the optimal energy optimization strategies should be coupled together for effectively utilizing the harvested energy.
1.1. A Distinctive Approach and Motivation. In this article, we have done an extensive study for classifying the energy optimization strategies for EH-WSNs; the uniqueness in this research that distinguishes this work from others is the indepth classification of energy optimization strategies for EH-WSNs by covering all possible aspects. Also, in this article, different energy optimization strategies have been discussed consisting of various algorithms along with further details of improvements in the coming future.
Although, there exist a large number of literature on energy management issue that mostly focus on a particular area and discuss energy management by considering only particular perspective and they do not concentrate on other areas affecting energy utilization in EH-WSNs, and we have tried our best to overcome this drawback of existing pieces of literature and try to cover as many as possible all perspectives in this article. We hope that the energy optimization strategies discussed in this article with a wide view definitely help the researchers working on this area to understand the key pillars of the energy optimization framework. Further, this article also provides space to explore new possibilities for energy efficiency in the field of EH-WSNs.

Salient Contributions towards
Findings. In this systematic review article, we have conducted a detailed study of energy optimization issues in EH-WSNs. The detailed study helps in finding the facts (research gaps) in the existing review papers and guides the authors to write a systematic review article covering all aspects related to energy optimization for EH-WSNs.
The major contribution of this systematic review article can be pointed out as follows: (1) This systematic review article explains the energy optimization strategies for EH-WSNs by considering eleven factors, namely, radio optimization schemes, optimizing the energy harvesting process, data reduction schemes, schemes based on cross-layer optimization, sleep/wake-up policies, schemes based on load balancing, schemes based on optimization of power requirement, optimization of communication mechanism, schemes based on optimization of battery operations, mobility-based schemes, and finally energy balancing schemes. The ultimate aim is to discuss the energy optimization schemes for EH-WSNs with a diversified view covering all possible areas having the sharp vision that affected the energy consumption in the network 2 Wireless Communications and Mobile Computing (2) Furthermore, this systematic review article provides the pointwise precise overview of various key pieces of literature in the last twenty years covering especially duty cycle schemes, MAC protocols, opportunistic routing schemes considering geographic characteristics, cluster-based routing, and also energy balancing schemes with key attributes (3) Next, this systematic review article provides complete coverage of challenging issues for formulating the optimization problems for EH-WSNs and also describes the solutions to optimization problems in the form of popular algorithms (4) Further, this systematic review article provides the paradigm shift for energy harvesting technologies considering the twenty years of an era starting from the year 2000 to 2020 and also describing the nextgeneration technologies and future of energy harvesting techniques We have structured this systematic review article as per the following: Section 1 comprises the brief introduction of WSNs, the need and role of energy harvesting techniques, and WSNs with energy harvesting techniques (EH-WSNs), and in the last, this section describes the need of energy optimization strategies for EH-WSNs. Section 2 provides a deep insight into the adopted research methodology. In Section 3, this article provides details about the energy harvesting techniques with corresponding sources of energy. In Section 4, this article provides the classification of energy optimization strategies with a broad view by covering multiple factors; also, the precise tabular representations of duty cycle schemes with key attributes, MAC protocols, opportunistic routing schemes considering geographic characteristics, key features of clustering approaches, and also energy balancing schemes are illustrated. The optimization problems with the respective solutions for EH-WSNs are provided in Section 5. Also, a paradigm shift for energy harvesting technologies along with the future of energy harvesting techniques for WSNs and further research directions is provided in Section 6. In Section 7, this review article provides a conclusion.

A Deep Insight into the Adopted Research Methodology
This systematic review article focuses on the energy optimization strategies for EH-WSNs with a broad view by covering all possible factors that affected the energy efficiency in EH-WSNs.
There exist a large number of literature with a focus on performance enhancement for EH-WSNs; further, it should be noted that researchers have already put lots of efforts towards proposing various strategies for EH-WSNs for enhancing the performance by covering the energy scarcity issue. We would like to highlight a few pieces of literature as [17][18][19][20]. We have tried our best level to select the appropriate literature for conducting the detailed study for collecting facts and thereby designing this systematic review article.
We have selected the era starting from 2000 to 2020 for selecting the appropriate good literature with good citations. Articles published in different journals/conferences have been studied for collecting the facts about energy optimization. Figure 1 illustrates the number of literature considered yearwise.
We have adopted a structured research methodology consisting of six major steps as depicted in Figure 2. In this systematic review article, we have structured the research into six major steps: In step 1, we select the potential research questions, and further, we understand the need of exploring the research questions; next, we define the research objectives corresponding to potential research questions. Further, in step 2, we prepare the new database consisting of high-quality research papers from three already existing databases, namely, Web of Science, SCI/SCIE, and Scopus databases, and next, we conduct an extensive research study which leads to further retrieval of data for formulating systematic research design. Furthermore, in step 3, we prepare the summarized solutions corresponding to research questions with the help of an analyzed literature review. Next, in step 4, we perform the quality assessment activities for verifying the facts and figures which we have collected from the newly constructed database. This step is very important because this systematic review article will provide significant information to researchers working on the energy optimization issues of EH-WSNs. Moreover, in step 5, we prepare the summary of outcomes after analyzing the literature in the newly constructed database. Besides, in step 6, we mark the corresponding references.
The newly constructed database consists of various literature, and we have considered three major parameters while considering the particular literature for this systematic review and three parameters are the articles that should focus on our primary objective, i.e., energy efficiency, the journal in which literature has been published that should be peerreviewed, and next, the selected literature that should have appropriate citations.
Initially, we have selected 610 pieces of literature. After, the analysis we have further selected 305 pieces of literature out of the database of 610 pieces of literature. Again, we perform further analysis and select 178 pieces of literature out of 305 pieces of literature previously selected.

