PaperlayerView exact revision on arXiv

Source-linked AI summary

Energy Harvesting Wireless Communications: A Review of Recent Advances

Sennur Ulukus, Aylin Yener, Elza Erkip, Osvaldo Simeone, Michele Zorzi, Pulkit Grover, Kaibin Huang

arXiv:1501.06026v1cs.IT

TL;DR

Energy-harvesting wireless communications require models and strategies that account for stochastic energy availability, storage, consumption, and transfer. This review synthesizes results from information-theoretic limits through scheduling, access, networking, and energy cooperation across network scales. It identifies established capacity results and continuing challenges, while emphasizing practical constraints and energy-information transfer as directions for network operation.

  • Problem

    Energy-harvesting networks must be understood across stochastic energy availability, storage, consumption, transfer, and multiple network scales rather than through conventional communication models alone.

  • Method

    The paper reviews theory, algorithms, protocols, models, and results spanning capacity, transmission scheduling, medium access, networking, energy cooperation, and simultaneous information and energy transfer.

  • Results

    Known capacity results differ sharply across battery regimes, with Gaussian signalling for unlimited batteries and discrete signalling for no batteries, while finite-capacity noisy channels remain incompletely solved.

  • Takeaways & Limitations

    Practical design insights depend on storage imperfections, consumption and processing costs, causal harvesting profiles, and the efficiency of energy transfer.

Abstract

from arXiv · show

This article summarizes recent contributions in the broad area of energy harvesting wireless communications. In particular, we provide the current state of the art for wireless networks composed of energy harvesting nodes, starting from the information-theoretic performance limits to transmission scheduling policies and resource allocation, medium access and networking issues. The emerging related area of energy transfer for self-sustaining energy harvesting wireless networks is considered in detail covering both energy cooperation aspects and simultaneous energy and information transfer. Various potential models with energy harvesting nodes at different network scales are reviewed as well as models for energy consumption at the nodes.

I. INTRODUCTION

Energy harvesting can support self-sustaining wireless networks and applications beyond conventional battery operation. The paper introduces energy-arrival channel models whose causal, battery-dependent constraints shape capacity and network design.

  • Motivation: Energy harvesting enables wireless nodes to acquire energy from natural or man-made sources, supporting self-sufficiency and potentially perpetual operation.Sources include solar, lighting, vibration, thermal, biological, chemical, and electromagnetic energy, as well as wireless energy transfer.
  • Motivation: Energy harvesting networks may enable medical, environmental, monitoring, surveillance, and safety applications that conventional battery-powered operation cannot support.
  • Scope: The review spans information-theoretic limits, scheduling and resource allocation, medium access, networking, energy transfer, cooperation, and energy consumption across network scales.
  • Channel model: Energy causality requires cumulative energy expenditure to remain no greater than cumulative harvested energy, imposing constraints at every channel use rather than one codeword-wide average constraint.An unlimited battery permits energy harvested earlier to be used later, whereas no battery creates instantaneous stochastic amplitude constraints.
  • Channel model: In finite-battery systems, transmission changes the next battery state, while the battery state determines the allowable transmission symbols.
  • Open problems: General noisy-channel capacity remains an open problem for finite battery sizes, despite known results for unlimited, zero, and one-unit batteries in specified settings.The known cases use substantially different capacity-achieving strategies, including Gaussian codebooks for unlimited batteries and discrete signalling for no batteries.

A. Capacity with an Unlimited-sized Battery, Emax = ∞

With an unlimited battery, energy-harvesting channel constraints are stricter than the classical average-power constraint but can still attain the classical AWGN capacity. This result relies on storing energy over time and does not require energy-arrival or battery-state knowledge at either terminal.

  • Energy-harvesting codewords automatically satisfy the classical average-power constraint, so their capacity is upper bounded by AWGN capacity at average recharge rate P.
  • The save-and-transmit scheme achieves AWGN capacity by reserving an o(n) initial saving phase and using Gaussian codewords afterward.Both the saving phase and data-transmission phase grow with n, while the saving phase remains sublinear.
  • An unlimited battery is essential because it can store enough energy during saving or absorb later arrivals to prevent energy shortages.
  • Unlimited-battery capacity is achievable without transmitter or receiver knowledge of the energy-arrival process or current battery state.The battery smooths stochastic arrivals, so battery-state information does not improve achievable rates.

B. Capacity with no Battery, Emax = 0

With no battery, harvested energy cannot be saved, so channel inputs face time-varying stochastic amplitude constraints. The capacity-achieving input distribution is discrete, and capacity is significantly lower than with an unlimited battery.

