A multi-user, multi-service resource sensing type scheduling method suitable for a low earth orbit satellite system

Abstract

The application discloses a multi-user and multi-service resource sensing type scheduling method suitable for a low-orbit satellite system and belongs to the field of satellite communication. First, the priority of each service is calculated in the initial time slot of the current scheduling period for the low-orbit satellite multi-beam communication scene, each user is sorted and then is sequentially scheduled, and the unoccupied PRB is distributed. The throughput of each user is calculated. In the next time slot, the throughput of each user is recalculated until the scheduling period cycle ends. The optimization target is constructed to maximize the system throughput, and the system total throughput under the current weight combination is recorded. Finally, the learning rate and the maximum iteration number are set, the weight is updated, the next scheduling period is entered, the system throughput under the current weight combination is recalculated, the optimal weight combination maximizing the system throughput is screened out through iteration, and the optimal weight combination is output. The application realizes the maximization of the on-orbit resource utilization efficiency and guarantees the fairness of user service.

Description

Technical Field

[0001] This invention belongs to the field of satellite communication, specifically relating to a multi-user, multi-service resource-aware scheduling method suitable for low-Earth orbit satellite systems. Background Technology

[0002] Based on their orbital altitude, satellites are classified into Geosynchronous Earth Orbit (GEO) satellites, Medium Earth Orbit (MEO) satellites, and Low Earth Orbit (LEO) satellites. Compared to GEO and MEO, LEO satellites, due to their lower orbital altitude, have lower signal transmission latency, lower path loss, higher spectral efficiency, and lower manufacturing cost, making them more compatible with terrestrial 5G communication networks in integrated space-ground networks. However, LEO satellite networks primarily consist of small satellites, meaning that effective payloads and spectrum resources are scarce. This can easily lead to spectrum congestion when satellites simultaneously serve a large number of users, resulting in degraded communication quality. To improve the utilization rate of onboard payloads and ensure that limited satellite resources are allocated on demand, research into resource scheduling strategies is crucial.

[0003] In low-Earth orbit (LEO) satellite systems, resource scheduling is a key technology for optimizing communication performance and ensuring user fairness. Traditional resource scheduling methods are primarily designed for terrestrial mobile communication networks and perform well in static or low-dynamic scenarios. However, in LEO satellite systems, the frequent changes in network topology due to continuous satellite movement and significant differences in user channel conditions across wide coverage areas severely limit the performance of traditional scheduling schemes. Particularly when allocating resources to edge users, it is often impossible to simultaneously guarantee system efficiency and user fairness. Furthermore, the lack of flexible priority adjustment mechanisms when handling diverse service demands prevents the optimal allocation of system resources.

[0004] In low-Earth orbit satellite systems, existing resource scheduling schemes mainly face the following technical challenges:

[0005] First, users are distributed in different geographical locations and have diverse business needs, including users with real-time services that are sensitive to latency and users with non-real-time services that have high throughput requirements. At the same time, users located at the edge of the satellite coverage area often face the problem of poor channel quality, while users in the central area have better channel conditions. This dual complexity of user distribution and business needs makes it difficult for traditional single-dimensional scheduling strategies to adapt.

[0006] Secondly, real-time service users require low latency and stable resource guarantees, placing extremely high demands on the timeliness of resource allocation. Non-real-time service users, on the other hand, are more concerned with system throughput and expect to obtain more resources to improve transmission efficiency when channel conditions are favorable. This difference in service requirements conflicts with limited system resources.

[0007] Furthermore, users at the edge of satellite coverage areas are easily marginalized in traditional scheduling algorithms that prioritize channel quality due to poor channel quality, often making it difficult for them to obtain fair service opportunities.

[0008] Therefore, how to design a resource scheduling scheme that can comprehensively consider user geographical location and business type under limited system resources, and achieve the joint optimization of system performance and user experience, is a key problem that urgently needs to be solved. Summary of the Invention

[0009] To address the aforementioned issues, this invention proposes a multi-user, multi-service resource-aware scheduling method suitable for low-Earth orbit satellite systems, which maximizes on-board resource utilization efficiency and ensures fairness in user services.