Energy Harvesting Techniques with
Corresponding Sources of Energy The major issue that has a severe impact on the performance of WSNs is the evergreen energy scarcity issue that needs to be handled with a focused approach. Traditionally, sensor nodes are operating with limited capacity batteries and batteries are having their limitations such as leakages of current, breakdown issues, and limited energy density issues. Also, the sensor nodes energy profile is regularly depleted with the time; thereby, sensor node is not able to provide the assigned duties especially in longer duration, but there exist many applications which require a much higher lifespan of sensor nodes ranging from months to several years, so in 3 Wireless Communications and Mobile Computing these emergencies, there exist only two usual solutions in terms of replacing the batteries or harvesting the energy to fill the gap that arises due to depleted energy. Also, battery replacement is sometimes not possible especially in cases of difficult remote deployment. So, the most general solution for enhancing the life of sensor nodes is by providing a regular power supply and further this is possible by utilizing the energy harvesting mechanism and this scheme enables the sensor nodes to work continuously without interruption, since, in most cases, sensor nodes need a continuous power supply. Further, Figure 3 describes the different types of energy harvesting mechanisms with different energy types available for harvesting and Table 1 illustrates the power density of various energy harvesting schemes.
Next, we are trying to illustrate the brief description with relevant references as per the following.
Mechanical energy harvesting [21,22] is defined as the method in which energy conversion from mechanical to electrical occurs. The harvester consists of the spring-mounted mass component installed inside, and the mechanical energy from displacements and oscillations is converted into electrical energy. This entire process utilized mechanical stress, vibrations, rotational movements of waste, fluid, force, and high-pressure motors. There exist three forms of mechanical energy harvesting, namely, electromagnetic, electrostatic, and piezoelectric. In the piezoelectric energy harvesting method [23,24], the two factors play a major role such as the use of piezoelectric material and the piezoelectric effect.
Here, the piezoelectric material is strained, and further, energy from vibrations, pressure, and force is converted to electrical energy. The electrostatic energy harvesting mechanism [25] utilized the principle in which energy conversion occurs from mechanical to electrical with the capacitance change. Here, the key factor is changing the capacitance of a varying capacitor which is depending on the vibrations. Next, the popular law of electromagnetic induction which is known as Faraday's law is used in the electromagnetic energy harvesting mechanism [26,27]. Here, the electromagnetic harvester utilizes the inductive spring-mass system which is the backbone of this type of harvesting system. The stationary magnet creates the magnetic field, and further, movement of magnetic material via the magnetic field induces the voltage. Now, energy conversion takes place from mechanical energy to electrical energy.
Photovoltaic energy harvesting is defined as the method in which conversion of photons to electrical energy occurs, and here, photons are coming from sources such as artificial or solar light. In this harvesting mechanism, photovoltaic cells are exploited. Here, at the P-N junction, an electrical field is formed [28]. The solar energy harvesting mechanism is illustrated in Figure 4.
Thermal energy harvesting [29] can be implemented in two ways as pyroelectric and thermoelectric energy harvesting. Further, thermoelectric energy harvesting consists of power generators that are used for electrical energy generation based on the Seebeck effect and known as thermoelectric  Wireless Communications and Mobile Computing power generators. The thermopile is the main element of a thermoelectric power generator. Further, two dissimilar conductors' arrays are used in the formation of the thermopile [30]. Pyroelectric energy harvesting is defined as the method in which voltage is generated by cooling or heating the pyroelectric materials. The basic core element of pyroelectric energy harvesting is the pyroelectric material's crystal structure, and due to change in temperature, atom location in crystal structure changes and which further results in producing the voltage. Pyroelectric materials require timevarying changes in the temperature as compared with ther-moelectric energy harvesting in which temperature gradient is needed [31]. Wireless energy harvesting exists in two major forms such as resonant and RF energy harvesting. Further, RF energy harvesting consists of a rectifying antenna or rectenna which is used for converting electromagnetic waves into electrical energy. In the RF energy harvesting mechanism, there are two sources from which energy can be harvested either through RF power or through electromagnetic signals with a specific wavelength. Further, sources, namely, microwaves, cell phones, broadcasting of television and radio, and Wi-Fi,  Figure 2: Structured research methodology with six major steps. 5 Wireless Communications and Mobile Computing are considered as ambient RF power. Resonant energy harvesting or resonant inductive coupling can be defined as the method of transmitting and accumulating electrical energy specifically between the two coils. There exist two coils named primary and secondary coil; further, an inductive transformer is attached to a primary coil. At the same frequency, these two coils are resonant. Next, through the air, power is sent to a specified device that is attached to a secondary coil. The magnetic flux is produced by the primary coil which is time-varying in nature, and further, voltage is induced whenever that magnetic flux crosses the secondary coil [32]. RF energy harvesting is depicted in Figure 5.
Wind energy harvesting is defined as the method in which the conversion of energy from airflow or wind energy to electrical energy occurs. In wind energy harvesting, a specified-sized wind turbine is used. Further, the wind turbine exploited the linear motion due to wind for the generation of electrical power [33]. Wind energy harvesting is explained in Figure 6.
Biochemical energy harvesting is defined as the method in which electrochemical reactions are used for the conver-sion of endogenous substances and oxygen to electrical energy [34].
Acoustic energy harvesting is defined as the method in which a resonator or transducer is used for the conversion of acoustic waves to electrical power [35]. Acoustic energy harvesting is illustrated in Figure 7.
Hybrid energy harvesting is defined as the method of combining any harvesting technologies and further concurrently using this hybrid model on a single platform. The concept of hybrid energy harvesting is illustrated in Figure 8.
In the current era of IoT, nanoenergy-based technology is booming which is known as nanogenerators. Further, nanogenerators are having the capability of harvesting energy from the surroundings and efficiently utilizing that harvested energy for facilitating the tiny sensors and other portable electronic devices. In a more general way, we can define the nanogenerators as the energy harvesting device for generating electrical energy from ambient mechanical energy. The nanogenerators are classified into four major categories such as piezoelectric nanogenerators (PENGs), triboelectric nanogenerators (TENGs), thermoelectric generator (TEGs), and      Wireless Communications and Mobile Computing futuristic energy harvesting devices with enormous capability [36,37].
Furthermore, in the case of a linear and rotatory generator for energy harvesting [38,39], the motion of the wave provides irregular mechanical energy. Further, this form of mechanical energy is needed to convert into regular mechanical motion. Next, there exist two types of motions, namely, linear or rotational. The rotational motion is dedicated for driving a turbine, and then, the rotating electrical generator is driven. But the linear motion is responsible for driving a linear electrical generator. Figure 9 illustrates the role of linear generator in converting ocean wave energy to electrical energy.
In the energy harvesting mechanism, there exists specific hardware which is known as an energy harvester and it is responsible for converting the environmental energy to electrical energy. Further, there is a need for efficient conversion of harvested energy to electrical energy and then appropriately conditioned by the power management circuit to an appropriate form then energy is stored in the energy storage elements, and finally, electrical energy can be utilized either for energizing the batteries or directly providing the supply to the load.

Classification of Energy Optimization Strategies for EH-WSNs
In traditional WSN, the remotely deployed energy constraint sensor nodes (SNs) require optimal energy optimization strategies for efficient utilization of capabilities of WSNs. Generally, traditional sensor nodes (SNs) are powered by limited capacity batteries that are attached to SNs and the use of SNs are particularly useful in an isolated and remote area, but the limited energy capabilities of SNs affected the performance in these remote areas. Further, in harsh challenging environmental conditions, it is very difficult or impossible to replace the batteries or energizing the batteries. Therefore, WSNs, which are powered by conventional batteries, are not able to provide long-lasting performance. The issue of the limited capacity of battery power has shifted the attention of researchers towards finding alternate sources of energy to energize the SNs by harvesting ambient energy. Therefore, the life span of WSN can be enhanced by utilizing the emerging technologies known as energy harvesting schemes, and further, energy optimization strategies can be used to effectively utilize the energy harvesting techniques. Also, these mechanisms achieve balance in the energy for the overall network. The uniqueness in this systematic survey is that we have considered the two views while conducting the survey, first considering the current enhancements in energy optimization strategies while in parallel comparing with the traditional approaches for handling energy scarcity issues. This overall systematic survey broadly consists of two branches, namely, energy harvesting techniques coupled with energy optimization strategies for efficiently utilizing the harvested energy.
The aim of conducting the survey about energy optimization strategies is to handle the challenging issue of energy scarcity, and in this way, we can keep the sensor nodes alive making the network more operational and efficient. It must be noted that in this systematic survey, we have studied the internal design details of the protocol with energy-saving capabilities for understanding the limitations with abilities in dealing with the system with energy harvesting characteristics.
The role of energy optimization schemes for EH-WSNs is eventually to save energy which is considered as the primary issue for the EH-WSNs for achieving the specified objective. Extensive deep studies have been conducted for the classification of energy optimization schemes for EH-WSNs in the following categories by considering eleven factors, namely, radio optimization schemes, optimizing the energy harvesting process, data reduction schemes, schemes based on cross-layer optimization, sleep/wake-up policies, schemes based on load balancing, schemes based on optimization of power requirement, optimization of communication mechanism, schemes based on optimization of battery operations, mobility-based schemes, and finally energy balancing schemes. The ultimate aim is to discuss the energy optimization schemes for EH-WSNs for achieving energy efficiency by decreasing energy consumption and also achieving energy balance among sensor nodes. Figure 10 depicts the deep classification of energy optimization schemes for EH-WSNs.
4.1. Radio Optimization Schemes. Radio optimization schemes are further divided into three categories such as schemes based on power control in transmission, cooperative communication schemes, and also modulation optimization schemes. A brief discussion about all these schemes is provided below.
In WSN, one of the main units in which heavy energy consumption occurs is the radio module. Therefore, some power control mechanism is needed for transmission. In the adaptive transmission power control framework, a separate model representing the relationship between power needed for particular transmission and corresponding quality of the link is built by the separate node for their neighbors. Further, these models help in maintaining link quality over a period of time dynamically with the help of a feedback-based algorithm for power control in transmission [40].
Next, for designing optimal WSN, optimization of energy is a crucial factor that needs to be carefully addressed. Also, in WSN, the energy consumption part in the circuit module as compared with actual energy consumption that occurs in the transmission part may not be negligible, and further, both these parts need to be recorded separately. Therefore, the general optimization techniques for energy may not be sufficient in the case of WSN, which are responsible for reducing the energy in the transmission process. Also, it has been Ocean wave energy Linear generator Rectifier Inverter Load Figure 9: Ocean wave energy converted to electrical energy. 9 Wireless Communications and Mobile Computing proved in the literature that two prominent factors, namely, energy consumption and delay in transmission, can be minimized under the cooperative environment among the sensor nodes in WSN for transmitting and/or receiving information [41,42].
Next, it should be noted that in communication link with point-to-point connection and for applications having shorter distance, one traditional fact that higher transmission duration results in minimum energy consumption may be false if energy consumption in circuit module is significantly considered. Also, the transmission time should be maximum in order to minimize the energy requirement in the transmission process. Further, there is a need of optimizing the transmission time for overall minimizing the total energy consumption in WSN [43,44].

Optimization of Energy Harvesting
Process. The research area towards the energy harvesting techniques is getting much higher attention by the researcher currently, and lots of research is going on exploring the optimal use of energy harvesting technique. The research activities on energy harvesting area can be broadly categorized into two categories such as optimal design and development of energy harvesting framework, and next category deals with efficiently storing the generated charge and this is possible only by proficiently designed capable electronic circuits. The microelectronic devices used in the harvesting process need to be optimized for reducing the overall energy consumption in the system. There exist a large number of literature discussing the opti-mization of energy harvesting mechanism; we are highlighting a few of the literature as follows.
Buchli et al. [45] have presented a unique scheme for dynamically adjusting the performance level of the system by designing an efficient power subsystem, and in this way, uninterrupted energy-neutral operation can be achieved in solar EH-WSNs. Next, Reis et al. [46] have discussed about the need of optimizing the extraction circuits for efficient utilization of the energy harvesting mechanism. Further, Tai and Zuo [47] have conducted a two-variable optimization analysis for deriving the exact optimization conditions for maximum power. Furthermore, Cai and Harne [48] have proposed an optimal optimization framework at the system level, particularly for a nonlinear vibration energy harvester. Besides, Fouad [49] has discussed the important role of the rectifier in energy harvesting applications and proposed an efficient CMOS rectifier.

