Energy management in harvesting enabled sensing nodes: Prediction and control
Introduction
The number of Internet of Things (IoT) devices is increasing exponentially (Ericsson mobility report.) and the energy demand for IoT related applications thus continues to grow (Ejaz et al., 2017). Energy efficiency and the life span of IoT devices are key challenges to next generation IoT based solutions (Kaur and Sood, 2017; Arshad et al., 2017; Kolios et al., 2016; Iannacci, 2018). Therefore, energy management in IoT technologies, is a critical issue. Since, wireless sensing nodes are an integral part of IoT technology (Wadud et al., 2017; Shin and Joe, 2016) the problem of energy management in IoT sensing nodes inherits many of the characteristics of the energy management problem in Wireless Sensor Networks (WSNs) (Ulukus et al., 2015; Anastasi et al., 2009; Ali et al., 2016). Thus throughout the paper, the term “node” is used for IoT and WSN nodes interchangeably. The problem was originally challenged by the limited energy storage capabilities and the independent operation of IoT nodes in remote, often hazardous environments (Sudevalayam and Kulkarni, 2011). New challenges and design options later emerged as a result of the integration of the harvesting capability of ambient energy and the emergence of new networking paradigms, as for example in smart cities (Ejaz et al., 2017), body sensor networks (Seyedi and Sikdar, 2010; Zuhra et al., 2017), habitat monitoring (Mainwaring et al., 2002), volcano monitoring (Werner-Allen et al., 2006), structural monitoring (Chebrolu et al., 2008), vehicle tracking (Karpiriski et al., 2006), and more recently nano-networking (Afsana et al., 2018).
The energy harvesting (EH) capability, whether this is solar, vibration, thermal or energy from radio waves, significantly affects the energy management design (Kim et al., 2014). It is well established that the harvesting energy capability penetrates in all layers of network protocol design leading to numerous harvesting aware solutions in a number of network related problems (Ulukus et al., 2015; Adu-Manu et al., 2018): in topology control (Tan et al., 2015; Li et al., 2016; Peng et al., 2015), in routing (Hieu Kimet al., 2016; Ding et al., 2016), in medium access control (Fafoutis et al., 2015; Tadayon et al., 2013; Nguyen et al., 2017; Sherazi et al., 2018) in transmission policies (Ho and Zhang, 2012; He et al., 2014; Gupta et al., 2016), in scheduling based congestion control (Li et al., 2017), in data cycling (Niyato et al., 2007; Jaggi et al., 2009) and in admission control (Gatzianas et al., 2010). A primary objective in many of the aforementioned works is to achieve energy neutral operation i.e. ensure that the energy consumed is always smaller than the energy harvested. In (Tan et al., 2015) for example, the problem is cast as a game theoretic topology control problem, motivated by the fact that the available and harvested energy affect the link formation between nodes, as long-distance links require large power transmissions to be established. The sensor behavior is modeled as a potential cooperation game to maintain connectivity and the achievable Nash equilibrium is shown to reduce energy consumption and balance the load distribution. In energy aware routing approaches, the residual and harvested energy affect the routing decisions with a recent approach adopting geographic routing and adapting greedy forwarding to account for the residual energy, the harvested energy and the link quality (Hieu Kimet al., 2016). MAC layer parameters may also be adapted based on energy considerations. For example, in (Fafoutis et al., 2015), a receiver initiated asynchronous design is considered and harvested energy considerations are used to control the duty cycle of beacon message generation. Finally, sensor data packet generation scheduling can also become energy aware with a recent work in (Li et al., 2017) attempting to minimize packet losses in terms of the nodes' energy consumption and data queue state information. It must be noted that the literature on harvesting enabled WSNs is extensive and a nice recent review can be found in (Adu-Manu et al., 2018).
