VLC-NOMA system power distribution method based on improved particle swarm optimization

Abstract

The invention relates to a VLC-NOMA system power distribution method based on improved particle swarm optimization. The method comprises the following steps: constructing a downlink model based on a VLC-NOMA system; constructing a power distribution optimization model with maximization of the total rate of the VLC-NOMA system as a target; and constructing an IPSO algorithm, wherein the IPSO algorithm comprises an adaptive dynamic inertia weight, a collaborative learning factor and an enhanced elite reverse learning mechanism. Through the introduced adaptive dynamic inertia weight, collaborative learning factor and enhanced elite reverse learning mechanism, the adaptive dynamic inertia weight can allocate different powers according to the difference of users, and the collaborative learning factor can dynamically adapt to the collective behavior of the group. And an enhanced elite reverse learning mechanism selects a better fitness through comparison of an elite reverse solution and a quasi-reverse solution, and these improvements jointly enhance the capabilities of the scheme in the aspects of global exploration and local development, and improve the convergence rate of the algorithm and the robustness in a complex interference environment.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a power allocation method for VLC-NOMA systems based on improved particle swarm optimization. Background Technology

[0002] In the information age, the security risks of wireless communication technology are becoming increasingly prominent. Traditional radio frequency signals have inherent defects such as strong electromagnetic wave penetration and easy interception and cracking. Especially in some special fields, communication confidentiality faces severe challenges. Visible Light Communication (VLC) has emerged as a solution to this problem. With its rich spectrum resources, strong anti-electromagnetic interference ability, and high security, it has become an indispensable key supplementary technology in indoor communication scenarios.

[0003] VLC systems still face many technical challenges in practical applications. For example, the limited modulation bandwidth of light-emitting diode (LED) light sources severely restricts the system channel capacity, making it difficult to meet the current service requirements of high-power communication. On the other hand, the transmission performance of VLC systems is easily affected by multiple factors such as changes in the lighting environment, photodetector position offset, and multipath effects, resulting in insufficient system stability and greatly limiting its large-scale application.

[0004] To overcome the capacity bottleneck of VLC systems, researchers introduced Non-Orthogonal Multiple Access (NOMA) technology. This technology achieves efficient allocation of user resources through power domain multiplexing. The core logic lies in configuring differentiated power for different users: based on real-time channel state information, the base station allocates lower power to users with good channel conditions and higher power to users with poor channel conditions. Then, with the help of serial interference cancellation technology, the accurate separation of multi-user information is completed at the receiving end, thereby improving resource utilization and system capacity.

[0005] However, VLC-NOMA power allocation is essentially a non-convex, multi-constraint, high-dimensional coupled optimization problem. Traditional solution approaches rely on convex relaxation and approximate transformations, making it difficult to obtain a truly global optimum. While swarm intelligence algorithms can directly handle non-convex problems, they suffer from inherent issues such as rapid degradation of population diversity, susceptibility to local optima, and late-stage convergence stagnation. Furthermore, their stability and convergence accuracy are unsatisfactory in complex electromagnetic environments or multi-user scenarios, resulting in poor algorithm robustness.

[0006] Therefore, it is necessary to provide a new technical solution to improve one or more of the problems existing in the above solutions. Summary of the Invention

[0007] The purpose of this disclosure is to provide a power allocation method for VLC-NOMA systems based on improved particle swarm optimization. By introducing adaptive dynamic inertia weights, collaborative learning factors, and enhanced elite back-learning mechanisms, these improvements collectively enhance the ability of this scheme in global exploration and local exploitation, thereby improving the algorithm's convergence speed and robustness in complex disturbance environments.

[0008] A power allocation method for a VLC-NOMA system based on improved particle swarm optimization, provided by the present invention, includes the following steps: A downlink model based on a VLC-NOMA system is constructed, in which the user channel gain is calculated using the Lambertian model. A power allocation optimization model is constructed with the objective of maximizing the total rate of the VLC-NOMA system. The total power of the system is constrained based on this power allocation optimization model, which is defined as follows: ; in, The total system rate is the sum of the achievable rates for all successfully demodulated users. This represents the maximum total system speed. This represents the rate that user k can achieve after considering the interference cancellation factor, where K represents the total number of users; The power allocation optimization model is solved using the IPSO algorithm to obtain the optimal power allocation scheme. The IPSO algorithm includes an adaptive dynamic inertia weight mechanism, a collaborative learning factor mechanism, and an enhanced elite reverse learning mechanism. The adaptive dynamic inertia weighting mechanism introduces a dynamic power function and random perturbation to adapt to the complex, time-varying and noisy weight changes in the simulation. The collaborative learning factor mechanism introduces a set of co-evolutionary learning factors. and ,pass and This enables learning factors to dynamically adapt to the collective behavior of the group; The enhanced elite reverse learning mechanism introduces a quasi-reverse point, which selects the one with better fitness by comparing the elite reverse solution and the quasi-reverse solution.