Data Reduction Schemes.
The data reduction schemes can be further classified as schemes based on aggregation, compression, adaptive sampling, and also network coding. A brief discussion about all these schemes is provided below.
In WSN, one of the ways to save energy is by efficiently handling the transmitted data. The aim of data aggregation schemes is to collect the sensitive data from the sensor nodes in the network and further sent it to the base station with lower data latency. Environment monitoring requires fresh data, and therefore, data latency is playing a crucial role in various monitoring applications [50]. It is a very tedious step   10 Wireless Communications and Mobile Computing or even impossible to access all original data after performing the aggregation function. The success of gaining original data depends on the type of aggregation function. There are two approaches, namely, lossy and lossless. In the lossless approach, there exists a high possibility of gaining the original data without error. For selecting a particular approach, various factors need to be evaluated such as application area, rate of data transmission, and network characteristics [51].
Further, the data samples should be minimum for energy conservation schemes. The impact of reduced data samples also results in the decrement of the total number of communications and thereby maintaining energy efficiency. Further, spatiotemporal correlations can be utilized for reducing the data samples and such type of scheme is known as adaptive sampling. Here, a joint approach covering the features of both temporal and spatial correlations can be used to minimize the total amount of data that is needed to be acquired [52]. Furthermore, in WSN, the number of transmissions can be further reduced by using the characteristics of multiple data packets coding scheme. This mechanism results in a reduced number of transmissions because multiple data packet coding occurs within a single transmission [53].

Schemes Based on Cross-Layer Optimization.
The crosslayer optimization schemes can be further classified as MAC and network layers; application, transport, and network layers; MAC, network, and physical layers; and finally all layers. A brief discussion about all these schemes is provided as follows.
Bouabdallah et al. [54] have presented a joint cross-layer framework considering MAC and network layers. For enhancing the lifetime of the network, energy-efficient protocols are needed. They proposed the scheme for traffic balancing at the network layer, and further, they proposed another scheme for controlling the retry limit of retransmissions at the MAC layer. In the proposed framework, energy efficiency is achieved by controlling the retry limit and thereby contributing towards enhancing the lifetime of the network.
You and Liu [55] have formulated a problem for maximizing the lifespan of the network which is known as a cross-layer problem. The counterpart of the formulated cross-layer problem can be decomposed into several individual subproblems such as problem for the application layer with the focus on improving the lifetime, problem for transport layer with the focus on controlling the source rate, and problem for network layer with the focus on optimizing the routing efficiency. Further, the subgradient methods are used for solving the dual problem.
Lee et al. [56] have presented a cross-layer optimization framework for improving system performance. Further, they suggested that this joint optimization framework efficiently handles the communications especially in the case of delayconstraint applications. The mechanism of the proposed cross-layer optimization framework can be illustrated with consideration of different individual protocol layers such as MAC layer for power allocation, network layer for optimal routing, and physical layer for energy management.
Phan et al. [57] have presented a cross-layer design in the form of an optimization problem with two stages. The first stage of the optimization problem is dedicated for maximizing the total number of admitted sensor nodes. Further, the second stage is aimed at enhancing the network lifetime. Therefore, this cross-layer design is covering all layers. Next, they have also shown that the optimization problem having only one stage with a compact and precise mathematical framework can be derived from the two-stage optimization problem.
4.5. Sleep/Wake-Up Policies. The sleep/wake-up policies can be further classified as schemes based on topology control and also schemes based on duty cycling schemes. A brief discussion about topology control schemes is provided as follows.
Topology control framework is aimed at maintaining efficient connectivity or coverage in WSN. In this framework, each sensor node dynamically adjusts the power for optimal transmission. Also, each sensor node selects the appropriate neighboring sensor nodes set for direct communication. All these steps in the topology control framework contribute towards the conservation of energy at each sensor node and consequently enhance the network lifetime.
The optimal deployment of WSN requires careful handling of two crucial factors, namely, network coverage and connectivity. Further, area monitoring applications need proper coverage of networks with efficient connectivity; therefore, the success of area monitoring applications primarily depends on these two prominent factors. Also, there may exist a probability of coverage area redundancy issue and this redundancy should be minimum for optimal resource utilization by the resource constraints sensor nodes in the WSN [58]. Next, the cooperative topology control framework has been presented in [59] and this framework is a distributed topology control scheme with the objective of maximizing the overall lifetime of the network.
Next, a detailed review for duty cycling schemes is provided in tabular form as described in Table 2.

Schemes Based on Load Balancing.
A brief discussion about various schemes based on load balancing is provided below.
Liu et al. [92] present various strategies for enhancing the performance of EH-WSNs with a focus on three areas, namely, efficient scheduling, energy-efficient relaying, and finally optimizing the medium access control layer protocols. Next, Cai et al. [93] have presented a new routing algorithm for EH-WSNs in which the flow of load is maximized under the life span of the network; also, in this framework, energy consumption is balanced for prolonging the lifetime of the network. Here, residual energy is considered as a prominent factor for updating the transmission capacity between any two particular sensor nodes. Further. Wu et al. [94] have proposed an efficient energy-balanced routing framework that is based on an autonomous load regulating scheme for rechargeable WSN. In this framework, significant steps are taken to control the relay radius, and consequently, nodes become capable of adjusting their load. Furthermore, Chai and Zeng [95] have presented a routing scheme for EH-WSNs with efficient load balancing characteristics. This (ii) In the proposed framework, the sleep time is adjusted dynamically.
(i) Future work may include extending the works towards the multihop data delivery scheme.
(ii) Further, future research directions include efforts towards the in-depth analysis of the energy harvesting process for transmission strategies.
Li et al. [64] Packet delivery ratio (i) The scheduling problem is defined in terms of a partially observable Markov decision process.
(ii) This article also presents that the formulated scheduling problem can be portioned as Markov decision process (MDP).
(i) The simulations are based on a single group consisting of the source node, relay, and destination node. Future work may include an extension towards considering multiple groups.
(ii) Future work may also include some more performance evaluation metrics.  (i) The future work may include an extension towards improving prediction quality.
(ii) The future work will focus on the global information consisting of cloud cover or snow warnings that can be used for enhancing the accuracy in long-term prediction.

2013
Akgün and Aykın [70] Life span of network (i) This article proposes a framework for enhancing the life span of the network.
(ii) In this framework, the scheduling scheme based on TDMA has been modified, and in this modified scheduling scheme, members will request a time slot based on their energy prediction.
(iii) Next, the cluster heads will be responsible for assigning particular slots to members.
(i) The future work may include the extension towards simulating the system in other simulation platforms such as NS3 and then assessing the system performance.
(ii) The future research direction also considers other energy sources such as vibration or pressure and comparing our framework with MAC protocols other than LEACH.
Sommer et al. [71] State of charge for battery (i) This article presents a framework for estimating the state of charge for a battery of sensor node by considering the two points (1) the current flow in the battery and (2)   Chetto [73] Energy and time constraints (i) This article proposes a semionline scheduling algorithm that is based on the earliest deadline first methodology.
(ii) Further, this framework depends on two specific factors such as energy demand and slack energy.
(i) The future work may include an extension towards incorporating fixed priority environments in the proposed framework.
(ii) Next, extension towards supporting dynamic voltage and frequency selection technology.
(iii) Furthermore, the extension towards considering the smallest harvester etc.
Castagnetti [74] Packet reception ratio & energy efficiency (i) This article proposes the power management framework for EH-WSNs.
(ii) Next, the proposed framework utilized two factors, namely, optimization of duty cycle and power control in transmission.
(i) The future work may include extension towards considering various adaptation methodologies, namely, adaptive modulation and coding for improving the performance and energy efficiency.
Li et al. [75] Total number of executed tasks (i) This article proposes the task scheduling scheme utilizing prediction data, and further, this framework is based on weather forecasts in EH-WSNs.
(i) The future work may include extension towards exploring the advantages of weather forecasting data for longer scheduling.

2015
Liu et al. [76] Energy utilization & deadline miss rate (DMR) (i) In this article, an intratask scheduling scheme is presented for the storage-less and converter-less channels.
(ii) Next, this framework utilizes neural network training.
(i) The future work may include extension towards utilizing the learning automata in place of the neural network and then assessing the system performance.
Zhang et al. [77] Total number of ready tasks & deadline miss rate (DMR) (i) This article proposes the scheduling algorithm utilizing the neural network for determining the appropriate scheduling pattern, the optimal size of the capacitor, and finally queue of tasks for improving the deadline miss rate.
(i) The future work may include extension towards utilizing other alternatives in place of ANN such as learning automata for further enhancing the deadline miss rate.