  • Without a battery, channel inputs are instantaneously amplitude constrained by the current stochastic energy arrival.The transmitter observes the energy-arrival state causally, while the receiver does not.
  • The no-battery capacity is obtained by choosing state-dependent inputs within the amplitude limits associated with each observed energy state.For states e1 and e2, the inputs lie in [-√e1, √e1] and [-√e2, √e2], respectively.
  • The capacity-achieving input distribution has discrete support rather than a continuous distribution.
  • Capacity with an unlimited battery is significantly larger than capacity with no battery.

C. Capacity with Unit-sized Battery, Emax = 1

Unit-sized-battery energy harvesting communication can be analyzed through a discrete, noiseless binary abstraction that is equivalent to a timing channel. More broadly, optimal transmission policies enforce energy causality and finite-storage constraints through tightest-string or directional water-filling constructions.

  • Capacity model: The unit-battery model uses binary energy arrivals and a noiseless binary physical layer to make finite-battery capacity analytically tractable.The abstraction assumes energy arrivals are multiples of a fixed quantity and restricts the physical layer to a discrete alphabet.
  • Capacity model: The binary channel is equivalently modeled as a timing channel with additive geometric service-time noise known causally at the transmitter.Information is represented by waiting times between transmitted 1s rather than their absolute positions.
  • Offline scheduling: The tightest-string policy selects a piecewise-linear energy-consumption profile within the feasible energy tunnel.Concavity makes the optimal policy constant between energy harvests, reducing optimization to a finite number of epochs.
  • Offline scheduling: Directional water-filling equalizes energy across epochs while right-permeable taps enforce energy causality and limit transfers to Emax.The algorithm is derived from KKT optimality conditions and prevents battery overflows.
  • Offline scheduling: In a five-epoch example, energy equalizes where feasible, while causality and finite capacity block leftward or excessive rightward energy flow.Fading-channel variants additionally account for channel-state strengths.

B. Multi-User Channels and Practical Considerations

Multi-user and practical energy-harvesting systems extend single-user power management to relays, processing costs, and imperfect storage. These extensions produce separable relay policies, bursty transmission under processing costs, and modified allocation rules for inefficient storage.

  • Multi-user channels: In energy-harvesting broadcast channels, optimal total power management matches the single-user policy, then allocates power using a cut-off structure.Only power above the cut-off level is assigned to the weaker user.
  • Multi-user channels: For two-hop relay channels, an optimal policy can be separable: the source and relay optimize throughput using their own energy and data profiles.The overall optimal policy is generally non-unique.
  • Practical considerations: Processing costs yield a directional glue-pouring solution with a threshold power p* and bursty transmission schedules.Transmission duration cannot be arbitrarily long because circuitry consumes energy whenever transmit power is non-zero.
  • Practical considerations: Energy leakage and charging or discharging inefficiency modify feasible energy allocation and can favor immediate use of harvested energy.An inefficient battery combined with a limited efficient super-capacitor can be optimized through multi-stage directional water-filling.

IV. ONLINE ENERGY MANAGEMENT FOR GENERAL REWARD MAXIMIZATION

Online energy management makes transmission decisions from causal system-state information, often using Markov decision processes. The resulting policies include state-dependent packet thresholds, numerical policy iteration, and performance dependence on harvesting dynamics as well as battery size.

  • Online decision-making: Online devices decide whether to transmit using causal knowledge of battery energy and, potentially, the harvesting-process state.Markov decision processes provide a framework for these state-dependent decisions.
  • Online decision-making: The battery state evolves by subtracting action-dependent energy use, adding harvested energy, and capping storage at Emax.This model explicitly represents finite storage and battery-state transitions.
  • Reward maximization: For binary packet-transmission decisions with random importance values, a threshold policy transmits packets whose importance exceeds a state-dependent threshold.The threshold depends on the system state, including available energy.
  • Reward maximization: Policy Iteration Algorithm numerically finds the optimal strategy, while the balanced policy is asymptotically optimal as Emax approaches infinity and performs well for modest Emax.The balanced policy sets the threshold so average consumed energy matches expected replenishment.
  • Correlated harvesting: With correlated harvesting, performance depends on the ratio between Emax and harvesting-process dynamics rather than battery size alone.Slow harvesting processes may require a larger battery to withstand prolonged low-harvest periods.

B. Optimal Transmission Policies with Imperfect Knowledge of the State-of-Charge

Imperfect battery state knowledge changes energy-harvesting transmission control from a fully observable decision process to a partially observable one. Battery degradation and sensing costs further require policies to account for time-varying storage and energy-aware activation decisions.

  • Battery-state uncertainty can reach 30%, while obtaining more accurate state-of-charge information consumes time and energy.These constraints motivate transmission-policy optimization under uncertain battery status.
  • With imperfect state observation, the battery remains a Markov process, but the resulting control problem is modeled as a Partially Observable MDP.The optimal policy requires the full history rather than only the current observed state.
  • Battery degradation is modeled by adding memory that tracks time-varying battery parameters within an MDP framework.A probabilistic multistage formulation limits state-space growth while characterizing lifetime energy delivery.
  • Sensing policies decide whether to activate a sensor or remain idle, balancing energy consumption against the risk of missing important events.For correlated events, activation uses event prediction together with an available-energy threshold.