[0010] The specific steps of the multi-user, multi-service resource-aware scheduling method applicable to low-Earth orbit satellite systems are as follows:

[0011] Step 1: Establish a low-orbit satellite multibeam system for the application scenario;

[0012] The scenario includes several users with different business needs and multiple satellites, with each user corresponding to a different business type; the service area is divided into multiple cells, and the network control center is responsible for the unified allocation of resources and the collection and processing of business demand information in each area.

[0013] Step 2: For each user in the scenario, randomize the user's initial task queue and calculate the SINR for each user.

[0014] User's initial task queue Q j j = 1, 2, 3, ..., J; the total number of users is J;

[0015] Step 3: Using different modulation and coding schemes (MCS), plot the BLER-SINR curves of bit error rate versus signal-to-noise ratio, and obtain the optimal MCS level based on the SINR of each user;

[0016] The specific acquisition method is as follows: find the SINR threshold value where BLER < 0.1 in the graph, compare the SINR calculated by each user with the threshold value, and the curve with a SINR greater than the threshold value is the MCS that meets the condition;

[0017] Step 4: In the initial time slot of the current scheduling cycle, calculate the priority of each service and sort each user according to the priority;

[0018] Business types are divided into: real-time business and non-real-time business;

[0019] The formula for calculating the priority of real-time services is:

[0020]

[0021] Where w1 is the weight controlling parameters such as latency and packet loss rate, and w2 is the weight controlling the geographic location factor, with initial values ​​set manually. j =-log(P j-loss ) / τ j P j-loss τ represents the maximum allowable packet loss rate for user j during data transmission. j This represents the tolerable latency threshold. j (t) represents the instantaneous data transmission rate of user j in time slot t. R j (t) represents the average data packet transmission rate of user j in time slot t, W j (t) represents the waiting delay of user j's data in time slot t. j (t) refers to the length of the remaining data queue in time slot t. j (t) refers to the position factor introduced in time slot t, which changes dynamically at each scheduling time; the formula for calculating this dynamically adjusted position factor is:

[0022]

[0023] Where d refers to the straight-line distance between the user and the point on the Earth's surface projected by the satellite. This refers to the standard deviation of Gaussian decay, which determines the rate at which the decay rate decreases with distance.

[0024] The priority calculation formula for non-real-time services is as follows:

[0025]

[0026] Step 5: Schedule each user in the sorted order and allocate unoccupied Physical Resource Blocks (PRBs) to each user in sequence according to channel quality;

[0027] B t =[rb1,rb2,rb3,...,rb b ,..,rb total ] represents the set of physical resource blocks in time slot t, if rb b A value of 1 indicates that the PRB is occupied;

[0028] Step 6: Update the number of Physical Resource Blocks (PRBs) and their occupancy status. Calculate the throughput for each user based on the optimal MCS tier, and record the remaining transmission queue q for each user. j (t);

[0029] The throughput calculation formula for user j is expressed as: thj = N PRB ·n RE_per_PRB ·M q ·C q ;

[0030] Where N PRB Indicates the total number of PRBs allocated to the user; n RE_per_PRB M represents the number of resource particles (REs) available for data transfer in each PRB; q C represents the modulation order of the selected MCS with index q. q This indicates the coding rate of the selected MCS with index q.

[0031] Step 7: Proceed to the next time slot, return to Step 4, and recalculate the throughput of each user in this time slot until the scheduling cycle ends.

[0032] Step 8: Based on the throughput of each user recorded in each time slot, construct an optimization objective to maximize system throughput, and record the total system throughput under the current weight combinations w1 and w2;

[0033] The optimization objective is:

[0034]

[0035] T represents the total time slots;

[0036] Constraint 1 states that the sum of the weighting factors must be 1; Constraint 2 states that the bit error rate threshold in the BLER-SINR curve is 0.1; ρ q This indicates the SINR required for the MCS with index q; constraint 3 states that each PRB can only be occupied once in each time slot; constraint 4 states that the number of PRBs allocated to all users in each time slot cannot exceed the total number of PRBs in that time slot, B. total,t Constraint 5 states that the number of PRBs allocated to each user in each time slot cannot be less than 0; Constraint 6 is the Jain fairness mechanism, which states that the minimum fairness constraint is 0.85, and N represents the number of users scheduled.

[0037] Step 9: Set the learning rate η and the maximum number of iterations N max Update weight w1 t+1 =w1 t +η, enter the next scheduling cycle, return to step four, adjust the weight factor w1, and recalculate the system throughput under the current weight combination;

[0038] Step 10: Record all weight combinations and their corresponding throughput values, iterate through and filter out the optimal weight combination that maximizes the system throughput, and output it.