2016
Ali [78] Lifetime of battery & consumption of power (i) This article proposes the eventdriven duty cycling framework for reducing the power consumption of the roadside unit which are having self-powered capabilities.
(i) The future work may include extension towards modifying the proposed framework by including few more factors, such as the weather conditions, for enhancing the performance of the distributed power management scheme.
Gomez et al. [79] Efficiency of system (i) This article proposes an efficient energy management unit to maintain power supply especially in those circumstances in which the harvested power is not sufficient for smooth system operation.
(i) In the proposed work, optimalsized buffer is used and the future work may include the variable-sized buffer and then measuring the system efficiency.
Zhang et al. [  wireless, and vibration and also some hybrid approaches.
Oueis et al. [81] Harvested energy, battery residual energy, and battery level variation (i) This article discusses the effect of photovoltaic energy harvesting on the duty cycle of sensor nodes in both scenarios outdoor and indoor.
(i) The future work may include extension towards evaluating and comparing several other performance metrics such as end-to-end delay and packet delivery ratio.
Housseyni et al. [82] Success ratio for deadline (i) This article proposes a new scheduling framework for a sporadic task model. (ii) Next, an energy-efficient offline task assignment heuristic is generated by the proposed framework.
(i) The future work may include an extension towards exploring the proposed framework in which arrival times are characterized with probabilistic distributions.
Maeda et al. [83] End-to-end delay, total number of packets sent to base station (sink) per cycle, and the fairness index for the node position (i) This article proposes a task scheduling scheme in which data is periodically collected from the sensor nodes.
(i) In this article, line topology is used for simulating the sensor network and future work may include extension towards adopting other topology and then measuring the performance of the system.

2017
Sanchez et al. [84] Final remaining energy of sensor node (i) The energy management strategy is based on a hybrid dynamical system approach.
(ii) The combination of continuous physical processes results in the hybrid nature; on the other hand, change in the functioning modes results in the discrete concept.
(i) In the current work, the sensor nodes' energy is modeled using a specific representation named as a hybrid dynamical systems representation; future work may include modeling the energy with other representations and then measuring the performance.

2018
Huang et al. [85] Optimal expected rewards (i) This article proposes a scheduling policy for EH-WSNs based on the threshold value.
(ii) In this framework, limited memory is needed for storing optimal threshold values at the sensor node for carrying out energy management activity.
(i) The future work may include an extension towards utilizing Markov channels for energy harvesting systems.
(ii) In this scenario, the system state will be represented by data packets, energy, and the state of the wireless channel.
Bengheni et al. [86] Mean latency, throughput, and packet delivery ratio (PDR) (i) This article proposes the enhanced energy management scheme in EH-WSNs.
(ii) Further, this framework utilizes receiver-initiated communication.
(iii) Next, a policy based on the energy threshold is used for controlling the active/sleep periods.
(i) Future work may include extension towards considering other performance metrics such as average energy consumption and network lifetime.
Anagnostou et al. [87] Rate of execution and energy efficiency (i) This article proposes a hardware scheduling framework that is poweraware.
(ii) Next, this framework also observes harvesting power.
(iii) Further, the available energy along with the software module is used to dynamically activating the other system modules.
(i) Future work may include extension towards using the dynamic schedulers and also supporting the interruptible tasks.

Schemes Based on Optimization of Power for the
Hardware Devices. Schemes based on optimization of power can be further classified into two categories such as processor power management and device management. A brief discussion about these schemes is provided below.
The rapid development in the IC fabrication technology and current innovations in semiconductor technology results in a tremendous transformation in the field of highperformance computing such as starting from the development of single-core architecture to multicore architecture with homogeneous characteristics and then multicore architecture with heterogeneous and dynamic reconfigurable characteristics. This rapid development further imposes a challenge in terms of increment in power density as well as heat dissipation and consequently affected the system reliability and availability. Currently, research is going on achieving high performance with low power consumption. Nagalakshmi and Gomathi [96] have presented various effective techniques for overall reducing the power dissipation in multicore processing architecture. Further, there is an urgent need to handle power management issues for designing effi-cient microprocessors in current scenarios of highperformance computing. The ultimate goal is to maximize the performance of the processor with low power consumption. The role of power management techniques is to maintain the balance between higher performance and power consumption with aggressive thermal effects. Next, Attia et al. [97] have explored the various schemes for managing the power in multicore processing architecture.
The designing of an electronic system should be harvesting-aware from two perspectives such as the hardware and software perspectives for achieving optimal performance in the energy harvesting environment. Also, the efficiency of the harvesting system can be enhanced by utilizing a proper power management framework with harvesting-aware characteristics. Further, an efficient power management framework will enable the harvesting system to operate uninterruptedly and achieve the expected objectives. One of the prominent operating modes is the energy-neutral mode that can provide assurance about longer system operation [98]. Further, an efficient and proper functional power management framework is the need of the hour in the current high-performance demanding environments for an energy harvesting device. The harvested energy is stored in the energy storage element until enough energy is currently corresponding hardware) platforms.
(ii) Next, geographic latitude and meteorological conditions of the deployment should be used for refining the energy harvesting model.
Cui [89] Mean relative error, and root mean square error (i) This article proposes the solar energy prediction method which is based on long short-term memory recurrent neural network.
(ii) Also, a predictive task scheduling framework is proposed based on the prediction of energy available for improving the overall performance of the WSN.
(i) Future work may include extension towards using other alternatives in prediction such as gated recurrent unit (GRU) and bidirectional LSTM (bi-LSTM) and then measuring the performance.

Schemes Based on Optimization of Communication
Mechanism. Schemes based on optimization of communication mechanism can be further classified into three major categories, namely, MAC protocols for EH-WSNs, transmission schemes classification, and also routing protocols for EH-WSNs. A brief discussion about all these schemes is provided below.
4.8.1. MAC Protocols for EH-WSNs. These are further categorized into three categories such as sender-initiated asynchronous protocols, receiver-initiated asynchronous protocols, and also sink-initiated asynchronous protocols. Tables 3-5 are used to briefly describe each respective category clearly as follows.

Transmission Schemes Classification.
Transmission schemes are divided into three categories, namely, fixed transmission schemes, variable transmission schemes, and also probability model-based transmission schemes. These schemes are briefly described below.
(1) Fixed Transmission Schemes. Reddy and Murthy [112] have effectively addressed the power management issue in EH-WSNs. They have proposed the framework that is based on the approach of energy-neutral for managing the power efficiency issues in communication and therefore enhancing the utility of sensor nodes that are utilizing the harvested energy. The designing of this framework clearly incorporates three factors such as fixed power consumption in the circuits, inefficiencies of battery, and finally storage capacity.
(2) Variable Transmission Schemes. The energy harvesting mechanism is considered as the prominent solution for sustained WSNs. In EH-WSNs, two important factors need to be considered while designing the optimal transmission policies such as the mechanism of energy refreshing and the storage constraints for rechargeable batteries. Tutuncuoglu and Yener [113] have tried to address these issues with optimum solutions. The proposed framework specifically tried to determine the transmission policy that enhances the data transmission rate in a bounded timeline. Finally, the proposed framework determined the optimum transmission policies with constraints such as the mechanism of energy refreshing and the storage constraints for rechargeable batteries. The mechanism of energy refreshing considers the model having discrete packets for energy arrivals. Next, the energy harvesting technique usually suffers from two vibrant issues such as instability and random behavior in harvested energy and these two issues need to be tackled efficiently since these issues cause temporal death of sensor nodes and thereby making a negative impact on the quality of service and overall performance starts degrading. Tang and Tan [114] have proposed the framework which considers the case of temporal death and analyzes the behavior of energy harvesting devices for data transmission characteristics, and further, the proposed framework determines the optimal transmission policy.
(3) Probability Model-Based Transmission Schemes. Berbakov et al. [115] have presented the approach for determining the optimal joint transmission scheme in a particular scenario consisting of two sensor nodes in which one is battery operated and another one is operated purely based on the harvested energy. Also, in this framework, the common message is transmitted to the base station cooperatively. The ultimate aim is to find the optimal joint transmission scheme for maximizing the throughput with the bounded deadline. Next, Michelusi et al. [116] have proposed an approach for handling the issue of uncertainty in estimating the state of charge for rechargeable batteries since it is almost impractical or very costly to accurately estimate the state of charge for the rechargeable battery. Also, the proposed approach noted the impact of deficient information about the state of charge on WSNs.

