Source-linked AI summary
Energy Harvesting-Aided Spectrum Sensing and Data Transmission in Heterogeneous Cognitive Radio Sensor Network
Deyu Zhang, Zhigang Chen Ju Ren, Ning Zhang, Mohamad Khattar Awad, Haibo Zhou, Xuemin, Shen
TL;DR
HCRSNs must jointly manage spectrum sensing and sensor energy under inaccurate sensing, EH constraints, and battery limitations. The paper develops tandem scheduling and resource-allocation algorithms, reporting near-optimal spectrum scheduling and lower data-sensor energy than random assignment while sustaining EH sensors.
Problem
HCRSN resource allocation must address spectrum under-utilization, energy inefficiency, and spectrum-sensing inaccuracy across EH-enabled spectrum sensors and battery-powered data sensors.
Method
The paper combines Cross-Entropy-based spectrum sensor scheduling with joint time, channel, and power allocation for data sensors.
Results
JTPA consumes 5% to 14% more energy than the optimal scheme but conserves 18% to 31% more energy than random assignment.
Takeaways & Limitations
The proposed solution adapts channel use to harvested energy while allocating scarce battery-powered data-sensor resources and protecting primary-user transmissions.
Abstract
from arXiv · showhide
The incorporation of Cognitive Radio (CR) and Energy Harvesting (EH) capabilities in wireless sensor networks enables spectrum and energy efficient heterogeneous cognitive radio sensor networks (HCRSNs). The new networking paradigm of HCRSNs consists of EH-enabled spectrum sensors and battery powered data sensors. Spectrum sensors can cooperatively scan the licensed spectrum for available channels, while data sensors monitor an area of interest and transmit sensed data to the sink over those channels. In this work, we propose a resource allocation solution for the HCRSN to achieve the sustainability of spectrum sensors and conserve energy of data sensors. The proposed solution is achieved by two algorithms that operate in tandem, a spectrum sensor scheduling algorithm and a data sensor resource allocation algorithm. The spectrum sensor scheduling algorithm allocates channels to spectrum sensors such that the average detected available time for the channels is maximized, while the EH dynamics are considered and PU transmissions are protected. The data sensor resource allocation algorithm allocates the transmission time, power and channels such that the energy consumption of the data sensors is minimized. Extensive simulation results demonstrate that the energy consumption of the data sensors can be significantly reduced while maintaining the sustainability of the spectrum sensors.
I. INTRODUCTION
HCRSNs combine cognitive radio, cooperative sensing, and energy harvesting to address interference, spectrum under-utilization, sensing errors, and sensor energy constraints. The paper proposes coordinated spectrum-sensor scheduling and data-sensor resource allocation for sustainable, energy-efficient operation.
- Cognitive radio lets sensor networks opportunistically access underutilized licensed spectrum, reducing interference and improving spectrum utilization.
- Frequent spectrum scanning improves availability estimates and PU protection but increases energy consumption in battery-constrained networks.
- Cooperative spectrum sensing improves accuracy because single-sensor decisions are vulnerable to shadowing and multipath fading.
- The HCRSN uses EH-enabled spectrum sensors for cooperative sensing and battery-powered data sensors for transmitting collected data to the sink.
- The proposed unified solution maximizes detected channel available time while respecting EH dynamics and minimizes data-sensor energy through joint channel, time, and power allocation.
- Each frame contains spectrum sensing and data transmission phases, with sensing divided into mini-slots for individual sensor-channel sensing.
B. Cognitive Radio Model
The cognitive radio model detects primary-user activity using cooperative energy detection over fading channels. Sensor decisions are combined at the sink under a Logic-OR rule, with false-alarm and detection probabilities characterizing sensing performance.
- Channels undergo slow, flat Rayleigh fading, while primary-user activity follows stationary exponential ON-OFF processes.
- Spectrum sensors use binary hypotheses in which H0 denotes an inactive, available channel and H1 denotes an active, unavailable channel.
- An energy detector compares the received-signal test statistic with threshold ε to decide whether the primary user is active.
- The false-alarm probability measures detecting a primary user when H0 is true, whereas the detection probability measures detecting one when H1 is true.
- Under the Logic-OR rule, the sink declares a primary user present if at least one scheduled sensor reports presence.
- The scheduled sensor set for channel k is represented by M_k, and each sensor sends a final 1-bit decision to the sink.