In this work, we examine the problem of determining the transmission policy of wireless sensing nodes, taking into account the constrained energy resources and the EH capability. Several works have appeared in the literature mostly involving stochastic representations of the underlying processes (Ulukus et al., 2015; Adu-Manu et al., 2018). Stochastic approaches in many cases provide a more realistic representation of the underlying processes and dynamics as energy storage and consumption on wireless nodes is commonly stochastic in nature. These approaches rely on different formulations of the problem, considering different decision variables of the transmission policy, different design objectives and different characteristics of the stochastic processes involved, whether these are used to represent the energy harvested, the storage devices, the data queuing disciplines or the communications channel. Decision variables include the wake up schedule of the sensing nodes (Jaggi et al., 2009; Wang et al., 2016), the transmission probability (Michelusi et al., 2013), the transmission mode (Seyedi and Sikdar, 2010), the energy allocated for transmission (Sharma et al., 2010), and the transmission rate (Yang and Ulukus, 2012). Different design objectives have also been considered as for example the probability of future energy depletion (Seyedi and Sikdar, 2010), the likelihood of data being correctly detected (Seyedi and Sikdar, 2010), the average long term importance of the reported data (Michelusi et al., 2013), the detection probability (Jaggi et al., 2009), the achieved throughput (Sharma et al., 2010; Ozel et al., 2011; Tutuncuoglu and Yener, 2012; Orhan et al., 2012; Tacca et al., 2007; Ho and Zhang, 2012; Huang et al., 2013; Xu and Zhang, 2014; Tadayon et al., 2013) and the packet delivery time or transmission completion time (Yang and Ulukus, 2012; Antepli et al., 2011; Ozel et al., 2012; Wang et al., 2016). In addition, memoryless (Lei et al., 2009) and temporally correlated (Michelusi et al., 2013; Jaggi et al., 2009) models have been considered, infinite (Kashef and Ephremides, 2012) and finite (Michelusi et al., 2013; Ozel et al., 2012; Srivastava and Koksal, 2013) buffer capacities have been accounted for, some cases in the presence of fading communication channels (Yang and Ulukus, 2012; Ho and Zhang, 2012), Gaussian relay channels (Huang et al., 2013) or additive white Gaussian channels (Xu and Zhang, 2014; Antepli et al., 2011; Ozel et al., 2012). Note also that battery imperfections have been considered in (Devillers and Gündüz, 2012), non ideal circuit power transmitter has been considered in (Xu and Zhang, 2014), multi-hop duplex communications have been considered in (Rezaee et al.,), reinforcement learning techniques have been incorporated in (Aoudia et al., 2018) and recently in (Luo et al., 2018) it has been identified that the transmission scheduling policy does affect the EH model, a process which is naturally modeled by feedback. The most prominent solutions which have been obtained include the save-and transmit where transmitter does not send anything for a fraction of time p (to save energy) and transmits data in the remaining time 1-p, the best-effort transmit where all the time is devoted exclusively for data transmission, the energy feasible tunnel approach and variations of the directional water filling algorithm e.g. directional iterative water-filling algorithms and two-dimensional directional water-filling algorithms (Ulukus et al., 2015). Significant above works involve stochastic representations of the underlying processes (Ulukus et al., 2015). However, stochastic approaches lead to representations which make analysis and design complex in nature and sometimes intractable. Deterministic approaches, however, are in many cases easier to analyze and lead towards simpler to implement solutions. This thrust has led to the adoption of deterministic approaches in other problems exhibiting similar characteristics to the energy control problem, as for example the congestion control problem in computer networks. So, in this work we adopt a deterministic model approach, we view the problem as a queue control problem and we demonstrate through simulations that the approach is able to outperform a characteristic stochastic approach, despite the simplicity of the deterministic model. The other advantage of considering the problem as a queue control problem in a deterministic model framework, is the availability of off the shelf solutions which have been developed in the context of other problems e.g. the congestion control problem. This constitutes the major contribution of this work which paves the way to adopt algorithms from the rich literature of congestion control algorithms available in the literature.
Recent research efforts in literature have established that EH aware protocols, can be greatly improved upon availability of energy prediction strategies, which generate predictions of harvested energy to be used for better energy provisioning. In (Zou et al., 2016), node clustering and routing algorithms are proposed to optimize data transmission, based on future predictions of harvested solar energy. Moreover, in (Saleem et al., 2016) a power management scheme for the throughput maximization problem using energy predictions is proposed for the autonomous mode of device-to-device (D2D) communications. In (Saidi et al., 2016), predicted energy based on Kalman filtering is used to regulate the number of bits sent by the transmitter during a time slot in point-to-point communication between wireless nodes. In (Jushi et al., 2016), power in WSNs is controlled using predictions of wind energy, while adaptive control of the packet transmission period with solar EH prediction is proposed in (Kwon et al., 2015). Such prediction schemes (Qureshi et al., 2017) can lead to better energy management of the available and harvested resources and lead to protocols with improved properties and an effective energy provisioning system.