[0009] Preferably, the downlink model based on the VLC-NOMA system includes an LED transmitter and K randomly distributed users, wherein the LED transmitter is located at the center of the ceiling and is allocated different powers according to the differences in user channel gain. At the receiving end, each user is equipped with a single photodetector to receive the mixed optical signal from the LED transmitter. The channel gain between the LED transmitter and the user is calculated using the Lambertian model, with the following formula: ; Where m is the order of the Lambert radiation. Indicates user The receiving area of ​​the photodetector. , and Representing users respectively The incident angle, the field of view of the photodetector, and the emission angle of the LED emitter. Indicates LED transmitter and user The distance between them This indicates the directional attenuation term of the LED emitter. Indicates user The gain of the optical filter is usually taken as ; This represents the cosine projection attenuation factor at the receiving end. Indicates user The gain of the optical concentrator.

[0010] Preferably, the adaptive dynamic inertia weighting mechanism is defined as follows: ; in, and These are the maximum and minimum values ​​of the inertia weight, respectively. This represents the current iteration number. The maximum number of iterations, An adaptive exponent for adjusting the decay rate based on the relative improvement value of the globally optimal fitness. The standard deviation of the current population fitness value. It is a positive integer.

[0011] Preferably, and This is a set of co-evolutionary learning factors, and the mechanism of the co-learning factors is defined as follows: ; ; in, , , , These represent the maximum and minimum values ​​of the learning factor, respectively, where t is the iteration time, and T is the maximum and minimum values. max Indicates the total number of times. and This is an adaptive exponent based on the relative difference between the average fitness and the global optimal fitness. Indicates the maximum speed. This represents the running speed of the i-th unit at time t. Indicates population spatial diversity, The peak value representing the spatial diversity of a population. This represents the average fitness of the population at time t. This represents the global optimal fitness of the population at time t.

[0012] Preferably, the enhanced elite reverse learning mechanism includes: Select the strain with the highest fitness from the current population. Individuals form an elite particle swarm ,for Each elite particle in This generates two different candidate solutions: the elite inverse solution and the quasi-inverse solution. The elite reverse solution is generated using a standard elite reverse learning strategy and is defined as follows: ; in, Indicates the first A dynamic interval of dimensions, which is adaptively updated with each generation based on the current population distribution; The quasi-inverse solution is generated between the center of the dynamic interval and the elite inverse solution, and is defined in each dimension as follows: ; in, Indicates the first The center of the dynamic interval; Reverse solution from generated elites Quasi-inverse solution Select the one with better fitness from the original elite particles Compare and select the one with better fitness to replace the one in the population. .

[0013] Preferably, the power allocation optimization model satisfies the following constraints: human eye safety regulations and total power constraints. The aforementioned eye safety regulations are defined as follows: ; in, This represents the DC bias used to ensure that the transmitted optical signal is non-negative. This represents the peak luminous intensity of the LED emitter. The total power constraint is defined as follows: ; in, This represents the power allocated to the k-th user. This indicates the maximum total transmit power allowed by the system.

[0014] Preferably, the power allocation optimization model further includes: non-negative power constraints and user service quality assurance constraints. The non-negative power constraint is defined as follows: ; in, This represents the transmit power of the k-th user.

[0015] The user service quality assurance constraint is defined as follows: ; in, This represents the actual signal-to-interference-plus-noise ratio (SIR) for the k-th user. This represents the minimum signal-to-interference-plus-noise ratio (SINNR) threshold required to ensure the quality of service for users.

[0016] Preferably, The definition of is: ; in, The relative improvement is calculated as follows: ((Optimized performance metrics - Unoptimized performance metrics) / Unoptimized performance metrics) × 100%.

[0017] Preferably, The definition of is: ; in, This represents the refractive index of the light concentrator.

[0018] Preferably, The order of the Lambert radiation is expressed as follows: ; in, This indicates the half-power angle of the LED emitter.