F. Optimal Random Multi-Access for a Network of Energy Harvesting Devices

Random multi-access in energy-harvesting networks must jointly manage collisions, intermittent energy, and unknown device backlogs. The reviewed protocols frame access policies through reward optimization and delivery–time-efficiency trade-offs, with TDMA providing an energy-limited delivery benchmark and DFA adapting frames to estimated backlogs.

  • Random multi-access: Uncoordinated multi-user access is more realistic but incurs collision errors and inefficiencies.Unlike the two-user upper-bound setting, no coordination is available among devices.
  • Random multi-access: Each user policy maps battery energy to a transmission probability or an importance threshold, and system reward sums successful users’ importance values.The reward is obtained when a user transmits while all others remain silent.
  • Random multi-access: The multi-user optimization is non-convex, so the analysis seeks symmetric Nash equilibria that are local optima rather than guaranteed global optima.Numerical investigation finds the resulting solution very good except under battery-capacity conditions described later in the paper.
  • MAC protocols: MAC performance is evaluated through time efficiency and delivery efficiency, which respectively measure successful packets per allocated slot and the fraction of devices reporting within an inventory round.Contention-based protocols trade higher delivery for lower time efficiency when more slots reduce collisions and wasted transmission energy.
  • MAC protocols: TDMA assigns every device an exclusive slot, making delivery efficiency limited only by device energy and potentially wasting slots when packets or energy are scarce.Thus, TDMA gives an upper bound on delivery efficiency for the considered ALOHA-based protocols but may be time-inefficient.
  • MAC protocols: DFA groups devices that still have data, previously collided, and retain transmission energy into successive frames sized from estimated backlog.This structure lets access adapt across frames within an inventory round.
  • MAC protocols: Increasing DFA or FA parameter ρ raises delivery efficiency but lowers time efficiency, whereas TDMA occupies a single point in the trade-off plane.The comparison is shown for different harvesting rates µH.

VI. JOINT WIRELESS ENERGY AND INFORMATION TRANSFER

The paper reviews wireless systems that transfer energy and information between nodes, covering energy cooperation and simultaneous transfer. Information-energy tradeoffs are characterized through capacity formulations, coupled circuits, and extensions to broader channel models.

  • Energy and information can be transferred simultaneously or through separate technologies, enabling energy sharing among self-sustaining wireless nodes.
  • The information-energy transfer problem maximizes mutual information subject to a receiver energy-delivery constraint B, yielding capacity C(B) and capacity-energy tradeoffs.
  • For finite alphabets, assigning each input symbol an energy-delivery cost reduces information-energy transfer to a classical cost-constrained capacity problem.
  • Inductively coupled circuits model energy and information transfer as communication over a frequency-faded channel with distinct optimal strategies for information and power transfer.Water-filling is optimal for information transfer, while a single resonant-frequency tone is optimal for power transfer.
  • The optimal information-power tradeoff outperforms every strategy that time-shares between water-filling and a resonant-frequency tone.
  • The literature extends simultaneous transfer to MIMO broadcast, fading, MIMO interference, and practical receiver implementations.

B. Coding for Energy and Information Transfer

The paper examines constrained code design and energy cooperation for managing energy timing, battery limitations, and network throughput. Structured codes match receiver utilization patterns, while coordinated energy and data transfer improves relay operation despite transfer inefficiency.

  • Coding for Energy and Information Transfer: Classical information-maximizing codes provide little control over energy-transfer timing, risking battery overflows or underflows.
  • Coding for Energy and Information Transfer: Constrained codes adjust energy-transfer properties to the receiver’s energy-utilization requirements.
  • Coding for Energy and Information Transfer: Type-0 (d, k)-RLL codes limit zero-run lengths and suit overflow-limited regimes, whereas type-1 codes target underflow-limited regimes.
  • Coding for Energy and Information Transfer: For periodic energy requests, small-k type-0 RLL codes substantially improve energy-transfer efficiency at low information rates; higher rates require larger k and small d.
  • Energy Cooperation: In relay networks, energy cooperation combines source-to-relay wireless energy transfer with signal-level forwarding to optimize overall operation.The relay receives αδ_i from a source transfer δ_i, with α < 1 representing transfer inefficiency.
  • Energy Cooperation: Energy cooperation extends to two-way and multiple-access channels, including directional water-filling and separate optimization of wireless transfer and temporal power allocation.