III. PROBLEM STATEMENT AND PROPOSED SOLUTIONS
The HCRSN separates spectrum-sensor scheduling from data-sensor resource allocation, while solving both problems in tandem. The first maximizes channel availability under sensing constraints, and the second minimizes battery-powered sensors’ energy consumption.
- The two problems are linked because spectrum-sensor scheduling determines channel availability for subsequent data-sensor allocation.
- The spectrum sensor scheduling problem maximizes channel availability while respecting energy-harvesting dynamics and primary-user access priorities.
- The data sensor resource allocation problem assigns available channels, transmission time, and power to minimize data sensors’ energy consumption.
- The first problem is formulated as nonlinear integer programming, whereas the second is formulated as biconvex optimization.
A. Spectrum Sensors Scheduling
Spectrum sensor scheduling assigns sensors to channels to maximize detected channel availability while satisfying primary-user protection, energy-harvesting, and sensing-time constraints. The resulting combinatorial problem is addressed with a Cross-Entropy-based solution.
- The SSS objective maximizes each channel’s average detected available time while protecting primary-user transmissions.Detected availability depends on actual channel availability and sensing-detection performance.
- The assignment matrix J records which spectrum sensors detect which channels using binary sensor-to-channel assignments.A value of 1 indicates that a sensor is assigned to detect a channel, while 0 indicates no assignment.
- The scheduling constraints require each spectrum sensor’s energy consumption not to exceed harvested energy within a frame.The sensing energy is determined by sensing duration and sensing power, while harvested energy depends on the sensor’s average EH rate.
- Sensing time for each channel is bounded by the spectrum-sensing phase duration.
- Exhaustive search is computationally prohibitive because the assignment space has size 2^MK, motivating the Cross-Entropy algorithm.The paper notes that the Cross-Entropy algorithm’s performance bound remains theoretically open.
2) Cross Entropy-based Algorithm:
The Cross-Entropy algorithm converts the constrained scheduling problem into a stochastic optimization procedure that samples assignments, retains high-performing solutions, and updates assignment probabilities until convergence. The selected best sample is mapped back to the sensor-to-channel assignment matrix.
- A penalty method transforms the constrained SSS problem into an unconstrained objective that penalizes infeasible assignments.A positive penalty makes infeasible solutions evaluate negatively when constraints are violated.
- The algorithm represents each sensor’s possible channel assignments as a binary assignment matrix and samples assignments from a probability-mass matrix.The probability matrix is updated iteratively during the Cross-Entropy procedure.
- Each iteration generates random samples, evaluates their objective values, and retains the largest fraction as elite solutions.The retained solutions determine the threshold used for subsequent probability updates.
- Assignment probabilities are increased for choices that produce large objective values across sampled solutions.
- The algorithm stops at the iteration limit or convergence condition, then selects the sampled assignment with the largest objective value.That selected sample is mapped to J and used to schedule spectrum sensors; the sink then estimates availability using the Logic-OR rule.
B. Transmission Time and Power Allocation in the Data Transmission Phase
During data transmission, the proposed approach allocates transmission resources for battery-powered data sensors and assigns detected channels to sink-mounted cognitive-radio transceivers. The objective is to reduce data-sensor energy consumption.
- The data-transmission phase reports collected data to the sink, making energy minimization important for battery-powered data-sensor lifetime.
- The transmission time and power allocation problem is formulated as biconvex optimization and addressed with the JTPA algorithm.
- Detected available channels are allocated to cognitive-radio transceivers mounted on the sink.If fewer channels are available than transceivers, all available channels are allocated; otherwise, channels with the largest sojourn times are selected.
1) Problem Formulation:
The data-sensor resource allocation problem minimizes total transmission energy while satisfying channel-access, duration, data-rate, and power constraints under PU collision protection.
- Because all channels share bandwidth and average power gain, longer average sojourn time implies greater capacity.
- The access time is chosen below the detected channel availability to maintain a low PU-collision probability under stationary exponential ON-OFF PU behavior.
- The formulation allocates transmission times and powers for each data sensor–channel pair, with total energy determined by these allocations.
- Each channel’s aggregate transmission time is limited by its maximum access time, while each sensor’s total time is limited by the data-transmission phase.
- Each sensor must transmit its required data amount using rates determined by channel gain, transmission time, and power.
- Maximum-power and nonnegativity constraints bound the transmission variables, and the resulting DSRA problem is non-convex but biconvex.
2) Joint Time and Power Allocation (JTPA) Algorithm:
The JTPA algorithm solves the biconvex data-sensor allocation problem by alternating convex time and power optimization, yielding a partially optimal solution rather than a guaranteed global optimum.