In this work, we consider the energy management problem in IoT sensing nodes and in particular the problem of regulating the transmission rate in the presence of harvested energy whose levels can be predicted. We adopt a design approach based on control theoretic considerations, which is different from the overwhelming body of existing works in the literature. We view the problem as a queue control problem where the objective is to regulate the transmission and thus the energy leaving the battery in order to ensure that at equilibrium the energy level within the battery converges to a constant reference chosen by the designer. Maintaining this reference value ensures that the battery is not depleted and that some energy is stored for crucial or emergency operations. In particular, we consider a non-linear model of the queuing dynamics first introduced in (Pitsillides et al., 2005), to derive non-linear robust controllers for serving the energy management policy. A notable feature of the proposed protocol is that it incorporates predictions of the energy to be harvested which are generated using the proposed Accurate Solar Irradiance prediction Model (ASIM) and is based on Markov Chains of increasing order (Ghuman et al., 2015). The stability of the proposed controllers is established analytically and the performance of the combined prediction and control policy is investigated using simulations conducted on the Network Simulator (NS-3). Our simulation experiments indicate that both derived controllers are successful in guiding the energy level to the desired values and that the addition of integral action is beneficial in terms of the throughput achieved. Moreover, a sensitivity analysis is performed to investigate the effect of changing the design parameter α and the sampling period. Finally, the proposed approach is compared against the Throughput Optimal (TO) policy (Sharma et al., 2010) which is chosen as the benchmark solution on the basis that it is a good representative of throughput maximization policies which are abundant in the literature. Higher throughput is achieved by the proposed strategy. The current work extends our previous works in (Ghuman et al., 2015; Ashraf et al., 2017). The main new contributions of the paper are the following:
- •
Extension of the ASIM model to account for both long term and short term prediction and demonstration of its effectiveness.
- •
Introduction of the proportional and integral controller and demonstration of its effectiveness using both analysis and simulations.
- •
Extensive performance evaluation that includes sensitivity analysis of the proposed schemes (to highlight the robustness) and a comparative analysis relative to an existing approach, namely the TO policy.
The paper is organized as follows: In section 2, we describe and validate the proposed prediction scheme, in section 3 we formulate the energy management problem as a queue control problem and we derive non-linear controllers whose convergence properties are established analytically, in section 4, we evaluate the performance of the combined prediction and control scheme using simulations and finally in section 5 we offer our conclusions and future research directions.
Section snippets
Energy prediction
The proposed energy management scheme incorporates predictions of the energy to be harvested. In the proposed scheme, the predictions are obtained using the ASIM model. In this section, we review background theory on the ASIM model and describe how Markov chains of increasing order are used as a baseline to develop the proposed prediction model. We present ASIM model as a generic model for both long term and short term prediction. Central elements of the proposed model are state dependencies of
Control based energy management
In this section, we develop and analyze control based energy management policies for IoT which incorporate energy predictions. The main feature of the proposed scheme, which differentiates it from previous approaches, is that the battery at each node is modeled as a M/M/1 queue accommodating energy “packets” and the energy management problem is viewed as a queue control problem. The objective of control problem is to regulate the amount of energy which is made available for packet transmission
Performance evaluation
In the previous section, using analysis, performance bounds were derived for provable controlled behavior. In this section, we demonstrate using simulations conducted on the NS-3 that the proposed approach achieves the desired behavior and performance. For this, a network of 100 harvesting enabled wireless sensor nodes is considered where nodes are placed in an area of 1500 × 1500 m2 following a uniform random distribution. 802.11 transceivers are used with the transmission power value set to
Conclusions
In this paper, we view the energy management problem in EH enabled IoT sensing nodes, as a queue control problem, where the objective is to regulate transmission so as to guide the battery energy level to a predetermined reference level, and we propose non-linear controllers whose stability properties are established analytically. The proposed strategies utilize predictions of the energy to be harvested which are generated using the proposed ASIM model. The effectiveness of the proposed method
Nouman Ashraf received the BS in Electrical (Telecommunication) Engineering and MS in Electrical (Control Systems) engineering from COMSATS Institute of Information Technology in 2012 and 2015 respectively. Currently, he is pursuing his PhD in Electrical Engineering from Frederick University, Cyprus under Erasmus Mundus Scholarship Program. His research interests include application of control theory and optimization methods for energy management in communication and power networks.