Routing Protocols for EH-WSNs.
The energy scarcity issue of sensor nodes is an evergreen research area for both traditional WSNs and also for EH-WSNs. In the case of traditional WSNs, sensor nodes operated with limited capacity battery usually start struggling for providing optimum performance after specified time duration, and as a result, the performance of the network starts degrading, and there is a need for efficient utilization and management of harvested energy in case of EH-WSNs. Therefore, the energy issue is considered as a primary factor in designing routing protocols for both traditional WSNs and for EH-WSNs. Further, routing protocols for EH-WSNs are divided into four major categories, namely, routing protocols based on the cost of routes, based on passive characteristics, based on geographic area, and also based on clustering. These are briefly described as follows.
(1) Routing Protocols Based on Cost of Routes. Pais et al. [117] have presented a newly designed function reflecting costbenefit for EH-WSNs. Their research work consists of the main innovation in terms of a new routing cost metric. Further, this routing cost metric is utilized for prolonging the lifetime of the sensor network. Next, Martinez et al. [118] have proposed an efficient framework for selecting the routes. Wireless Communications and Mobile Computing Special emphasis has been given to the amount of the wastage of network energy which is generally not considered in the previous literature while selecting the appropriate routes.
The main source of this network energy wastage is due to overcharging of limited capacity batteries in the network, and this is the main innovation in this framework. Therefore, this framework achieves higher levels of residual energy by reducing energy consumption.
(2) Routing Protocols Based on Passive Characteristics. The routing protocols based on passive characteristics utilize the past information about the network and do not actively consider the current state of the network for constructing the routing tables. Further, Kollias and Nikolaidis [119] have implemented a seasonally aware routing protocol. In this routing protocol, two factors are considered for the construction of routing tables such as solar energy harvesting rates and information about the several routes created in the past years. These passively attributed routing protocols are having an advantage in the specific scenario in which they reduce the wastage of energy in routes establishment overhead in the network, but on the other side, if a particular sensor node dies, then these protocols unable to recover it due to their passive nature.
(3) Routing Protocols Based on Geographic Area. Jumira et al. [120] have proposed an efficient geographic routing scheme for EH-WSNs. The salient features of this energy-efficient beaconless geographic routing consist of loop-free nature with minimum communication overhead and stateless, and also, this routing does not require any prior knowledge about the neighborhood for communication. Further, the mechanism of this routing framework is different from other geographic routing and it started by first sending the data (ii) The proposed MAC protocol is significantly reducing the collision probability, outage probability, and also application layer loss rate; on the other hand, energy efficiency has been increased. (iii) The proposed framework considered fairness in the access mechanism during the design step.
(i) Future research directions may consider the work towards formulation of complex optimization problem concerning to relationship among multiple parameters which are part of the proposed framework. (ii) Next, multiple sleep probability values may be considered for applying to various devices for guaranteeing the packet delay time. (iii) Further, the energy harvesting 802.11ah devices may be explored for analyzing the design space. 18 Wireless Communications and Mobile Computing  the performance under different harvesting models.  20 Wireless Communications and Mobile Computing packet in place of sending the usual control messages which is a general trend in geographic routing. Next, only those neighbors are selected for communication that initially received the data packet successfully and, in this way, appropriate neighbors are selected for efficient communication.
Next, Hieu and Kim [121] have proposed a unique geographic routing scheme for EH-WSNs. In this routing framework, four parameters are considered for selecting the appropriate routes, namely, quality of link, residual energy, location information, and also energy harvesting rate. The ultimate aim of this geographic routing is to select reliable routes and thereby enhancing the lifetime of the network. Further, the detailed review of opportunistic routing utilizing the concept of geographic characteristics is provided in Table 6.
(4) Routing Protocols Based on Clustering. The clustering mechanism significantly contributes towards energy efficiency in EH-WSNs.
A detailed review of clustering-based routing is provided in Table 7.

Schemes Based on Optimization of Battery Operation.
In this section, the main focus is on the issue of efficiently managing the operations of sensor nodes. The brief discussion is provided as follows.
In EH-WSNs, the currently available energy profile is considered for the transmission of data packets; these data packets are generated by the sensor nodes after periodically sensing the field and further stored in the queue. Sharma et al. [144] have presented the optimal energy management schemes for sensor nodes having energy harvesting capabilities. Further, these schemes are attributed as optimal throughput and mean delay. The proposed schemes are aimed at achieving the energy-neutral operation condition in the network. Further, Sinha and Chandrakasan [145] have proposed a scheme for enhancing the energy efficiency of nodes which is actually a power management scheme based on the microoperating system. Next, Chetto and Ghor [146] have analyzed the energy harvesting system having dynamic power management capabilities for selecting the appropriate scheduler on a uniprocessor platform with applicative conditions. 4.10. Mobility-Based Schemes. The mobility-based schemes are further classified into two categories, namely, mobile relay-based and also mobile sink-based schemes. A brief discussion about these schemes is provided below.
The mobility-enabled WSNs are suffering from higher latency factors in specific activities towards the collection of data. The mobile base stations are generally moving at a slower speed for collecting the data which further results in increasing the latency. This issue severely degrades the performance of mobility-enabled WSNs. Researchers explore the issue and find that energy efficiency can be achieved by resolving this issue. Xing et al. [147] have proposed a scheme and tried to address this issue with efficiency. The proposed scheme makes a significant contribution towards balancing the two factors such as energy efficiency and latency factor in the mobility-enabled WSNs. In the proposed scheme, few sensor nodes buffer and aggregate the data and play the role of rendezvous points for the base station for the collection of data.
Next, another challenging issue arises due to the fact that data should be transmitted to the base station only during the lifetime of the particular application. This requirement is hard to achieve in the environment of bounded supplies of power. Moukaddem et al. [148] have proposed a scheme and tried to resolve this issue by introducing the concept of mobile relays and thereby reducing the energy consumption in data-rich WSNs.
In WSNs, the energy holes issue can be resolved by using the concept of sink mobility. The sink mobility scheme efficiently balances the energy and consequently handles the energy hole issue. In the sink mobility scheme, sensor nodes consume less energy since data collection activities occurred in a single hop by the mobile sink which is generally moving throughout the network. In the sink mobility environment, one major issue arises in data forwarding to mobile sink only during time-bound emergency circumstances, and in these situations, sensed data start losing its relevance in the bounded course of time; this sensitive issue needs to be resolved carefully. Ghosh et al. [149] have proposed an efficient scheme in which the Moore curve trajectory motion is used by the mobile sink for the collection of data. The other salient features including any forwarding and also strict sleep/wake-up policy have been followed for sensor nodes.
Next, for specific applications, the unpredictable movement behavior of the mobile sink causes performance degradation since the source sensors tried to locate the continuous moving mobile sink before actual reporting of data. The ultimate aim is to use the minimum number of hops for sending the data to a continuous moving mobile sink and the reason is obvious with the minimum number of hops leads to energy efficiency. Cheng et al. [150] have proposed the framework that addresses both the concerns such as locating the mobile sink efficiently and also reducing hops count.
4.11. Energy Balancing Schemes. The tiny sensor nodes collect the information for the specific event, and further, this information is transmitted to the sink; this is considered as the basic mechanism of WSNs. But, the transmission range issue creates hurdles in monitoring the large area in WSNs and now the need for relay nodes arises further with the help of relay nodes information reached to the sink node. Again, this step creates an imbalance in traffic share which results in an imbalance in energy consumption. The imbalance of energy also results in variation in the lifetime of sensor nodes. The reason is clear as some relay nodes are highly occupied with a portion of traffic and therefore consume a high amount of energy and soon they become dead. Dead nodes cause network portioning and thereby collected information unable to reach the destination, and the basic objective cannot be attained. Now all these circumstances force us to think about the need for an energy balance framework in WSNs. The energy balancing framework efficiently manages the traffic load in WSNs and therefore balances the lifetime of sensor (i) As future work, the proposed framework can be used for a 2D topology.
(ii) Next, future work includes the use of network coding for enhancement of performance.

2012
Eu and Tan [123] AOR Throughput and fairness (i) An adaptive opportunistic routing framework is proposed for multihop EH-WSNs.
(ii) The proposed framework provides a higher throughput.
(iii) The proposed framework utilized the regioning scheme and also capable enough to change as per conditions of network and energy profiles.
(i) Future research directions may consider the work towards addressing factors that are not addressed in this framework.
(ii) The cost of multihop transmission is affected by the duty cycle of relay nodes and it results in the selection of inappropriate routes.

2013
Beheshtiha et al. [124] OR-AhaD Hop count, data delivery ratio, goodput, and efficiency (i) An opportunistic routing framework is presented with a duty cycle approach having awareness of adaptive harvesting mechanism.
(ii) The proposed framework utilized the geographical zoning approach for prioritizing the candidates set.
(iii) The model is presented for managing the energy, and in this model, the duty cycle is adaptively adjusted based on the estimation of harvesting rate in the future.
(i) Future research directions may consider the work towards consideration of the effect of prediction accuracy in the performance of the routing framework.
(ii) Next, future research directions may consider the work for the requirement of the unified evaluation framework for comparing the existing solutions.
(iii) Further, future research directions may consider the work towards test-bedding the existing routing frameworks for exploring the implementation issues.