[0019] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects: By introducing adaptive dynamic inertia weights, collaborative learning factors, and an enhanced elite back-learning mechanism, the adaptive dynamic inertia weights can allocate different powers according to the differences between users, the collaborative learning factors can dynamically adapt to the collective behavior of the group, and the enhanced elite back-learning mechanism selects the one with better fitness by comparing the elite back-learning solution and the quasi-back-learning solution. These improvements together enhance the ability of this scheme in global exploration and local development, improve the algorithm's convergence speed and robustness in complex interference environments. Attached Figure Description

[0020] Figure 1A flowchart illustrating a power allocation method for a VLC-NOMA system based on improved particle swarm optimization in an exemplary embodiment of this disclosure is shown. Figure 2 This diagram illustrates a downlink model of a VLC-NOMA-based system in an exemplary embodiment of this disclosure. Figure 3 This diagram shows the performance curves of the total system power variation under different signal-to-interference-plus-noise ratios in an exemplary embodiment of this disclosure. Figure 4 A comparison graph showing the change of total system power with the number of users when the interference cancellation factor η=0 in an exemplary embodiment of this disclosure is shown; Figure 5 A comparison graph showing the total system power as a function of the number of users with an interference cancellation factor η=0.01 in an exemplary embodiment of this disclosure is shown. Figure 6 A comparison graph showing the total system power as a function of the number of users with an interference cancellation factor η=0.02 in an exemplary embodiment of this disclosure is shown. Figure 7 A comparison graph showing the total system power as a function of the number of users with an interference cancellation factor η=0.03 in an exemplary embodiment of this disclosure is shown. Figure 8 A comparison graph showing the change of total system power with iteration number when the interference cancellation factor η=0 in an exemplary embodiment of this disclosure is shown; Figure 9 A comparison graph showing the change of total system power with the number of iterations when the interference cancellation factor η=0.01 is shown in the exemplary embodiments of this disclosure; Figure 10 A comparison graph showing the change of total system power with the number of iterations when the interference cancellation factor η=0.02 is shown in the exemplary embodiments of this disclosure; Figure 11 A comparison graph is shown showing the change of total system power with the number of iterations when the interference cancellation factor η=0.03 is used in an exemplary embodiment of this disclosure. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0022] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0023] This example implementation first provides a power allocation method for a VLC-NOMA system based on improved particle swarm optimization, such as... Figure 1 As shown, it includes the following steps: S1: Construct a downlink model based on the VLC-NOMA system. The Lambertian model is used to calculate the user channel gain in the downlink model.

[0024] S2: Construct a power allocation optimization model with the objective of maximizing the total rate of the VLC-NOMA system. Constrain the total system power based on the power allocation optimization model. The power allocation optimization model is defined as follows: ; in, The total system rate is the sum of the achievable rates for all successfully demodulated users. This represents the maximum total system speed. This represents the rate that user k can achieve after considering the interference cancellation factor, where K represents the total number of users.

[0025] S3: Construct the IPSO algorithm and solve the power allocation optimization model based on the IPSO algorithm to obtain the optimal power allocation scheme. The IPSO algorithm includes an adaptive dynamic inertia weight mechanism, a collaborative learning factor mechanism, and an enhanced elite back-learning mechanism.

[0026] S31: The adaptive dynamic inertial weighting mechanism introduces dynamic power functions and random perturbations to adapt to complex, time-varying, and noisy weight changes in simulation.

[0027] S32: The collaborative learning factor mechanism introduces a set of collaborative evolutionary learning factors. and ,pass and This enables learning factors to dynamically adapt to the collective behavior of the group.

[0028] S33: The enhanced elite reverse learning mechanism introduces quasi-reverse points, and selects the one with better fitness by comparing the elite reverse solution and the quasi-reverse solution.

[0029] By introducing adaptive dynamic inertia weights, collaborative learning factors, and an enhanced elite back-learning mechanism, the adaptive dynamic inertia weights can allocate different powers according to the differences between users, the collaborative learning factors can dynamically adapt to the collective behavior of the group, and the enhanced elite back-learning mechanism selects the one with better fitness by comparing the elite back-learning solution and the quasi-back-learning solution. These improvements together enhance the ability of this scheme in global exploration and local development, improve the algorithm's convergence speed and robustness in complex interference environments.

[0030] The method described above in this example implementation will now be explained in more detail.

[0031] In S1, the main architecture of the downlink model is as follows: the LED transmitter is located at the center of the room ceiling, and K users are randomly distributed and communicate via line-of-sight links. The LED transmitter acts as the LED transmitter, and the users are equipped with photodetectors to receive the mixed light signals.

[0032] Channel gain model: The user channel gain is calculated using the Lambert model, taking into account factors such as receiving area, incident angle, field of view (FOV), distance, optical filter gain, and optical concentrator gain. The channel gain directly affects the power allocation strategy.

[0033] Power allocation mechanism: Different powers are allocated according to the differences in user channel gain. Users with poor channel conditions receive more power to overcome interference. This mechanism improves spectrum efficiency through power domain multiplexing.