References (70)
- et al.
Energy efficient techniques for m2m communication: a survey
J. Netw. Comput. Appl.
(2016) - et al.
Energy conservation in wireless sensor networks: a survey
Ad Hoc Netw.
(2009) Internet of things (iot); internet of everything (ioe); tactile internet; 5g–a (not so evanescent) unifying vision empowered by eh-mems (energy harvesting mems) and rf-mems (radio frequency mems)
Sensor Actuator Phys.
(2018)- et al.
The impact of node position on outage performance of rf energy powered wireless sensor communication links in overlaid deployment scenario
J. Netw. Comput. Appl.
(2016) - et al.
Energy harvested roadside ieee 802.15. 4 wireless sensor networks for iot applications
Ad Hoc Netw.
(2017) - et al.
Energy neutral clustering for energy harvesting wireless sensors networks
Ad Hoc Netw.
(2015) - et al.
A comprehensive review on energy harvesting mac protocols in wsns: challenges and tradeoffs
Ad Hoc Netw.
(2018) - et al.
Power management in smac-based energy-harvesting wireless sensor networks using queuing analysis
J. Netw. Comput. Appl.
(2013) - et al.
Energy harvesting aware topology control with power adaptation in wireless sensor networks
Ad Hoc Netw.
(2015) - et al.
An energy-efficient sdn based sleep scheduling algorithm for wsns
J. Netw. Comput. Appl.
(2016)
Routing protocols in wireless body sensor networks: a comprehensive survey
J. Netw. Comput. Appl.
Energy-harvesting wireless sensor networks (eh-wsns): a review
ACM Trans. Sens. Netw.
An energy conserving routing scheme for wireless body sensor nanonetwork communication
IEEE Access
Optimal packet scheduling on an energy harvesting broadcast link
IEEE J. Sel. Area. Commun.
Rlman: an energy manager based on reinforcement learning for energy harvesting wireless sensor networks
IEEE Trans. Green Commun. Network.
Green iot: an investigation on energy saving practices for 2020 and beyond
IEEE Access
Active energy management for harvesting enabled wireless sensor networks
Brimon: a sensor network system for railway bridge monitoring
Stochastic modelling of energy harvesting for low power sensor nodes
A general framework for the optimization of energy harvesting communication systems with battery imperfections
J. Commun. Network.
Optimizing routing based on congestion control for wireless sensor networks
Wireless Network
Efficient energy management for the internet of things in smart cities
IEEE Commun. Mag.
Energy-efficient medium access control for energy harvesting communications
IEEE Trans. Consum. Electron.
Control of wireless networks with rechargeable batteries [transactions papers
IEEE Trans. Wireless Commun.
Asim: solar energy availability model for wireless sensor networks
Energy harvesting aided device-to-device communication underlaying the cellular downlink
IEEE Access
Recursive waterfilling for wireless links with energy harvesting transmitters
IEEE Trans. Veh. Technol.
Stability-aware geographic routing in energy harvesting wireless sensor networks
Sensors
Optimal energy allocation for wireless communications with energy harvesting constraints
IEEE Trans. Signal Process.
Throughput maximization for the Gaussian relay channel with energy harvesting constraints
IEEE J. Sel. Area. Commun.
Rechargeable sensor activation under temporally correlated events
Wireless Network
Wind energy harvesting for autonomous wireless sensor networks
Sensor networks for smart roads
Optimal packet scheduling for energy harvesting sources on time varying wireless channels
J. Commun. Network.
Cited by (18)
A deep learning architecture for power management in smart cities
2022, Energy ReportsCitation Excerpt :The researchers reached new cutting-edge results on ISO New York and Latin America electrical utilities. Ashraf et al. use deep learning in computer vision applications, IoT, cybersecurity, medicine, etc., to provide more accurate predictions (Ashraf et al., 2019). Popular deep learning approach in energy predicting published studies, like Silva et al. suggested STLF employing resident behavior learning and LSTM, focuses on projection for apartment complexes (Silva et al., 2020).