2017
Shafieirad et al. [125] Max-SNR Data delivery ratio and transmitted messages per node (i) A novel opportunistic routing framework is presented for multihop large-scale EH-WSNs with energy awareness.
(ii) The proposed scheme considered the three factors while selecting the next forwarder such as energy profile, distance, and also the amount of data.
(i) Future research directions may consider the work towards addressing factors that are not addressed in this framework.
(ii) The proposed scheme may result in packet loss as well as poor delay in the case of heterogeneous and dynamic duty cycle and consequently selecting the sleeping relay nodes for packet transmission.
(iii) The duty cycle cannot be adjusted accurately and this results in the limited performance of EH-WSNs.

2018
Zhang et al. [126] OPEH Delivery ratio and delay (i) A novel opportunistic packet forwarding framework is presented with consideration of heterogeneous and dynamic duty cycle.
(ii) The echo state network is utilized for adjusting the duty cycle.
(i) Future research directions may consider the work towards enhancing the efficiency of the echo state network since it is utilized in the current work for adjusting the duty cycle.

2019
Cheng et al. [127] ORDTP Packet redundancy ratio, delay, and delivery ratio (i) An opportunistic routing framework is presented with (i) Further, future research directions may consider the work 22 Wireless Communications and Mobile Computing nodes. The aim is to improve the lifetime of WSNs and to try to maintain the ideal environment in which all sensor nodes share the same lifetime with no variation. But in multihop WSNs, this condition is not feasible to maintain since the traffic density is increased near the sink. Therefore, in multihop WSNs, the objective is to balance the energy consumption.
A detailed review of energy balancing schemes is provided in Table 8.

Challenges in EH-WSNs for Formulating
Optimization Problems along with Solutions EH-WSNs are having various challenging issues which are further formulated as optimization problems. These multiobjective optimization problems can be solved by utilizing various efficient optimization algorithms that are currently available. A detailed classification of various existing challenges for green WSNs is illustrated in Figure 11. Next, with the help of another (Figure 12), we are trying to illustrate some key challenges in EH-WSNs for formulating optimization problems along with algorithms for providing the solution to optimization problems in EH-WSNs. Therefore, Figure 12 is having two blocks; further, in the first block, some key challenges are represented, and in the second block, algorithms are depicted for providing the solution to multiobjective optimization problems. In the following sec-tion, we are trying to address challenges as well as solutions in detail.
In [166], two important metrics such as aggregate utility and network lifetime are considered for performance evaluation. Further, to achieve enhanced network lifetime along with improved aggregate utility, a multiobjective stochastic algorithm is utilized. There exist a large number of real-life optimization problems in different fields, and the optimal solutions of these optimization problems require specially designed optimization models along with the algorithms having multiobjective computational solution capabilities with stochastic nature.
Further, in WSNs, the sensor node is having limited capabilities in processing the particular event and therefore a single node would require high processing energy to solve the particular problem and the only solution in these circumstances is the distributed approach. In the distributed framework, the part of the solution concerning the particular node is required to be sent to the respective nodes. The overall communication overhead in the distributed approach is less as compared to a centralized approach since in the centralized approach, transmission mechanism occurs from the sensor node (solving) to all the other nodes as compared to the distributed approach in which the transmission mechanism occurs to a subset of nodes [167]. Further, in [168], for determining the trade-off between enhanced network lifetime and QoS, a multiobjective routing framework is towards test-bedding the existing routing framework for exploring the performance in actual deployment with implementation issues. 2020 Rathore et al. [128] OOR Delay, packet delivery ratio, and throughput (i) An optimal opportunistic routing framework is presented with consideration of a dynamic duty cycle based on the modified echo state network for EH-WSNs.
(ii) In the proposed framework, the whale optimization algorithm is used for enhancing the prediction efficiency of the echo state network.
(i) Further, future research directions may consider the work towards test-bedding the existing routing framework for exploring the performance in actual deployment with implementation issues.

2020
Singh and Pattanayak [129] MEHOR Efficiency, throughput, goodput, source sending rate, data delivery ratio, and hop counts (i) An opportunistic routing framework is presented with consideration of cross-layer design.
(ii) The proposed framework utilized the Markovian model for cross-layer design.
(i) Further, future research directions may consider the work towards key challenges in the proposed framework such as reducing the duplicate packets. Number of nodes available and throughput (i) A new clustering routing protocol is presented for EH-WSNs which is known as an adaptive energy harvesting-aware clustering routing protocol.

Wireless Communications and Mobile Computing
(ii) In the proposed framework, energy state of the node is considered for selecting the cluster head.
(iii) The parameters in the proposed framework can be adjusted as per the deployed environment.
(i) Future research directions may consider the work towards some more advanced energy-efficient schemes with more metrics for cluster head elections.

2013
Xiao et al. [131] Throughput, number of dead nodes, and data failure rate (i) A new function has been proposed which is known as the energy potential function for assessing the capabilities of sensor nodes in terms of harvested energy.
(ii) Next, for EH-WSNs, a new protocol known as energy potential LEACH has been presented which is the extended version of the LEACH.
(i) Future research directions may consider the work towards applying the proposed framework in the practical environment with consideration of two constraints related to sensor nodes such as the capacity of the battery and limited computing capabilities.
Zhang et al. [132] Number of rounds until the first node dies (i) A new framework has been proposed for enhancing the lifespan of the network.
(ii) In the proposed framework, the relay nodes for cluster heads are actually the energy harvesting sensor nodes.
(i) Future research directions may consider the work towards proposing a new distributed EH clustering scheme.
(ii) The next work may be related to exploring energy harvesting nodes with different configurations.
(iii) The next work may be related to using the testbed for evaluating the proposed framework.

2014
Mostafa and Hassan [133] Number of alive nodes (i) A new clustering algorithm has been presented for EH-WSNs in which three factors are considered primarily such as sensor node centrality, amount of energy harvested, and the total neighbors. (ii) Next, fuzzy petri nets are used in the proposed algorithm for selecting the cluster heads.
(i) Future research directions may consider the work towards using type-2 fuzzy petri nets and other advanced versions such as extended TOPSIS for selecting the cluster heads.

2015
Peng et al. [134] Accumulated cluster failure time and total amount of information bits (i) A new framework has been proposed for clustering in EH-WSNs and is known as the distributive energy-neutral clustering protocol.
(ii) The proposed framework utilizes the novel cluster head group mechanism.
(iii) Next, for efficiently handling the high traffic load, a cluster is allowed to use multiple cluster heads.
(i) Future research directions may consider the work towards using hybrid advanced metaheuristic algorithms for finding the solution to the optimization problem in terms of the optimal number of clusters.
Yukun et al. [135] Total number of cluster heads, alive nodes, success ratio of data transmission, and residual energy (i) A new clustering routing framework has been presented for solar energy-harvested WSNs.
(ii) The proposed framework utilized the energy threshold factor for reviving the sensor nodes.
(i) Future research directions may consider the work towards using other energy harvesting environments since the proposed framework dedicated to solar energy harvesting.
Li and Liu [136] Normalized average throughput, awake nodes, and residual energy (i) A new clustering routing framework has been presented for EH-WSNs.
(ii) The proposed framework utilized discrete particle swarm optimization.
(i) Future research directions may consider the work towards optimization of the sensor nodes number in each cluster.
(ii) Next, the future work may consider two 24 Wireless Communications and Mobile Computing (iii) Next, in the proposed framework, the modified discrete particle swarm optimization algorithm has been utilized for determining the optimal topology. performance metrics, namely, throughput and alive nodes for comparing the proposed framework with the distributed framework.

2016
Li and Liu [137] Normalized average throughput, awake nodes, and residual energy (i) A new distributed clustering routing framework is presented for EH-WSNs.
(ii) In the proposed framework, the cluster head electing mechanism is based on the current residual energy profile of the sensor nodes and the amount of harvested energy.
(iii) Also, the proposed framework utilized the model for predicting solar energy based on a neural network.
(i) Future research directions may consider the work towards dynamic tuning of various coefficients that are used in the framework for overall improving the performance.
(ii) The next future work may consider other different prediction models for solar energy for improved accuracy.

2017
Bozorgi et al. [138] Alive nodes and average energy of alive nodes (i) A new hybrid framework has been proposed consisting of both static and dynamic clustering mechanisms.
(ii) The distributed-centralized approach with multihop routing has been used in the proposed framework.
(iii) The proposed framework considers the three primary factors for the clustering mechanism such as the current level of energy, the number of neighbors, and the amount of harvested energy.
(i) Future research directions may consider the work towards some more advanced energy-efficient schemes with more metrics for cluster head elections in the clustering mechanism.