[0034] Specifically, such as Figure 2 As shown, a VLC-NOMA-based downlink model is constructed. In this scenario, there is one LED transmitter and K randomly distributed users, where the LED transmitter is located at the center of the room ceiling, and different power is assigned to each user based on their channel gain differences.

[0035] Given the importance of line-of-sight links in indoor visible light communication, the channel gain between the LED transmitter and the user is calculated using the Lambertian model, with the following formula: ; Where m is the order of the Lambert radiation. Indicates user The receiving area of ​​the photodetector. , and Representing users respectively The incident angle, the field of view of the photodetector, and the emission angle of the LED emitter. Indicates LED transmitter and user The distance between them. This represents the directional attenuation term of the LED emitter, indicating the change in luminous intensity with emission angle. The varying radiation mode attenuation term, Indicates user The gain of the optical filter is usually taken as ; This represents the cosine projection attenuation factor at the receiving end. Indicates user The gain of the optical concentrator.

[0036] Furthermore, ; in, This represents the refractive index of the optical concentrator. In a VLC-NOMA system, the gain of the optical concentrator directly affects the signal strength at the receiver and the total system power. By optimizing the power allocation algorithm, the system's performance in complex interference environments can be improved.

[0037] Furthermore, the order m of the Lambert radiation is expressed as: ; in, This indicates the half-power angle of the LED emitter.

[0038] Receiver mechanism: The receiver decodes the signal of the strong channel user first and eliminates its interference to the weak channel user through step-by-step decoding, and then decodes the remaining signal. Each user is equipped with a single photodetector to receive the mixed optical signal, and SIC assists in decoding to obtain its own signal.

[0039] Specifically, at the receiving end, each user is equipped with a single photodetector to receive the mixed optical signal from the LED transmitter. The receiving device obtains its own signal through successful interference cancellation (SIC)-assisted decoding, while considering the non-ideal interference cancellation factor to represent the interference components remaining for users in weak channels.

[0040] In S2, to study the power allocation optimization problem of a VLC-NOMA system, the power allocation optimization model is constructed. It is defined as follows: ; in, The total system rate is the sum of the achievable rates for all successfully demodulated users. This represents the maximum total system speed. This represents the rate that user k can achieve after considering the interference cancellation factor, where K represents the total number of users.

[0041] For the physical feasibility of achieving the above objectives to be met, the power distribution of the system must adhere to the following constraints: 1) Eye safety regulations: Limit the total power consumption of the LED emitter to avoid hardware overload. The upper limit of the constraint is the minimum value of the bias current limit and the residual current limit.

[0042] ; in, This represents the DC bias used to ensure that the transmitted optical signal is non-negative. This is the peak luminous intensity of the LED emitter; this condition is used to limit the total power consumption of the LED emitter to avoid hardware limitations.

[0043] 2) Total power constraint: The sum of the transmit power of all users cannot exceed the upper limit of the total transmit power of the system.

[0044] ; in, This represents the transmit power of the k-th user. This indicates the maximum total transmit power allowed by the system.

[0045] 3) Non-negative power constraint: The total transmit power of the system is finite and the power allocated to each user must be non-negative.

[0046] ; in, This represents the transmit power of the k-th user.

[0047] 4) User Service Quality Assurance Constraints: In order to ensure the minimum service quality for all users and maintain the continuity of communication, the signal-to-interference-plus-noise ratio of users must not be lower than the preset minimum threshold (to ensure service quality).

[0048] ; in, This represents the actual signal-to-interference-plus-noise ratio (SIR) for the k-th user. This represents the minimum signal-to-interference-plus-noise ratio (SINNR) threshold required to ensure the quality of service for users.

[0049] In S3, to address the problems of traditional PSO (Particle Swarm Optimization) algorithm being prone to getting trapped in local optima when dealing with complex optimization problems, and having slow convergence speed and low accuracy in the later stages, an IPSO (Improved Particle Swarm Optimization) algorithm is proposed.

[0050] Compared to the standard PSO algorithm, three key modifications are introduced: an adaptive dynamic inertia weight mechanism, a collaborative learning factor mechanism, and an enhanced elite back-learning mechanism. These improvements collectively enhance the application's capabilities in global exploration and local exploitation, improving the algorithm's convergence speed and robustness in complex disturbance environments.