Impact of internet of things paradigm towards energy consumption prediction: A systematic literature review
2022, Sustainable Cities and SocietyCitation Excerpt :Moreover, the sensed variable was only transmitted following the identification of a relevant shift to facilitate low computational costs and significant savings in sensor node power consumption and increase battery lifespan. Ashraf et al. (2019) suggested a design model following theoretical considerations to regulate transmission and guide the battery energy level to a predetermined reference level in EH-enabled IoT sensing nodes. The sustenance of reference values ensured against battery depletion and facilitated energy storage for vital or emergency operations.
Energy Prediction for Energy-Harvesting Wireless Sensor: A Systematic Mapping Study
2023, Electronics (Switzerland)A new generalized stochastic Petri net modeling for energy-harvesting-wireless sensor network assessment
2023, International Journal of Communication SystemsHarvested Energy Prediction Technique for Solar-Powered Wireless Sensor Networks
2023, IEEE Sensors JournalEnergy-Aware AI-Driven Framework for Edge-Computing-Based IoT Applications
2023, IEEE Internet of Things Journal
Nouman Ashraf received the BS in Electrical (Telecommunication) Engineering and MS in Electrical (Control Systems) engineering from COMSATS Institute of Information Technology in 2012 and 2015 respectively. Currently, he is pursuing his PhD in Electrical Engineering from Frederick University, Cyprus under Erasmus Mundus Scholarship Program. His research interests include application of control theory and optimization methods for energy management in communication and power networks.
Muhammad Faizan Ghumman received his Masters degree in Electrical Engineering from School of Electrical Engineering and Computer Science, NUST, Pakistan and BS in Computer Engineering from COMSATS Institute of Information and Technology, Pakistan. Currently, he is working as Senior System Engineer at Snskies solutions, Pakistan working in the domain of WSN, SDN, Virtualization and Cyber Security. His previous jobs include Research Assistant in Namal College, Pakistan, Research Assistant in WisNet Lab, NUST, Pakistan and Design Engineer at Plumgrid, Pakistan. His research interests include Efficient Energy management, Wireless Sensor Networks, cloud computing and Software Defined Networks.
Waqar Asif received his Ph.D. degree from City University of London in 2016 and since then he is working as a Research Fellow at the same. He did his MS and BEng in 2012 and 2009 respectively from well reputed institutions in Pakistan. Before moving to the UK, he served as a Lecturer in Bahria University Pakistan for a year. He secured multiple merit based scholarships which includes an Erasmus Mundus STrongTies Scholarship for one and a half year for Cyprus where he worked as a researcher in Frederick University. His research interest include but are not limited to graph theory, sensor networks, network performance metrics, BlockChian, network privacy, network security, power networks and The Internet of Things.
Hassaan Khaliq Qureshi received his Ph.D degree in Electrical Engineering from City University, London, England in 2011. Earlier, he received the M.Sc degree in Electrical Engineering with a first class honors from Blekinge Institute of Technology, Sweden in 2006. He is a recipient of EU Erasmus Mundus staff research mobility and post-doctoral fellowship under STRONG TIES and INTACT program, respectively. Currently, he is working as an Assistant Professor in the School of Electrical Engineering and Computer Science (SEECS), National University of Sciences and Technology (NUST), Pakistan. His main research interests include wireless networks, Network optimization techniques, D2D communications, Internet of Things (IOTs), Network Intrusion Detection Systems, and energy provisioning issues for infrastructure-less networks.
Adnan Iqbal has a PhD in I.T. from National University of Sciences and Technology (NUST), Pakistan. He is currently serving as Associate Professor in Computer Science Department, Namal College, Mianwali. His research interests include Software Defined Networking, Cloud Gaming, Disaster Recovery and Energy Harvesting Sensor Networks.
Marios Lestas received the B.A and M.Eng degrees in Electrical and Information Engineering from the University of Cambridge U.K and the PhD degree in Electrical Engineering from the University of Southern California in 2000 and 2005 respectively. He is currently an Associate Professor at Frederick University. His research interests include application of non-linear control theory and optimization methods in Computer and Transportation Networks.