2019
Afsar and Younis [139] Live nodes, throughput, consumed energy, and maximum hop count (i) A new cross-layer design framework has been presented with capabilities such as energy scavenging and transfer for EH-WSNs.
(ii) The proposed framework adopted a distributed two-tier routing topology.
(iii) The proposed framework divided the network into virtual tracks further the width is based on the density of node and distance to the base station.
(i) Future research directions may consider the work towards some more advanced energy-efficient schemes with more metrics for cluster head elections in the clustering mechanism.

2020
Ge et al. [140] Packets received by the base station, packets received by the cluster head, and failure node ratio (i) A new uneven clustering framework has been presented for EH-WSNs.
(ii) In the proposed framework, the energy prediction model is based on a long shortterm memory.
(iii) Also, the proposed framework considers the energy consumption of the node with supplement models for clustering mechanism.
(i) Future research directions may consider the work towards enhancing the rate of dynamic data transmission by automatically adjusting the network parameters and measuring the performance with the help of more simulations.
Sah and Amgoth [141] Network lifetime, alive nodes, and total commutative packets arrived at base station (i) A new clustering framework has been presented for solar EH-WSNs. (ii) The proposed framework utilized the hierarchical clustering approach.
(i) Future research directions may consider the work towards using advanced efficient harvesting devices for EH-WSNs. Kumar and Chaparala [142] Energy consumption, throughput, delay, delivery ratio, drop ratio, and network lifetime (i) A new clustering framework has been presented utilizing a hybrid optimization approach consisting of two optimization algorithms such as bacterial foraging and fruit fly optimization algorithm for cluster head selection in the EH-WSNs.
(i) Future research directions may consider the work towards using other advanced hybrid models of optimization algorithms for selecting the cluster heads.

Wireless Communications and Mobile Computing
presented. The framework focused on battery cost/budget along with maximizing the residual energy of the forwarding set. Both the abovementioned literature used the heuristic strategy. The heuristic approach is usually utilized for finding the solution to multiobjective optimization problems, and this fact is attested by currently available various kinds of literature. Further, for finding the solution in a relatively practical time, a heuristic algorithm can be used which utilized the concept of the trial-and-error approach. The heuristic method majorly provides reasonably accurate solutions.
Furthermore, the next important issue related to WSNs is regarding efficiently localizing the sensor nodes. An efficient optimization framework with multiobjective consideration is proposed in [169] for localizing the sensor nodes accurately to assess the geographical relevance of data. Next, coverage efficiency is also considered as one of the crucial issues for the performance evaluation of WSNs. Again, various literature have already been published with multiobjective formulations. They focus on enhancing the coverage with consideration of other desirable objectives also. The best solution is selected among multiple optimal solutions in the multiobjective optimization framework with consideration of a particular objective that is to be achieved [170]. Both the abovementioned literature used the evolution-based strategy. A population-based approach is used in an evolutionary multiobjective optimization framework. Here, in a particular iteration, multiple solutions take part, and a further new set of solutions evolves in the subsequent iteration. Next, derivative information is not needed in the evolutionbased optimization framework, and therefore, their implementation is easy. Evolution-based optimization frameworks have a wide area of applicability and also the capability for providing solutions to complex optimization problems with multiobjective nature.
Moreover, the deployment strategies in WSNs have an immense impact on the performance since the strategies adopted in deployment have a direct relationship with the power efficiency of sensor nodes and therefore deployment issue is also considered an important constraint for WSNs. Therefore, it can be concluded that optimal deployment strategies improve the energy efficiency of WSNs and consequently lifetime of the network enhanced; also, these strategies contribute towards enhancing the overall performance of the network. In [171], metaheuristic algorithms based on various approaches are presented for resolving the deployment issue in WSNs. The metaheuristic approach is considered better than the heuristics approach. The metaheuristic framework [172] has two primary components such as randomization and the selection of the optimal solution. The randomization components contribute to avoiding the trapping of solutions at local optima. The best solution selection step helps in ensuring that the solutions will converge to optimality.
Next, in WSNs, two conflicting objectives such as efficient connectivity and network lifetime are required to be optimized by optimization formulation with multiobjective consideration for enhancing the overall performance [173]. The ultimate aim of the framework is to provide superior connectivity to the other schemes by considering the same energy conservation profile. Since superior connectivity along with enhanced lifetime are two prominent factors that greatly affected the performance, there is an urgent requirement of optimizing all objectives simultaneously by a single solution, but traditional multiobjective formulations are not capable to satisfy this requirement. The properly designed multiobjective optimization formulation can provide plenty of alternative solutions. Further, the location of these alternative solutions is nearby or on the Pareto optimal front. Also, only in a single run, plenty of Pareto optimal solutions can be discovered by nondominated sorting algorithm II.
Besides, in WSNs, energy-efficient or energy-aware routing frameworks are needed for the effective transmission of data packets. Various network attributes such as lifetime, overhead in communication, and availability of data are influenced by energy-efficient routing frameworks. Energyefficient or energy-aware routing frameworks are aimed at maximizing system performance. For optimization and computational intelligence, bio-inspired algorithms are extensively used currently. Bio-inspired schemes consist of dual approaches such as reactive and proactive approaches. Bioinspired schemes are capable enough for accomplishing adaptive routing, enhanced load balancing, and also discovering network topology [174].
Also, in WSNs, the bit error rate should be minimum, and on the other hand, signal-to-noise ratio should be maximum, and for achieving these conditions, transmission power should be increased. The increment in the transmission power has a severe impact on several key attributes of  [143] Energy consumption, delay, delivery ratio, and throughput (i) A new clustering framework is presented utilizing a hybrid optimization approach consisting of whale and grey wolf optimizer.
(ii) In the proposed framework, the exploitation and exploration abilities are significantly improved with the help of the hybrid approach.
(i) Future research directions may consider the work towards using other advanced hybrid models of optimization algorithms for the clustering mechanism. 26 Wireless Communications and Mobile Computing Transmission quality (i) An efficient framework has been proposed for controlling the rates of source coding adaptively by appropriately selecting the most eligible data packets for transmission.
(ii) The proposed framework considers four primary factors for the selection of eligible packets such as power efficiency, energy cost, energy-neutral constraint, and also multimedia distortion reduction.
(i) The future research directions may consider extension towards utilizing other energy harvesting model since the current work utilizes the solar energy harvesting system.

2013
Gregori and Payaró [152] Mean minimum completion time, percentage of solutions (i) An optimal data transmission scheme has been proposed.
(ii) The proposed scheme considers the three constraints such as constraints for quality of service, data causality constraints, and also constraints for energy efficiency.
(i) The future research directions may consider extension towards fulfilling some other constraints for the optimal data transmission scheme.
Roseveare and Natarajan [153] Sum rate, inverse mean squared error (i) Models have been presented for different applications in the EH-WSNs with system-wide utility.
(ii) The proposed models consider the presence of factors like uncertain resource availability with system parameters.
(i) The future research directions may consider extension towards developing the framework for utility maximization decentralized problems also the proposed framework will allow the problem decomposition.

2014
Wang et al. [154] Average reward rate (i) An optimal offline framework has been proposed for maximizing the weighted throughput.
(ii) Two optimal deterministic online algorithms have also been proposed along with the randomized algorithm.
(i) The future research directions may consider extension towards developing the framework for semionline algorithms.
(ii) The aim of the proposed framework is to evaluate the joint optimal transmission policy for EH-WSNs.
(iii) In the proposed framework, throughput has improved for a given deadline.
(i) The future research directions may consider extension towards considering other metrics for the evaluation of the proposed framework under different energy harvesting models.

2015
Michelusi and Zorzi [156] Network utility (i) Efficient access schemes are proposed with decentralized nature.
(ii) Also, each sensor node can individually decide whether to transmit a specific eligible packet or discarding the particular packet.
(iii) The decision on packets is based on certain factors such as estimation of packet's utility, the current state of energy harvesting mechanism, and finally energy profile.
(iv) The aim is to enhance the aggregate longterm utility of the packets.
(i) The future research directions may consider extension towards considering some other operational regimes since the current work identified two operational regimes such as the network-limited scenarios and energy-limited scenarios.
Peng and Low [157] Distinct packet delivery ratio (i) A new framework has been proposed which is driven by query and having characteristics of energy-neutral with directed diffusion.
(ii) The proposed framework is aimed at maintaining the energy-neutral state at the network level.
(iii) The proposed framework results in enhanced data delivery and also the improved lifetime of the network.
(i) The future research directions may consider extension towards considering some more performance metrics for evaluating the performance of the proposed framework.  [158] Average reward, percentage of used slots, average fidelity (i) Efficient policies have been proposed to control lossy data compression over fading channels.

Wireless Communications and Mobile Computing
(ii) The proposed schemes consider the replenishable energy buffer along with the specific process having stochastic energy input.
(i) The future research directions may consider extension towards modeling of the coupled dynamics of multiple energy harvesting sensors with multiuser scenarios.

