Source-linked AI summary
Dynamic Spectrum Access in Cognitive Radio Networks with RF Energy Harvesting
Xiao Lu, Ping Wang, Niyato Dusit, Hossain Ekram
TL;DR
RF-powered CRNs seek to combine spectrum-efficient dynamic access with RF energy harvesting, but their dynamic access problem has not been rigorously studied. This article surveys the network setting and challenges, then formulates multi-channel channel selection as an MDP; numerical results examine throughput under sensing and harvesting tradeoffs.
Problem
Dynamic spectrum access for RF-powered CRNs has not been rigorously studied, despite the need to coordinate idle-channel transmission with occupied-channel energy harvesting.
Method
The article overviews RF-powered CRNs and formulates multi-channel channel selection as an optimization problem based on a Markov decision process.
Results
The MDP formulation obtains an optimal channel-selection policy that maximizes secondary-user throughput, with numerical results showing sensitivity to sensing decisions and RF energy availability.
Takeaways & Limitations
Effective channel selection must jointly consider spectrum sensing, data transmission, and RF energy harvesting rather than transmission opportunities alone.
Abstract
from arXiv · showhide
Spectrum efficiency and energy efficiency are two critical issues in designing wireless networks. Through dynamic spectrum access, cognitive radios can improve the spectrum efficiency and capacity of wireless networks. On the other hand, radio frequency (RF) energy harvesting has emerged as a promising technique to supply energy to wireless networks and thereby increase their energy efficiency. Therefore, to achieve both spectrum and energy efficiencies, the secondary users in a cognitive radio network (CRN) can be equipped with the RF energy harvesting capability and such a network can be referred to as an RF-powered cognitive radio network. In this article, we provide an overview of the RF-powered CRNs and discuss the challenges that arise for dynamic spectrum access in these networks. Focusing on the tradeoff among spectrum sensing, data transmission, and RF energy harvesting, then we discuss the dynamic channel selection problem in a multi-channel RF-powered CRN. In the RF-powered CRN, a secondary user can adaptively select a channel to transmit data when the channel is not occupied by any primary user. Alternatively, the secondary user can harvest RF energy for data transmission if the channel is occupied. The optimal channel selection policy of the secondary user can be obtained by formulating a Markov decision process (MDP) problem. We present some numerical results obtained by solving this MDP problem.
1 School of Computer Engineering, Nanyang Technological University (NTU), Singapore
RF-powered CRNs combine dynamic spectrum access with RF energy harvesting to improve spectrum and energy efficiency. Their access protocols must jointly balance sensing, transmission, and harvested energy, motivating MDP-based channel selection.
- Motivation: RF energy harvesting offers a relatively predictable supply because it does not depend on nature, although harvested energy depends on wavelength and source distance.The harvested amount can be calculated using the Friis transmission equation.
- Motivation: RF-powered CRNs let secondary users harvest RF signals, convert them to DC electricity, store the energy, and use it for operation and data transmission.RF sources can include primary users, cellular base stations, and other ambient sources.
- Dynamic access challenge: Secondary users must identify idle spectrum for transmission and occupied spectrum for RF harvesting, unlike traditional CRNs focused primarily on spectrum holes.This requirement arises because RF-powered devices harvest energy from occupied bands while respecting primary-user interference constraints.
- Research focus: Dynamic sensing and channel access must be optimized across throughput, energy efficiency, and RF energy supply because traditional CRN protocols may be inefficient here.The article identifies dynamic spectrum access for RF-powered CRNs as insufficiently studied and focuses on this problem.
- Research focus: The paper studies multi-channel channel selection by mapping the secondary user’s data queue, energy storage, and channel status to sensing, transmission, or harvesting actions.The policy is formulated as an optimization problem based on a Markov decision process to maximize throughput.
A. RF Energy Harvesting in Cognitive Radio Device
An RF-powered cognitive radio device combines cognition-cycle functions with RF harvesting, energy storage, power management, and wireless communication components. Its harvester converts RF input into usable DC power through antenna, matching, rectification, and storage stages.
- Device components: The device includes a software-defined radio transceiver, spectrum analyzer, knowledge extraction unit, decision-making unit, node equipment, A/D converter, and power controller.These components support spectrum observation, learning, decision-making, applications, and network control.
- Device components: RF harvesting adds an energy storage device, power management unit, and RF harvester to the cognitive radio device.The storage may be a battery or capacitor, while power management decides whether to store or immediately dispatch harvested energy.
- Harvester design: The RF harvester consists of an antenna, impedance matching unit, voltage multiplier, and capacitor.The matching unit maximizes power transfer, the multiplier rectifies RF AC into DC, and the capacitor smooths delivery or temporarily stores energy.
- Harvester design: The harvester may operate at single or multiple frequencies, with multi-frequency acquisition potentially supplying enough RF energy as input.A multi-frequency design may require an antenna capable of operating at multiple frequencies simultaneously.