[0039] The advantages of this invention are:

[0040] This invention presents a multi-user, multi-service resource-aware scheduling method applicable to low-Earth orbit satellite systems. It comprehensively considers the differentiated service requirements of real-time and non-real-time services, as well as the impact of user geographical distribution characteristics on resource allocation. By quantitatively analyzing key factors such as user transmission rate, queue waiting time, and geographical location, a scientific priority ranking mechanism is established. Simulation results show that compared to the M-LWDF algorithm and the proportional fairness algorithm, the system described in this invention significantly improves throughput and user fairness, while achieving a packet loss rate similar to that of the M-LWDF algorithm. Attached Figure Description

[0041] Figure 1 This is a flowchart of a multi-user, multi-service resource-aware scheduling method applicable to low-Earth orbit satellite systems according to the present invention;

[0042] Figure 2 This is a scene diagram of the low-orbit satellite multi-beam system constructed according to the present invention;

[0043] Figure 3 This is the BLER-SINR curve plotted by this invention;

[0044] Figure 4 This is a comparison chart of the user fairness between the method described in this invention and existing algorithms;

[0045] Figure 5 This is a comparison chart of the system throughput between the method described in this invention and existing algorithms;

[0046] Figure 6 This is a comparison chart of the packet loss rate between the method described in this invention and existing algorithms; Detailed Implementation

[0047] The specific implementation method of the present invention will be further described in detail below with reference to the accompanying drawings.

[0048] This invention proposes a multi-user, multi-service resource-aware scheduling strategy for low Earth Orbit satellite systems. Addressing the user-level resource allocation problem, it combines the differentiated needs of user services with the characteristics of user geographical distribution to maximize resource utilization efficiency and improve the fairness of user services.

[0049] This invention achieves dynamic allocation of resources among multiple users and multiple PRBs through a resource-aware strategy. First, the optimization objective is to maximize system throughput while ensuring service fairness. Second, based on the Maximum Weighted Delay First (M-LWDF) algorithm, a multi-dimensional priority criterion for resource allocation is established, and a priority function is designed. The system comprehensively evaluates user service needs, queue status, channel quality, and geographical location, and completes each round of resource allocation based on the priority sequence. The allocation output is the resource acquisition plan for each user. Within this resource allocation plan, the system can further determine which users will receive service resources in a given time slot.

[0050] The aforementioned multi-user, multi-service resource-aware scheduling method applicable to low-Earth orbit satellite systems, such as Figure 1 As shown, the specific steps are as follows:

[0051] Step 1: Establish a low-orbit satellite multibeam system for the application scenario;

[0052] like Figure 2 As shown, the scenario includes several users with different business needs and multiple satellites. The users are geographically distributed, and each user corresponds to a different business type.

[0053] The service area is divided into multiple zones, and the network control center is responsible for the unified allocation of resources and the collection and processing of business demand information from each zone.

[0054] Step 2: For each user in the scenario, randomize the user's initial task queue and calculate the SINR for each user.

[0055] User's initial task queue Q j j = 1, 2, 3, ..., J; the total number of users is J;

[0056] The path loss model for satellite and ground station is as follows:

[0057] PL = PL b +PL g +PL s +PL e (1)

[0058] PL b It is the basic path loss, PL g It is atmospheric decay, PL s It is ionospheric decay, PL e It is penetration loss.

[0059] The basic path loss is calculated as follows:

[0060] PL b =FSPL(d,f c )+SF+CL(α,f c (2)

[0061] Here, SF refers to shadow fading, and CL(α,f) c ) refers to group fading, FSPL(d,f c This refers to the path loss in free space, and its calculation formula is:

[0062] FSPL(d,f c ) = 32.45 + 20log 10 (f c )+20log 10 (d) (3)

[0063] Where d refers to the transmission distance between the satellite and the user, and f c This refers to the carrier center frequency, measured in GHz.

[0064]

[0065] Where R E α represents the Earth's radius, and α represents the user's position relative to the satellite's elevation angle.