- JTPA divides the biconvex DSRA problem into separate convex time-allocation and power-allocation subproblems.
- Alternate Convex Search fixes one variable block while optimizing the other, then repeats this process until termination.
- The algorithm starts from an arbitrary time–power point and returns the iteratively updated allocation after convergence or reaching the iteration limit.
- JTPA uses the Simplex method for time allocation and the Interior Point method for power allocation.
- Convergence to the global optimum is not guaranteed, but convergence to a stationary partially optimal solution is guaranteed.
IV. PERFORMANCE EVALUATION
The study evaluates the C-E algorithm for spectrum sensing and JTPA for data transmission through Matlab simulations.
- The evaluation uses Matlab simulations to assess C-E in spectrum sensing and JTPA in data transmission.Simulations ran on a computer with an Intel Core i7-4510U CPU and 8 GB RAM.
A. Simulation Setup
The simulations examine C-E convergence, parameter sensitivity, optimality, and comparisons with greedy assignment under varied EH rates, sensing durations, and sensor counts.
- A. Simulation Setup: The simulated HCRSN contains M = 10 spectrum sensors and N = 30 data sensors, with seven licensed channels by default.
- B. Performance Evaluation of the C-E Algorithm: C-E achieves 87%−94% of exhaustive search’s DAATC and produces a solution 2 to 3 times larger than random assignment.
- B. Performance Evaluation of the C-E Algorithm: For EH rates from 3 mW to 7 mW, C-E converges to maximum DAATC after 30 iterations, with DAATC increasing as EH rate rises.
- B. Performance Evaluation of the C-E Algorithm: With EH rate fixed at 7 mW, DAATC increases as spectrum-sensing duration τs grows from 2 ms to 6 ms.
- B. Performance Evaluation of the C-E Algorithm: A smaller stopping tolerance ǫ can yield larger DAATC, while ρ = 0.6 maximizes DAATC for the parameters studied.
- B. Performance Evaluation of the C-E Algorithm: C-E outperforms the greedy algorithm in DAATC across EH rates and when the number of spectrum sensors varies at a fixed EH rate of 7 mW.
C. Performance Evaluation of the JTPA Algorithm
The evaluation examines JTPA’s energy consumption, convergence, and comparisons with random, optimal, and maximum-power schemes. Results indicate that JTPA substantially reduces data-sensor energy while supporting efficient channel and resource allocation.
- Comparison with benchmark schemes: JTPA consumes 5% to 14% more energy than the optimal scheme but 18% to 31% less than random assignment.The comparison uses three data sensors and three channels, with time and power optimized after channel assignment.
- Convergence: JTPA converges after 10 iterations, with energy consumption decreasing 97% during the first 6 iterations.The convergence experiment uses ten spectrum sensors, thirty data sensors, and five channels.
- Comparison with maximum-power allocation: The pmax scheme fixes data-sensor transmission power at the maximum available value and determines transmission time through a linear program.This scheme is comparable to channel allocation with fixed transmission power.
- Comparison with maximum-power allocation: JTPA consumes less energy than the pmax scheme across different data amounts and maximum-power values.The reported advantage is attributed to jointly allocating transmission time and power.
- Overall outcome: The proposed solution allocates channels to maximize detected available time and allocates transmission time and power to prolong data-sensor lifetime.The conclusion describes the solution as adapting to harvested-energy availability while optimizing battery-powered sensors’ scarce resources.
APPENDIX A DERIVATION OF THE COLLISION PROBABILITY pk
The appendix derives collision probability for a channel whose primary-user OFF period is exponentially distributed. It relates the probability of a primary user returning within a sensing or transmission interval to channel collision probability.
- OFF/Inactive-period model: The OFF/Inactive period is modeled with an exponential probability density function.The appendix introduces the sojourn time of an OFF/Inactive period and its density.
- Return probability: The probability that the OFF/Inactive period is less than ᾱ_k represents the primary user returning on channel k within [0, ᾱ_k].This return probability is used in the collision-probability derivation.
- Collision probability: The appendix expresses collision probability p_k^coll(ᾱ_k) using the channel-availability probability and the primary-user return event.The displayed derivation follows the exponential ON/OFF model.
- Optimization structure: The data-sensor resource-allocation problem is biconvex when either transmission-time variables T or power variables P are fixed.Fixing one variable set makes the remaining problem convex and solvable through the stated optimization subproblems.