2016
Zhou et al. [159] Average mean square error (i) A new framework has been proposed for investigating the source estimation problem in EH-WSNs.
(i) The future research directions may consider extension towards finding the more efficient solution to the stochastic constraints since the current proposed framework may increase the estimation distortion.
Gong et al. [160] Throughput (i) An efficient centralized scheme has been proposed for achieving the optimal throughput by utilizing the monotonicity which exists in the problem structure.
(ii) Further, a distributed suboptimal scheme has been proposed in a game-theoretic approach.
(iii) Next, the source and the relays are responsible for the proposed framework to iteratively update two primary factors such as beamforming vector and energy harvesting schedule.
(i) The future research directions may consider extension towards developing relay network with multihop transmission and also having energy harvesting constraints.
Sunny [161] Quality of monitoring (i) An efficient joint scheduling and sensing allocation problem has been formulated which is attributed with a long-term time-averaged characteristic.
(ii) The formulated problem considers the finite data buffers, specific energy profile, and constraints such as quality of service related to battery and certain data constraints.
(i) The future research directions may consider extension towards utilizing the variable energy profile with different size data buffers also considering various other constraints.  [175], for minimizing the interferences and maximizing the throughputs, a multiobjective memetic algorithm is used to design efficient power allocation techniques along with a spectrum sensing module. Memetic algorithms are having the computational intelligence structures; further, the trial-and-error approach is used for discovering the Pareto optimal solution set. Next, in WSNs, reliability and delay are also important parameters that need to be optimized for enhancing the system performance. In [176], an efficient routing framework is presented which utilized the multiobjective optimization approach. One of the most desired objectives is the imple-mentation of quality of service, and this objective conflicts with other objectives such as lifetime, delay, and network cost. The routing optimization model utilized the fuzzy random variables for representing the randomness and fuzziness of the objectives and constraints. The mathematical notations are used for representing human reasoning in the Fuzzy logic approach. Fuzzy logic utilized the inference rules and linguistic variables for establishing the approximation for the truth value of a proposition.
It should be noted that high transmission power is required for increasing the transmission range of sensor nodes and consequently the degree of sensor node increases, but it has a high impact on energy efficiency and results in reduced performance in terms of energy-saving. Also, one of the attributes of WSN known as network reachability is  Figure 11: A broad classification of various existing challenges for green WSNs. 29 Wireless Communications and Mobile Computing usually low in case of poor connectivity of WSNs; on the other hand, strong connectivity in the network requires a high demand of power and consequently decreases the quality of service in WSNs. An extensive investigation is needed to analyze the optimal transmission power requirement of a sensor node. Further, for obtaining the energy-efficient clustering-based routing framework, an optimization formulation with multiobjective consideration is presented in [177]. The trade-offs among total packets delivered to the base station, the lifetime of the network, energy utilization, and dead nodes are achieved in the proposed framework. Particle swarm optimization is used in the framework. Particle swarm optimization is based on swarm behavior. In recent years, various optimization formulations have been solved by the particle swarm optimization algorithm, and therefore, it has attained immense recognition as compared to other optimization algorithms.
Also, in WSNs, plenty of tiny sensor nodes are deployed densely; also, the spatial distribution of sensor nodes affected the overall performance which results generally in terms of the error in the reconstruction of the physical signal. In a distributed tracking framework, the sensors' assignment to fusion centers in a dynamic fashion is presented in [178]. A framework is proposed in which the original problem is decomposed into subproblems by taking the concept of a genetic algorithm approach for finding real-time suboptimal solutions. Over the last two decades, genetic algorithm has been proved as an efficient search technique for solving various optimization problems. The genetic algorithm is inspired by Darwin's principle of natural evolution and starts with a set of randomly generated candidate solutions, denoted as population. Each individual in the population corresponds to a candidate solution to the problem, and it is represented by a genetic code.

Trendwise Technical Analysis of Energy
Harvesting Technologies Development with Future Aspects The development trends for various energy harvesting techniques can be majorly divided into four eras yearwise as illustrated in Figure 13. Development trends in the summarized form with the scientific progress for different energy harvesting technologies have been demonstrated in this figure.
The first era covers the development activities in a duration of ten years, i.e., from 2000-2010. The key attributes of this era in terms of significant development for energy harvesting technologies are the development of energy harvesting devices with enhanced fabrication technologies and further inclination of trends towards developing the hybrid model for energy harvesting systems.
The second era covers the development activities in the duration of five years, i.e., from 2010-2015. The significant key advancement towards energy harvesting technique can be highlighted as development towards triboelectric nanogenerators technology, development of multiple system architectures for energy harvesting devices and also the development of hybrid energy harvesting system model, and further inclination of trends towards integrating the Industry 4.0 with energy harvesting technologies.
The third era covers the development activities in the duration of five years, i.e., from 2015-2020. The important key attention-seeking developments for energy harvesting  Furthermore, the trends have started towards exploring the use of energy harvesting technologies for uplifting the current scenario of smart cities and societies as well and in the future; this trend will be further boosted with the acceleration towards the development of the technologies for small-scale selfpowered devices with the system-level implementation. Next, machine learning will be used for analyzing the signal of the energy harvester for predicting the futuristic environmental conditions. The research is going on exploring new metamaterials for energy harvesting with machine learning. Further, research is going on exploring new emerging technologies for enhancing the performance of the energy harvesting system.

Conclusion
In this systematic survey, we have conducted a deep study for the energy optimization issue of EH-WSNs. The ultimate aim is to enhance the life span of sensor nodes. In this high-level systematic and taxonomical survey, we have organized the energy optimization strategies for EH-WSNs into eleven main classes such as schemes based on optimization of radio, schemes based on optimization of the energy harvesting process, schemes based on reduction of data in the system, schemes based on sleep/wake-up policies, schemes based on load balancing, schemes based on optimization of power needed by the hardware devices in the system, optimization of communication mechanisms, schemes based on optimization of battery operations, mobility-based schemes, and finally energy balancing schemes. This systematic and taxonomy survey also provides a progressive detailed overview and classification of various optimization challenges for the EH-WSNs that require attention from the researcher followed by a survey of corresponding solutions for corresponding optimization issues. Further, this systematic and taxonomical survey also provides a deep analysis of various emerging energy harvesting technologies in the last twenty years of the era. From the study, it can be concluded that energy harvesting technology can be used as an alternative source of energy for sensor nodes, but environmental heterogeneity compels us to think about strategies for effective utilization of harvested energy and energy optimization strategies play a major role in achieving this objective. There is a scope of enhancing the efficiency of the energy harvesting system, and therefore, a lot of research is still going on optimizing the energy harvesting mechanism. Although optimization of the energy harvesting mechanism possesses several challenging issues that need to be handled effectively, further, it should be worth mentioning that we should explore the possibility of new sources of energy for energizing the tiny sensor nodes. Besides, a hybrid approach can be considered for handling the energy scarcity issue of tiny sensor nodes that comprising all three pillars of energy in WSN, namely, harvested energy from the environment, batteries, and finally wireless energy transfer mechanism, but again this approach is an open research area and sincere efforts are needed to counter several challenges and, in this way, a sustained WSN may be realized.

Conflicts of Interest
The authors declare that there is no conflict of interests regarding the publication of this paper.

2000-2010
In this era, the summary of key developments is given below:-In this era, the summary of key developments is given below:- The integration of various fabrication technologies is another significant achievement is this era.
The research had been started towards developing various hybrid models for energy harvesting systems.

2015-2020
In this era, the summary of key developments is given below:-This era had achieved main achievement in integrating energy harvesting technology with IoT. The use of energy harvesting technique in the cross layer application is another significant achievement in this era. In decenber 2015, the concept of industry 5.0 had arised. Self-powered IoT, the new paradigm, had evolved.
Significant achievements towards development of renewable energy harvesting devices.
This era is the witness of revolutionary evolution in the field of energy harvesting technologies.
This era is the witness of development towards promising research field, namely, triboelectric nanogenerator (TENG) technologies.
In this era, various system architectures along with hardware devices had development energy harvesting system.
The concept of hybrid energy harvesting system had also evolved in this era.
The research had been started towards integrating energy harvesting technologies into industry 4.0.

Next generation technologies & further of energy harvesting techniques
In the coming years, energy harvesting techniques will pay crucial role in various areas such as:- The energy harvesting techniques will play a major role in various application for advancement of smart cities and societies.
The machine learning will be used to analyze the energy harvester's signal for predicting the futuristic environmental conditions.
In 2011, the concept of industry 4.0 had arised.

2010-2015
Artificial intelligence and energy harvesting techniques will be used for enhancing the life time of sensors and IoT devies.
The research is going on new metamaterials for energy harvesting with machine learning.
Use of various emerging technologies for enhancing the performance of energy harvesting system.