[0066] In communication links between low-Earth orbit satellites and ground users, the signal-to-noise ratio plus interference ratio (SINR) is a core indicator for measuring signal quality, directly affecting the reliability of the link and data transmission efficiency. SINR can be viewed as the ratio of the received effective signal power to the interference and noise power. Essentially, it describes the ability of a signal to be received and decoded even after being subjected to external environmental interference. The formula for calculating the SINR of the j-th user is as follows:

[0067]

[0068] EIRP j Refers to the effective isotropic radiated power of user j. is the antenna gain to noise-temperature ratio of user j, k is the Boltzmann constant, equal to -228.6 dBW / K / Hz, B is the system bandwidth, and PL refers to the path loss.

[0069] Step 3: Using different modulation and coding schemes (MCS), plot the BLER-SINR curves of bit error rate versus signal-to-noise ratio, and obtain the optimal MCS level based on the SINR of each user;

[0070] The specific method for obtaining the data is as follows: Based on an existing link-level simulation platform, system performance is simulated for different modulation and coding strategies to obtain standard BLER-SINR curves, such as... Figure 3 As shown in the figure, find the SINR threshold value where BLER < 0.1, compare the SINR calculated by each user with the threshold value, and the curve with a SINR greater than this value is the MCS that meets the condition.

[0071] Step 4: In the initial time slot of the current scheduling cycle, calculate the priority of each service and sort each user according to the priority;

[0072] To meet the diverse needs of different service types in low-Earth orbit satellite multi-beam scenarios, achieve on-demand allocation, satisfy diverse requirements such as low latency and high bandwidth, balance the quality of service for each user, and achieve coverage consistency, this embodiment considers M-LWDF, incorporating service type and user location considerations to dynamically adjust priority order and resource allocation decisions.

[0073] In satellite systems, based on the sensitivity of services to latency, service types are divided into two main categories: real-time services and non-real-time services. M=0 represents real-time services, and m=1 represents non-real-time services.

[0074] For real-time services, the priority calculation formula is as follows:

[0075]

[0076] Where w1 is the weight controlling parameters such as latency and packet loss rate, and w2 is the weight controlling the geographic location factor, with initial values ​​set manually. j =-log(P j-loss ) / τ j P j-loss τ represents the maximum allowable packet loss rate for user j during data transmission. j This represents the tolerable latency threshold. R j (t) represents the average data packet transmission rate of user j in time slot t, W j (t) represents the waiting delay of user j's data in time slot t. j (t) refers to the length of the remaining data queue in time slot t. j (t) represents the instantaneous data transmission rate of user j in time slot t, calculated using the following formula:

[0077]

[0078] in, This indicates how many bits of data each OFDM symbol can transmit, which is determined by the MCS selected by the user. This indicates the number of OFDM symbols contained in each time slot. This indicates the number of time slots contained in each scheduling cycle. This indicates the number of subcarriers contained in each RB.

[0079] L j (t) refers to the position factor introduced in time slot t, which changes dynamically at each scheduling time; the formula for calculating this dynamically adjusted position factor is: Where d refers to the straight-line distance between the user and the point on the Earth's surface projected by the satellite. This refers to the standard deviation of Gaussian decay, which determines the rate at which the decay rate decreases with distance. The smaller the value, the faster the decay, meaning that the majority of the priority is distributed within a smaller radius. The larger the value, the slower the attenuation; that is, some priority can be retained at greater distances. For low-Earth orbit satellite systems, which have a large coverage area, a larger value is designed based on the range of the edge and near-satellite regions. To control the decay.

[0080] The location factor is used to adjust the priority of marginal users or users who have been inactive for a long time during the resource allocation process based on their recent service status (such as the quality of allocated resources, duration of continuous inactivity, etc.), thereby finding a balance between fairness and resource efficiency. j The introduction of (t) combines user location with channel conditions, giving sub-satellite users a higher score in priority calculation and thus prioritizing resource allocation. For edge users, due to their poor channel conditions and high throughput improvement costs, the location factor L is weakened. j (t) Reduce its priority to avoid consuming too many resources in resource allocation.

[0081] The specific adjustment principle is as follows: for users whose service needs cannot be met (such as edge users), their priority will be appropriately increased based on the number of times they have not been served. When resources are sufficient, the needs of users at the base station will be prioritized; however, if base station users repeatedly over-consume resources, reducing fairness to some edge users, the service access level (L) of base station users may be appropriately limited. j (t). Location factor considerations have been incorporated to ensure that users at the beam edge can also be scheduled.