- Harvester design: A shared or separate wireless interface determines whether data communication and RF harvesting can occur simultaneously.Different interfaces can support concurrent transmission and harvesting on different frequencies, whereas a shared interface prevents simultaneous transmission and harvesting.
B. Architecture of RF-Powered Cognitive Radio Network
RF-powered CRNs use RF signals for both data transmission and energy transfer across primary and secondary network components. Their architecture may be infrastructure-based or infrastructure-less, with centralized or distributed spectrum access.
- Network architecture: Secondary users can harvest RF energy from primary base stations, primary users, secondary base stations, or other secondary users.The architecture therefore supports multiple possible RF energy sources.
- Network zones: In a primary base station’s transmission zone, a secondary user in the RF harvesting zone can harvest strong primary RF signals.A secondary user in an interference zone cannot transmit when primary transmissions occupy the spectrum.
- Network architecture: RF-powered CRNs can adopt infrastructure-based or infrastructure-less communication architectures, including centralized or distributed dynamic spectrum access.An infrastructure-based secondary base station coordinates communication among secondary users.
- Control architecture: Centralized control can use global radio-environment and RF-energy information, whereas autonomous distributed decisions may not achieve network-wise optimal control.The distributed limitation applies because individual secondary users independently decide spectrum access and energy harvesting.
- Design implications: Spectrum sensing, access, and handoff functionalities must be revisited to optimize RF-powered CRN performance.The paper identifies these issues as part of designing dynamic spectrum access for RF-powered CRNs.
III. RESEARCH CHALLENGES IN DYNAMIC SPECTRUM ACCESS IN RF-POWERED COGNITIVE RADIO NETWORKS
Dynamic access in RF-powered CRNs must identify channels for both data transmission and RF harvesting. The resulting channel-selection problem trades sensing accuracy and harvested energy against available transmission time and throughput.
- Spectrum sensing and access: RF-powered CRNs must sense occupied channels for energy harvesting as well as idle channels for data transmission, so traditional sensing and access methods may be insufficient.The paper frames these requirements as research challenges for spectrum sensing and access.
- Channel selection: Channel selection must account for channel occupancy, channel quality, RF-signal strength, and the secondary user’s energy level.A channel with high idle probability may still have low quality, while an energy-poor user may prefer an occupied channel with strong harvestable RF energy.
- Channel selection: Multi-channel RF-powered CRNs require policies specifically designed to balance harvested energy against communication throughput.Traditional channel-selection schemes are described as insufficient for the more complex decision problem.
- Spectrum sensing and access: Longer or more frequent sensing improves sensing accuracy and harvested energy but reduces data-transmission time and can harm throughput.This creates a tradeoff among sensing duration, sensing frequency, sensing accuracy, harvested energy, and throughput.
- Channel selection: Channel-selection policies may use proactive sensing of different channels or on-demand sensing when switching to a target channel.The sensing order and switching choice depend on primary-user activity and the secondary user’s remaining energy.
B. Spectrum Access
RF-powered CRNs require spectrum-access protocols to coordinate data transmission, RF energy harvesting, collision avoidance, and resource allocation. Both random and fixed spectrum access must account for users’ energy states and harvesting opportunities.
- Spectrum-access objectives: RF-powered CRNs retain throughput, primary-user protection, collision-avoidance, and fair-sharing objectives while adding energy-state constraints.These objectives motivate adopting fixed or random spectrum access protocols.
- Fixed spectrum access: Fixed access protocols statically allocate time slots and channels or subcarriers, requiring optimal resource allocation given RF-energy availability.Resources should be allocated to users according to whether they are harvesting energy or are outside transmitting sources’ range.
- Random spectrum access: Random access protocols must address collisions while coordinating backoff, energy harvesting, and data transmission among secondary users.Under high contention, some users can back off and harvest energy instead.
- Random spectrum access: RF-powered CRNs extend conventional spectrum access by requiring secondary users to choose between harvesting RF energy and transmitting data.This choice must account for remaining energy and the amount of RF energy available for harvesting.
- Random spectrum access: Secondary users may switch to RF energy harvesting when a primary user reoccupies the selected channel.In RF-powered CRNs, channel switching can also be required when the user needs to harvest energy.
IV. CHANNEL SELECTION IN RF-POWERED COGNITIVE RADIO NETWORKS
The paper models channel selection for a secondary transmitter in an RF-powered CRN with multiple primary-user channels. The secondary user senses the selected channel, transmitting when idle and harvesting energy when busy, and the policy is optimized through an MDP.
- System model: The system contains N primary users and one RF-powered secondary user operating across N channels that can be idle or occupied.The secondary user has an RF energy harvester and storage with capacity E units.
- System model: The model includes packet arrivals, a data queue of capacity Q, transmission energy requirement W, and channel-specific transmission success probabilities σn.Harvesting success is characterized by γn for each selected channel.