[0051] Furthermore, in S31, an adaptive dynamic inertia weight mechanism is introduced. This weight considers not only the iteration progress but also population diversity and convergence state. The proposed definition of inertia weight is as follows: ; in, and These are the maximum and minimum values ​​of the inertia weight, respectively. This represents the current iteration number. This represents the maximum number of iterations. The adaptive exponent for adjusting the decay rate based on the relative improvement value of the global optimal fitness is the amount of relative improvement of the global optimal fitness value. To adjust the attenuation power; This represents the standard deviation of the current population fitness values, reflecting the population diversity. This is a relatively small positive constant used to adjust for the impact of diversity on the inertia weights. In the early stages of the algorithm, a larger [value] is used. To facilitate global exploration, the population can quickly traverse the solution space. As the algorithm progresses, the inertia weights adaptively decrease in a non-linear manner based on the convergence speed and population diversity. This allows the algorithm to smoothly transition from global exploration to local exploration, improving its ability to avoid getting trapped in local optima while maintaining accurate convergence to the neighborhood of the global optimum. Fitness variance is introduced. Furthermore, the inertia weight can be dynamically adjusted to respond to the search status of the population. When the population distribution is relatively dispersed or fitness improvement is insufficient and the population stagnates, the inertia weight increases; when the population diversity decreases and converges stably, the inertia weight decreases, thereby achieving a more intelligent and robust trade-off between exploration and exploitation.

[0052] Compared to traditional algorithms, the decay type of the algorithm in this application is dynamic power-law decay (exponential). (Variable), attenuation power varies Dynamic adjustment, more flexible, and includes · The random perturbation term can be used to simulate complex, time-varying, and noisy weight changes (such as dynamic systems and financial models).

[0053] Furthermore, in S32, a co-evolutionary learning factor mechanism is designed. Traditional nonlinear learning factors typically adjust independently, neglecting the synergistic relationship between cognitive learning and social learning during the search process. To address this issue, this application proposes a co-evolutionary learning factor mechanism, enabling... and It not only possesses nonlinear time-varying characteristics, but also achieves interactive coupling through a feedback mechanism based on group convergence behavior.

[0054] The co-evolutionary learning factor is defined as follows: ; ; in, , , , These represent the maximum and minimum values ​​of the learning factor, respectively, where t is the iteration time, and T is the maximum and minimum values. max Indicates the total number of times. and This is an adaptive exponent based on the relative difference between the average fitness and the global optimal fitness. For average fitness, For optimal global fitness, through and The relative difference between them adjusts the nonlinear decay and growth rate; This represents the velocity of the i-th unit at time t, used to increase the velocity when the particle's motion stops. To enhance cognitive learning, Indicates the maximum speed. Population spatial diversity is represented by the average Euclidean distance between each particle and the global optimum, and increases when population diversity is high. To accelerate social learning, This represents the peak value of spatial diversity in a population.

[0055] In the early stages of the search, emphasis was placed on This encourages individual exploration and maintains population diversity. As evolution progresses, Gradually gaining dominance, the system promotes convergence towards the global optimum through enhanced social collaboration. The introduced speed and diversity feedback coupling mechanism enables the learning factor to dynamically adapt to the collective behavior of the group, thereby mitigating premature convergence and improving search efficiency. This collaborative adjustment mechanism, in conjunction with adaptive inertia weights, balances global exploration and local development at different stages of the optimization process.

[0056] Compared to traditional nonlinear learning factors, the learning factor in this application depends only on the number of iterations and the total number of iterations. Its adjustment is not only related to the iteration time t, but also introduces dynamic state variables (such as velocity and standard deviation) during algorithm operation. Furthermore, it achieves more flexible nonlinear changes through adaptive exponents p(t) and q(t), making it suitable for scenarios requiring dynamic parameter adjustments based on the algorithm's real-time state (such as particle swarm optimization and adaptive evolutionary algorithms). In addition, this application is more suitable for adaptive algorithms requiring dynamic feedback, such as in particle swarm optimization (PSO), where it is used to dynamically adjust inertia weights or learning factors, allowing parameters to change in real-time with the algorithm's convergence state.

[0057] Furthermore, in S33, an enhanced elite reverse learning mechanism is introduced. This mechanism, based on traditional elite reverse learning, introduces quasi-reverse points (randomly generated in the region between the interval center and the reverse point). This allows for a more balanced exploration of the search space, effectively coordinating large-scale leaps with local fine-grained search. Specifically, it selects the most fit individuals from the current population. Individuals form an elite particle swarm This guides the reverse learning process. Each elite particle in This will generate two different candidate solutions.

[0058] Elite Backward Learning Solution: This solution is generated using a standard elite backward learning strategy and is defined as follows: in, Indicates the first A dynamic range of dimensions is defined, which is adaptively updated with each generation based on the current population distribution to ensure accurate searching in hyperspace.

[0059] Quasi-inverse solution: This solution is generated between the center of the dynamic interval and the elite inverse solution. It is defined in each dimension as follows: in, Indicates the first The center of the dynamic interval.