[0082] The priority calculation formula for non-real-time services is as follows:

[0083]

[0084] Step 5: Schedule each user in the sorted order and allocate unoccupied Physical Resource Blocks (PRBs) to each user in sequence according to channel quality;

[0085] In this scenario, a single PRB resource block can only be allocated to one user at a time, but a user can acquire multiple PRB resource blocks simultaneously. The system allocates resources to users based on priority. Users select the resource block with the best channel quality from the available PRBs and mark it as occupied. Each round of allocation updates the resource allocation results, the list of unallocated resources, the user queue status, and the user priority sequence. After several rounds of allocation, the final resource allocation scheme for each user and the user to whom each PRB belongs are determined, ultimately completing a many-to-many matching of users and resources.

[0086] B t =[rb1,rb2,rb3,...,rb b ,..,rb total ] represents the set of physical resource blocks in time slot t, if rb b A value of 1 indicates that the PRB is occupied;

[0087] Step 6: Update the number of Physical Resource Blocks (PRBs) and their occupancy status. Calculate the throughput for each user based on the optimal MCS tier, and record the remaining transmission queue q for each user. j (t);

[0088] The throughput calculation formula for user j is expressed as: thj = N PRB ·n RE_per_PRB ·M q ·C q ;

[0089] Where N PRB Indicates the total number of PRBs allocated to the user; n RE_per_PRB This represents the number of resource elements (REs) available for data transmission in each PRB. One resource element (RE) occupies one subcarrier in the frequency domain and one OFDM symbol in the time domain; M q C represents the modulation order of the selected MCS with index q. q This indicates the coding rate of the selected MCS with index q.

[0090] Step 7: Proceed to the next time slot, return to Step 4, and recalculate the throughput of each user in this time slot until the scheduling cycle ends.

[0091] Step 8: Based on the throughput of each user recorded in each time slot, construct an optimization objective to maximize system throughput, and record the total system throughput under the current weight combinations w1 and w2;

[0092] To maximize the utilization of onboard resources and ensure fairness in user services, the optimization objective is as follows:

[0093]

[0094] T represents the total time slots;

[0095] Constraint 1 states that the sum of the weighting factors must be 1; Constraint 2 states that the bit error rate threshold in the BLER-SINR curve is 0.1; ρ q This indicates the SINR required for the MCS with index q; constraint 3 states that each PRB can only be occupied once in each time slot; constraint 4 states that the number of PRBs allocated to all users in each time slot cannot exceed the total number of PRBs in that time slot, B. total,t Constraint 5 states that the number of PRBs allocated to each user in each time slot cannot be less than 0; Constraint 6 is the Jain fairness mechanism, which states that the minimum fairness constraint is 0.85, and N represents the number of users scheduled.

[0096] Step 9: Set the learning rate η and the maximum number of iterations N max Update weight w1 t+1 =w1 t +η, enter the next scheduling cycle, return to step four, adjust the weight factor w1, and recalculate the system throughput under the current weight combination;

[0097] Step 10: Record all weight combinations and their corresponding throughput values, iterate through and filter out the optimal weight combination that maximizes the system throughput, and output it.

[0098] like Figure 4 , Figure 5 and Figure 6 As shown in the figure, the algorithm proposed in this invention has the best performance. Figure 4 The results show that as the number of users increases, resource competition intensifies. The algorithm proposed in this invention exhibits the smallest decrease in fairness and the most stable performance, while proportional fairness and M-LWDF show a larger decrease in fairness. Therefore, the algorithm proposed in this invention performs best in terms of resource allocation and user fairness optimization. Figure 5 The results show that as the number of users increases, the throughput of the three algorithms generally shows an increasing trend, but the algorithm proposed in this invention shows the most obvious trend, maintaining a high level even when the number of users is 30, with the highest throughput approaching 2.6*10^4bps. Figure 6 The results show that as the number of users increases, the packet loss rate of all three algorithms increases. The algorithm proposed in this invention shows the smallest increase in packet loss rate, indicating that it performs best in terms of resource allocation and packet loss rate control.