D. Interactive Exchange of Energy and Information

The paper studies interactive energy and information exchange in two-way systems and energy-aware sensing networks. Adaptive codebooks and joint source-communication resource management address energy-state dependence, queue stability, and reconstruction quality.

  • Interactive Exchange of Energy and Information: In a two-way system with energy reuse, nodes exchange information while harvesting received energy under a shared energy-state model.
  • Interactive Exchange of Energy and Information: Codebooks are selected according to the current distribution of energy units, with higher-energy nodes using codebooks containing more “1” symbols to facilitate transfer.
  • Interactive Exchange of Energy and Information: Adaptive codebooks optimize symbol probabilities by energy state, whereas non-adaptive codebooks use the same zero-one fractions across states.
  • Energy Harvesting Wireless Sensor Networks: Sensor networks must allocate energy between source acquisition and radio transmission because sensing, sampling, and compression can have energy costs comparable to transmission.
  • Energy Harvesting Wireless Sensor Networks: The energy management unit jointly chooses compression distortion and energy allocations using measurement SNR, channel SNR, and data-queue state.
  • Energy Harvesting Wireless Sensor Networks: For a desired average distortion, optimal policies stabilize the data queue while satisfying the distortion constraint whenever possible.
  • Energy Harvesting Wireless Sensor Networks: The optimal policy’s reconstruction distortion is compared with greedy and non-adaptive alternatives as harvesting variability changes.

VIII. LARGE-SCALE WIRELESS NETWORKS WITH ENERGY HARVESTING

The paper analyzes how energy harvesting affects large-scale wireless networks through network architecture, node density, interference, and energy arrivals. In MANETs, energy availability controls transmitter activity, creating a throughput-power tradeoff under outage constraints.

  • Large-scale energy-harvesting network models relate mobile ad hoc and cellular performance to energy arrivals, node distributions, architectures, and realistic interference.
  • In the MANET model, transmitters form a homogeneous Poisson point process, communicate with unit-distance receivers, and require SINR above θ for reliable decoding.
  • The steady-state transmitter activation probability is ρ = min(1, λ_e/P), where λ_e is the energy-arrival rate and P is transmission power.
  • Transmission power must balance interference, transmitter sparsity, and decoding reliability while satisfying the outage constraint and maximizing throughput.

B. Cellular Networks with Energy Harvesting

The section reviews cellular and heterogeneous networks powered by harvested energy, including spatial energy-field models, aggregators, and stochastic base-station operation. It also emphasizes that circuit and wiring energy materially affect total-energy optimization.

  • Cellular Networks with Energy Harvesting: A Boolean random-function model represents spatially varying renewable energy intensity by combining Poisson-distributed energy centers with exponentially decaying influence.The model is used to study coverage in cellular networks with renewable-powered base stations.
  • Cellular Networks with Energy Harvesting: Outage probability decreases exponentially with the product λeν in noise-limited cellular networks with on-site harvesters.The result applies under a downlink outage constraint and a spatial energy-field model.
  • Cellular Networks with Energy Harvesting: Aggregators spatially average harvested energy, stabilizing base-station transmission power as the number of connected harvesters increases.The transmission power converges to a constant by the law of large numbers, reducing outage probability.
  • Cellular Networks with Energy Harvesting: Heterogeneous cellular networks can model renewable-powered base stations as independent PPP tiers, yielding independent Bernoulli on/off states when energy arrivals are independent.This framework is used to study the impact of base-station operation on network coverage performance.
  • Energy Consumption Models: Wiring energy can diverge to infinity faster than node energy, making total-energy-optimal communication operate increasingly far from channel capacity as error probability approaches zero.The result challenges strategies that keep rate close to capacity when circuit energy is included.
  • Energy Consumption Models: Ignoring circuit energy can produce optimistic strategies that significantly underestimate practical total-energy requirements.The relevant energy components include both computational-node and wiring energy.

X. CONCLUSION AND FORWARD LOOK

The paper surveys advances across energy harvesting wireless communications, from physical-layer limits and scheduling to medium access, networking, and energy transfer. It concludes that practical models of storage, consumption, processing, and causal harvesting remain central to deriving useful design insights.

  • Conclusion and Forward Look: The article summarizes recent advances in energy harvesting wireless communication networks.Its scope spans theoretical limits, protocols, energy transfer, and cooperation.
  • Conclusion and Forward Look: The review covers information-theoretic limits, scheduling, medium access control, energy transfer and cooperation, and networks ranging from single-hop to large-scale systems.It includes simultaneous information and energy transfer as an emerging paradigm.
  • Conclusion and Forward Look: The field combines theoretical challenges with physical and practical concerns, including storage imperfections, consumption models, processing costs, and causal harvesting profiles.These considerations support mathematical formulations aimed at obtaining design insights.
Loading 1501.06026v1…