[0060] The generated elite reverse solution Quasi-inverse solution A fitness evaluation is performed. The candidate solution with the better fitness is selected and paired with the original elite particle. The results are compared. If the optimal candidate solution has better fitness, it replaces the one in the population. Quasi-reverse solutions offer a more refined search strategy. They probe the region between the population center and elite reverse points, leveraging the fact that quasi-reverse points are more likely to find better solutions, thus enhancing global search capabilities and convergence robustness.

[0061] To further evaluate the impact of the improved particle swarm optimization (PSO) algorithm on system performance, a definition and evaluation method for the total system power are provided. The proposed improved PSO algorithm was evaluated using MATLAB simulation software in an indoor VLC scenario. For comparison, the traditional FPA (Fixed Power Allocation) algorithm and the traditional PSO algorithm were used as benchmarks. In the simulation, the room size was assumed to be 7m × 7m × 3m, the LED emitter was located at the center of the ceiling, and users were randomly distributed. The simulation was run 100 times, and the final average value is presented in this application.

[0062] To better evaluate the performance of the three schemes, this application analyzes the total power and convergence performance of the system under different interference cancellation factors and user numbers. The simulation uses a controlled variable method, with a default user number of 3 and a minimum SINR (signal-to-interference-plus-noise ratio) threshold of 3 dB.

[0063] like Figure 3 As shown in the attached figure, the performance curves of the total system power change under different signal-to-interference-plus-noise ratios (SNR) demonstrate the differences in system power performance under different interference cancellation factors η.

[0064] Specifically: the horizontal axis represents SINR (signal-to-interference-plus-noise ratio), measured in decibels, ranging from 0 to 4.5. A higher SINR indicates better signal quality and stronger anti-interference capability; the vertical axis represents the total system power, reflecting the system's data transmission capability.

[0065] The attached figure is a comparison of the total system power as a function of the number of users under different interference cancellation factors. The four curves represent different values ​​of η: 1. η=0 (dotted line): Throughout the entire SINR variation range, the total system power remains at a high level of approximately 185Mbps and is very stable, unaffected by SINR increases.

[0066] 2. η=0.01 (square line): The power is stable at about 70Mbps, and is also unaffected by SINR changes.

[0067] 3. η=0.02 (triangle line): The power is stable at about 60Mbps and is basically unaffected by SINR.

[0068] 4. η=0.03 (diamond line): The first half is stable at about 55Mbps, but when SINR exceeds 3.5dB, the power begins to drop sharply, and drops to almost 0Mbps when SINR=4.5dB.

[0069] When the value of η is small (0, 0.01, 0.02), the system power performance is stable, and the smaller the value of η, the higher the total system power.

[0070] When the value of η increases to 0.03, the system power will collapse under high SINR conditions, which indicates that a larger value of η will significantly reduce the transmission stability of the system under high signal quality.

[0071] like Figure 4-7 As shown in the attached figure, the total system power varies with the number of users, which is used to compare the performance of four different algorithms in a multi-user scenario.

[0072] The horizontal axis represents the number of users (from 2 to 6), and the vertical axis represents the total power of the system.

[0073] IPSO (Improved Particle Swarm Optimization Algorithm, solid circles, solid lines).

[0074] PSO (Particle Swarm Optimization Algorithm, square, dashed line).

[0075] FPA (Fixed Power Allocation Algorithm, solid triangle, dotted line).

[0076] OMA-TDMA (Orthogonal Time Division Multiple Access, solid diamond, solid line).

[0077] Figures 4-7 The order is: eta=0; eta=0.01; eta=0.02; eta=0.03.

[0078] like Figure 4 As shown, when η=0, the performance of the four algorithms is as follows: IPSO exhibits the best performance, with total system power showing the most stable and fastest growth as the number of users increases. When the number of users increases from 2 to 6, the power increases from approximately 150Mbps to approximately 270Mbps, which is the most significant improvement among the four options.

[0079] PSO performance is second only to IPSO, and its growth trend is very similar to that of IPSO. Power has increased from approximately 130 Mbps to approximately 260 Mbps.

[0080] FPA performance is moderate, with a relatively slow growth trend. Power increased from approximately 115 Mbps to approximately 190 Mbps.

[0081] OMA-TDMA has the worst performance, with the total system power remaining almost unchanged with the number of users, consistently around 30Mbps.

[0082] like Figure 5 As shown, when η=0.01, the performance of the four algorithms is as follows: IPSO offers the best performance, with total system power increasing rapidly and linearly with the number of users. It increases from approximately 60 Mbps with 2 users to approximately 110 Mbps with 6 users, representing the most significant improvement among the four options.