Claims

1. A multi-user, multi-service resource-aware scheduling method suitable for low-Earth orbit satellite systems, characterized in that, The specific steps are as follows: Step 1: Establish a low-orbit satellite multibeam system for the application scenario; Step 2: For each user in the scenario, randomize the user's initial task queue and calculate the SINR for each user. User's initial task queue Q j j = 1, 2, 3, ..., J; the total number of users is J; Step 3: Using different modulation and coding schemes (MCS), plot the BLER-SINR curves of bit error rate versus signal-to-noise ratio, and obtain the optimal MCS level based on the SINR of each user; Step 4: In the initial time slot of the current scheduling cycle, calculate the priority of each service and sort each user according to the priority; Business types are divided into: real-time business and non-real-time business; The formula for calculating the priority of real-time services is: Where w1 is the weight of the parameter, w2 is the weight of the geographic location factor, and the initial value is set manually, a j =-log(P j-loss ) / τ j P j-loss τ represents the maximum allowable packet loss rate for user j during data transmission. j R represents the tolerable latency threshold. j (t) represents the instantaneous data transmission rate of user j in time slot t, R j (t) represents the average data packet transmission rate of user j in time slot t, W j (t) represents the waiting delay of user j's data in time slot t, q j (t) refers to the length of the remaining data queue in time slot t, L j (t) refers to the position factor introduced in time slot t; The priority calculation formula for non-real-time services is as follows: Step 5: Schedule each user in the sorted order and allocate unoccupied Physical Resource Blocks (PRBs) to each user in sequence according to channel quality; B t =[rb1,rb2,rb3,...,rb b ,..,rb total ] represents the set of physical resource blocks in time slot t, if rb b A value of 1 indicates that the PRB is occupied; Step 6: Update the number of Physical Resource Blocks (PRBs) and their occupancy status. Calculate the throughput for each user based on the optimal MCS tier, and record the remaining transmission queue q for each user. j (t); The throughput calculation formula for user j is expressed as: th j =N PRB ·n RE_per_PRB ·M q ·C q ; Where N PRB Indicates the total number of PRBs allocated to the user; n RE_per_PRB M represents the number of resource particles (REs) available for data transfer in each PRB; q C represents the modulation order of the selected MCS with index q. q This indicates the coding rate of the selected MCS with index q; Step 7: Proceed to the next time slot, return to Step 4, and recalculate the throughput of each user in this time slot until the scheduling cycle ends. Step 8: Based on the throughput of each user recorded in each time slot, construct an optimization objective to maximize system throughput, and record the total system throughput under the current weight combinations w1 and w2; Step 9: Set the learning rate η and the maximum number of iterations N max Update weight w1 t+1 =w1 t +η, enter the next scheduling cycle, return to step four, adjust the weight factor w1, and recalculate the system throughput under the current weight combination; Step 10: Record all weight combinations and their corresponding throughput values, iterate through and filter out the optimal weight combination that maximizes the system throughput, and output it.

2. The scheduling method as described in claim 1, characterized in that, In step one, the scenario includes several users with different business needs and multiple satellites, with each user corresponding to a different business type; the service area is divided into multiple cells, and the network control center is responsible for uniformly allocating resources and collecting and processing business demand information for each area.

3. The scheduling method as described in claim 1, characterized in that, The specific method for obtaining step three is as follows: find the SINR threshold value where BLER < 0.1 in the BLER-SINR curve graph, compare the SINR calculated by each user with the threshold value, and the curve with a SINR greater than the threshold value is the MCS that meets the condition.

4. The scheduling method as described in claim 1, characterized in that, In step four, the position factor L j (t) changes dynamically at each scheduling time, and the calculation formula is: Where d refers to the straight-line distance between the user and the point on the Earth's surface projected by the satellite. This refers to the standard deviation of Gaussian decay, which determines the rate at which the decay rate decreases with distance.

5. The scheduling method as described in claim 1, characterized in that, The optimization objective of step eight is: T represents the total time slots; Constraint 1 states that the sum of the weighting factors must be 1; Constraint 2 states that the bit error rate threshold in the BLER-SINR curve is 0.1; ρ q This indicates the SINR required for the MCS with index q; constraint 3 states that each PRB can only be occupied once in each time slot; constraint 4 states that the number of PRBs allocated to all users in each time slot cannot exceed the total number of PRBs in that time slot, B. total,t Constraint 5 indicates that the number of PRBs allocated to each user in each time slot cannot be less than 0; Constraint 6 is the Jain fairness mechanism, indicating that the fairness constraint is at least 0.85, and N represents the number of users scheduled.

Images