- Channel selection: The channel-selection policy maps the secondary user’s state, consisting of data-queue packets and stored energy, to a channel-selection action.The optimal policy is obtained by formulating and solving an MDP.
- Channel selection: The secondary user selects one channel using statistical information about channel occupancy, transmission success, and energy-harvesting success.The user does not know the selected channel’s status before sensing.
- Channel selection: After selecting a channel, the secondary user senses its status, then transmits on an idle channel or harvests RF energy on a busy channel.A single wireless interface prevents simultaneous transmission and harvesting.
B. Optimization Formulation
The optimization formulation represents the RF-powered secondary user as an MDP whose state tracks stored energy and queued packets. It maximizes long-term average throughput through channel-selection decisions and state transitions.
- State and action: The action is the selected available channel, and channel status determines whether the system follows transmission or harvesting transitions.Idle-channel transitions depend on α and σn, whereas busy-channel transitions depend on α and γn.
- State transitions: State transitions combine packet arrivals with successful or unsuccessful transmission and RF energy harvesting outcomes.The model enumerates increases, decreases, or unchanged levels for both queue occupancy and stored energy.
- State transitions: A packet transmits successfully when stored energy is available, the data queue is nonempty, and the wireless transmission has no error.Queue and energy levels cannot exceed their respective capacities or decrease below empty states.
- Optimization objective: The MDP maximizes the secondary user’s long-term average throughput using an optimal channel-selection policy π⋆.If stored energy is insufficient, the user cannot transmit and is forced to harvest RF energy; standard solution methods include linear programming, value or policy iteration, and Q-learning.
- State and action: The MDP state is θ = (e, q), where e is stored energy and q is the number of packets in the data queue.Both state components are bounded by energy-storage and queue capacities.
C. Performance Evaluation
The evaluation considers a secondary user with finite energy and queue capacities, two licensed channels, and specified arrival, occupancy, transmission, and harvesting probabilities. It compares the proposed policy with a static channel-selection policy.
- Parameter Setting: The experiment uses two licensed channels, c1 and c2, with idle probabilities 0.1 and 0.9, respectively.The successful packet transmission probability is 0.95 on both channels.
- Parameter Setting: When occupied by primary users, successful one-unit RF-energy harvesting probabilities are 0.95 on c1 and 0.70 on c2.These channel-specific harvesting probabilities are part of the evaluation setting.
- Performance Evaluation: The evaluation compares the state-aware optimal policy with a static policy that selects channels without considering data-queue and energy-storage states.The study examines how channel selection for sensing affects secondary-user throughput.
2) Numerical Results:
The numerical results show that throughput depends on balancing channel sensing, packet transmission, and RF energy harvesting. An MDP-based policy adapts channel selection to queue and energy states and outperforms a static policy.
- Numerical Results:: Throughput can be low when channel 1 is sensed too little or too much, revealing a sensing tradeoff.More sensing increases harvested RF energy but reduces transmission opportunities; less sensing has the opposite effect.
- Numerical Results:: Peak throughput is higher at a packet arrival rate of 0.5 packets/time slot than at 0.2 packets/time slot.At a low arrival rate, the secondary user does not need much energy to transmit packets.
- Numerical Results:: The optimal policy selects c1 when energy and queue levels are low, but c2 when both are high.c1 is more likely to support RF energy harvesting, whereas c2 has a higher chance of being idle for packet transmission.
- Numerical Results:: The policy favors c1 more than c2 because successful RF energy harvesting from c2 is less probable.Channel choice is determined by the secondary user's data queue, energy storage, and channel status.
- Numerical Results:: As channel c1 becomes less busy, optimal-policy throughput first increases because packet-transmission opportunities increase, then decreases when harvested energy becomes insufficient.When c1 is mostly idle, the secondary user cannot harvest enough RF energy to transmit packets.
- Numerical Results:: The MDP-based optimal policy achieves higher throughput than the static policy.The static policy adjusts channel-selection ratios without considering the data queue or energy-storage level.
V. CONCLUSION
The article formulates RF-powered CRN channel selection as an optimization problem based on an MDP, seeking a throughput-maximizing policy under incomplete information and RF harvesting constraints. It also identifies energy availability as a key determinant of whether frequently idle channels are beneficial and discusses extensions involving dedicated RF energy sources and energy trading.
- The channel selection problem considers multiple primary users and a secondary user with RF energy-harvesting capability under incomplete information.
- An MDP-based optimization formulation obtains the optimal channel selection policy that maximizes the secondary user's throughput.
- Frequent channel idleness is not always beneficial because the secondary user may fail to harvest enough RF energy for its own data transmission.
- Insufficient harvested RF energy can result in reduced throughput performance.
- The channel selection problem can be extended to multi-channel RF-powered CRNs and to settings with dedicated RF energy sources and energy trading.