[0083] PSO performance showed steady growth, with power consistently lower than IPSO, but significantly better than FPA and OMA-TDMA. It increased from approximately 52 Mbps with 2 users to approximately 97 Mbps with 6 users.

[0084] The increase in FPA was gradual, with power increasing only from approximately 42 Mbps for 2 users to approximately 59 Mbps for 6 users.

[0085] OMA-TDMA has the worst performance and remains unchanged; regardless of the number of users, the total system power always remains at around 25Mbps.

[0086] like Figure 6 As shown, when η=0.02, the performance of the four algorithms is as follows: IPSO has the best performance. As the number of users increases from 2 to 6, the total system power continues to increase from about 38Mbps to about 82Mbps, showing the fastest growth rate and the highest final value.

[0087] PSO's performance was second only to IPSO. Power increased from approximately 38 Mbps to approximately 73 Mbps, with a stable growth trend, consistently slightly lower than IPSO.

[0088] The growth rate of FPA was relatively slow. Power increased from approximately 33 Mbps to approximately 62 Mbps, but its overall performance was weaker than IPSO and PSO.

[0089] OMA-TDMA performance was the worst and remained completely unchanged. Regardless of the increase in the number of users, the total system power remained stable at approximately 25 Mbps.

[0090] like Figure 7 As shown, when η=0.03, the performance of the four algorithms is as follows: IPSO has the best performance, and the total system power increases continuously with the number of users, growing from about 33Mbps to nearly 70Mbps, which is the fastest growth rate.

[0091] PSO's performance is second only to IPSO, and its growth trend is similar to that of IPSO, with power increasing from about 33Mbps to about 67Mbps.

[0092] FPA performance is moderate, with power increasing from 30Mbps to approximately 56Mbps, a relatively gradual increase.

[0093] OMA-TDMA had the worst performance and remained unchanged, with the total system power consistently at around 24Mbps, unaffected by an increase in the number of users.

[0094] Combination Figure 4-7 We can obtain: Algorithm performance ranking: In terms of total system power performance, IPSO > PSO > FPA > OMA-TDMA.

[0095] Impact of user number: The total system power of the three algorithms, IPSO, PSO and FPA, increases with the increase of user number, with IPSO showing the largest increase.

[0096] Limitations of OMA-TDMA: The total system power of OMA-TDMA is almost constant, indicating that it cannot effectively improve system capacity in multi-user scenarios, and its performance is significantly weaker than the other three intelligent optimization algorithms.

[0097] In summary, the improved particle swarm optimization (IPSO) has advantages in multi-user wireless communication systems. Through more efficient resource allocation, it significantly improves the total system power, which is far superior to the traditional OMA-TDMA scheme, as well as PSO and FPA algorithms.

[0098] like Figure 8-11 As shown, the iterative convergence curves for IPSO and PSO are shown in sequence for η=0; η=0.01; η=0.02; η=0.03.

[0099] Specifically: the horizontal axis represents the optimization and iteration process of the algorithm, which is the number of steps the algorithm takes.

[0100] The horizontal axis represents the optimization target, namely the total transmission power of the communication system. The higher the value, the better the performance.

[0101] Combined with appendix Figure 8-11 In terms of performance, IPSO converges faster and achieves higher final performance, while PSO converges slower and achieves lower final performance.

[0102] In summary, the improved IPSO algorithm significantly outperforms the traditional PSO algorithm in power allocation optimization of VLC-NOMA systems, both in terms of convergence speed and final performance.

[0103] It should be noted that although the various steps for the algorithm execution flow are mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more steps described above can be embodied in a single step. Conversely, the features and functions of one step described above can be further divided into multiple steps for embodiment. Some or all of the steps can be selected to achieve the purpose of the solution in this application, depending on actual needs. Those skilled in the art can understand and implement this without any inventive effort.

[0104] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

Claims

1. A power allocation method for a VLC-NOMA system based on improved particle swarm optimization, characterized in that, Includes the following steps: A downlink model based on a VLC-NOMA system is constructed, in which the user channel gain is calculated using the Lambertian model. A power allocation optimization model is constructed with the objective of maximizing the total rate of the VLC-NOMA system. The total power of the system is constrained based on this power allocation optimization model, which is defined as follows: ； in, The total system rate is the sum of the achievable rates for all successfully demodulated users. This represents the maximum total system speed. This represents the rate that user k can achieve after considering the interference cancellation factor, where K represents the total number of users; The power allocation optimization model is solved using the IPSO algorithm to obtain the optimal power allocation scheme. The IPSO algorithm includes an adaptive dynamic inertia weight mechanism, a collaborative learning factor mechanism, and an enhanced elite reverse learning mechanism. The adaptive dynamic inertia weighting mechanism introduces a dynamic power function and random perturbation to adapt to the complex, time-varying and noisy weight changes in the simulation. The collaborative learning factor mechanism introduces a set of co-evolutionary learning factors. and ,pass and This enables learning factors to dynamically adapt to the collective behavior of the group; The enhanced elite reverse learning mechanism introduces a quasi-reverse point, which selects the one with better fitness by comparing the elite reverse solution and the quasi-reverse solution.

2. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 1, characterized in that, The downlink model based on the VLC-NOMA system includes an LED transmitter and K randomly distributed users. The LED transmitter is located at the center of the ceiling and is allocated different powers according to the differences in user channel gain. At the receiving end, each user is equipped with a single photodetector to receive the mixed optical signal from the LED transmitter. The channel gain between the LED transmitter and the user is calculated using the Lambertian model, with the following formula: ； Where m is the order of the Lambert radiation. Indicates user The receiving area of ​​the photodetector. , and Representing users respectively The incident angle, the field of view of the photodetector, and the emission angle of the LED emitter. Indicates LED transmitter and user The distance between them This indicates the directional attenuation term of the LED emitter. Indicates user The gain of the optical filter is usually taken as ; This represents the cosine projection attenuation factor at the receiving end. Indicates user The gain of the optical concentrator.

3. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 1, characterized in that, The adaptive dynamic inertia weighting mechanism is defined as follows: ； in, and These are the maximum and minimum values ​​of the inertia weight, respectively. This represents the current iteration number. The maximum number of iterations, An adaptive exponent for adjusting the decay rate based on the relative improvement value of the globally optimal fitness. The standard deviation of the current population fitness value. It is a positive integer.

4. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 1, characterized in that, and This is a set of co-evolutionary learning factors, and the mechanism of the co-learning factors is defined as follows: ； ； in, , , , These represent the maximum and minimum values ​​of the learning factor, respectively, where t is the iteration time, and T is the maximum and minimum values. max Indicates the total number of times. and It is an adaptive exponent based on the relative difference between the average fitness and the global optimal fitness. Indicates the maximum speed. This represents the running speed of the i-th unit at time t. Indicates population spatial diversity, The peak value representing the spatial diversity of a population. This represents the average fitness of the population at time t. This represents the global optimal fitness of the population at time t.

5. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 1, characterized in that, The enhanced elite reverse learning mechanism includes: Select the strain with the highest fitness from the current population. Individuals form an elite particle swarm ,for Each elite particle in This generates two different candidate solutions: the elite inverse solution and the quasi-inverse solution. The elite reverse solution is generated using a standard elite reverse learning strategy and is defined as follows: ； in, Indicates the first A dynamic interval of dimensions, which is adaptively updated with each generation based on the current population distribution; The quasi-inverse solution is generated between the center of the dynamic interval and the elite inverse solution, and is defined in each dimension as follows: ； in Indicates the first The center of the dynamic range; Reverse solution from generated elites Quasi-inverse solution Select the one with better fitness from the original elite particles Compare and select the one with better fitness to replace the one in the population. .

6. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 1, characterized in that, The power allocation optimization model satisfies the following constraints: human eye safety regulations and total power constraints. The aforementioned eye safety regulations are defined as follows: ； in, This represents the DC bias used to ensure that the transmitted optical signal is non-negative. This represents the peak luminous intensity of the LED emitter. The total power constraint is defined as follows: ； in, This represents the power allocated to the k-th user. This indicates the maximum total transmit power allowed by the system.

7. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 6, characterized in that, The power allocation optimization model also includes: non-negative power constraints and user service quality assurance constraints. The non-negative power constraint is defined as follows: ； in, This represents the transmit power of the k-th user; The user service quality assurance constraint is defined as follows: ； in, This represents the actual signal-to-interference-plus-noise ratio (SIR) for the k-th user. This represents the minimum signal-to-interference-plus-noise ratio (SINNR) threshold required to ensure the quality of service for users.

8. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 3, characterized in that, The definition of is: ； in, The relative improvement is calculated as follows: ((Optimized performance index - Unoptimized performance index) / Unoptimized performance index) × 100%.

9. A power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 2, characterized in that, The definition of is: ； in, This represents the refractive index of the light concentrator.

10. The power allocation method for a VLC-NOMA system based on improved particle swarm optimization according to claim 2, characterized in that, The order of the Lambert radiation is expressed as follows: ； in, This indicates the half-power angle of the LED emitter.

Images