User scheduling method, device and equipment and readable storage medium

Abstract

The invention provides a user scheduling method, device and equipment and a readable storage medium, and relates to the technical field of communication, and the method comprises the steps: determining the transmission error probability of each downlink user to be served under the condition that a limited block length is adopted for data transmission; according to the transmission error probability, the limited block length and the transmitter transmission rate, determining a first service user number, the first service user number including a first uplink user number and a first downlink user number scheduled in a time-frequency domain resource; according to the first service user number and the transmission error probability, user grouping is carried out on a user set to be served, and the user set to be served comprises uplink users to be served and downlink users to be served. According to the scheme, the uplink user number and the downlink user number scheduled in the time-frequency domain resource are determined based on the transmission error probability, a basis is provided for selection of the service user number, the transmission error probability of the system can be effectively reduced, and service requirements are met.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a user scheduling method, device, equipment and readable storage medium. Background Art

[0002] With the rapid development of massive Internet of Things (mIoT) technology, the massive machine type communication and ultra-reliable low latency communication requirements in the 5G standard have been integrated into the new 6G service standard, forming the massive ultra-reliable low latency communication (mURLLC) requirement. In the indoor IoT scenario, the overall coverage of the system is huge. All smart devices and mobile devices of staff in the factory can be used as potential access devices of the system. If they are served at the same time, transmission errors will occur frequently, making it difficult to meet the user's quality of service (QoS) requirements; dense user access will also cause cross-link interference (CLI) between users to occur more frequently at the symbol level, deteriorating the service quality. Summary of the invention

[0003] The embodiments of the present application provide a user scheduling method, apparatus, device and readable storage medium, which solve the problem that transmission errors frequently occur when dense IoT devices access the system and CLI frequently occurs at the symbol level, thereby deteriorating the service quality.

[0004] In a first aspect, in order to achieve the above-mentioned object, an embodiment of the present application provides a user scheduling method, comprising:

[0005] Determining a transmission error probability for each downlink user to be served when a finite block length is used for data transmission;

[0006] Determine a first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes a first number of uplink users and a first number of downlink users scheduled in a time-frequency domain resource;

[0007] According to the first number of service users and the transmission error probability, a user set to be served is grouped into users, wherein the user set to be served includes uplink users to be served and downlink users to be served.

[0008] Optionally, in the case of limited block length data transmission, determining the transmission error probability of each downlink user to be served includes:

[0009] For a first downlink user, determining a first relationship between a transmitter transmission rate of the first downlink user and a transmission error probability of the first downlink user;

[0010] Determine the transmission error probability of the first downlink user according to the first relationship using random block length equivalent capacity and first-order Taylor approximation;

[0011] The first downlink user is any downlink user among the downlink users to be served.

[0012] Optionally, determining the transmission error probability of the first downlink user according to the first relationship using a random block length equivalent capacity and a first-order Taylor approximation includes:

[0013] According to the formula:

[0014]

[0015] determining a transmission error probability of the first downlink user;

[0016] Among them, R k (·) represents the transmitter transmission rate of the first downlink user, n represents the finite block length, γ k represents the signal-to-interference-noise ratio of the first downlink user, ε k represents the transmission error probability of the first downlink user, represents the effective signal power of the first downlink user, γ R represents the channel capacity corresponding to the transmitter rate, σ IF,k represents the variance of the channel estimation error of the first downlink user and the finite block length error, λ c,k represents the total interference signal power received by the first downlink user, μ IF,k Represents the mean of the channel estimation error and the finite block length error of the first downlink user.

[0017] Optionally, determining the number of first service users according to the transmission error probability, the finite block length, and a transmitter transmission rate includes:

[0018] Determining an average service rate of the downlink users to be served according to the transmission error probability of each of the downlink users to be served;

[0019] When the constraint condition is that the ratio of the total number of uplink users in a time domain resource is greater than or equal to a preset ratio, the number of uplink users and the number of downlink users are traversed with the goal of achieving the maximum value of the average service rate to solve the first number of service users.

[0020] Optionally, grouping the user set to be served according to the number of the first service users and the transmission error probability includes:

[0021] The number of the first service users, the transmission error probability of each downlink user to be served, and the user set to be served are input into a pre-trained deep Q network model, the user set to be served is grouped, and a grouping result is obtained.

[0022] Optionally, the method further comprises: training the deep Q network model based on the following steps:

[0023] Establishing an optimization problem, wherein the optimization objective of the optimization problem is to minimize the inter-user interference IUI, the optimization variable of the optimization problem is the grouping mode, and the constraints of the optimization problem include: the transmission error probability of the downlink user is less than or equal to the transmission error probability threshold and each of the downlink users to be served or each of the uplink users to be served is located in only one user group;

[0024] Determine the number of second service users corresponding to the user training set, wherein the number of second service users includes the number of second uplink users and the number of second downlink users scheduled in one time-frequency domain resource;

[0025] Grouping users in the user training set;

[0026] The deep Q network model is iteratively trained according to the current grouping method, the optimization problem and the number of the second service users.

[0027] Optionally, iteratively training the deep Q network model according to the current grouping mode, the optimization problem and the number of the second service users includes:

[0028] Calculate the inter-user interference IUI corresponding to the current grouping mode;

[0029] When the current grouping method satisfies the constraint conditions of the optimization problem and the number of users of the second service, regroup the users in the user training set according to the current grouping method, and return to the step of calculating the IUI corresponding to the current grouping method;

[0030] After obtaining the IUIs corresponding to the multiple grouping modes, calculating the relative training error of the deep Q network model according to the minimum IUI;

[0031] According to the relative training error, the deep Q network model is iteratively learned until the IUI converges to a preset feedback value.

[0032] Optionally, calculating the inter-user interference IUI corresponding to the current grouping mode includes:

[0033] For each user group, calculating the signal to interference and noise ratio between each downlink user and each uplink user;

[0034] For each of the user groups, a sum of target signal to interference plus noise ratios is calculated; wherein the target signal to interference plus noise ratio is less than or equal to a transmission signal to noise ratio of an uplink user of the user group;

[0035] The IUI is determined according to the sum of the target signal to interference and noise ratios of the user groups.

[0036] Optionally, regrouping the users in the user training set according to the current grouping method includes:

[0037] Select the first user whose user group needs to be adjusted;

[0038] When the first target user group to which the first user belongs is a preset user group, randomly regrouping the first users;

[0039] When the first target user group is not the preset user group, the first user is randomly adjusted with a second user in a second target user group; wherein the first target user group and the second target user group are two adjacent groups, and the second user is any user in the second target user group who is of the same type as the first user, and the type includes an uplink user or a downlink user.

[0040] Optionally, iteratively learning the deep Q network model according to the relative training error until the IUI converges to a preset feedback value includes:

[0041] Adjusting parameters of the deep Q network model according to the relative training error;

[0042] Perform iterative learning based on the adjusted parameters;

[0043] When the state of the deep Q network model reaches the first state after iterative learning, the deep Q network model is retrained based on the grouping method corresponding to the minimum IUI until the IUI converges to the preset feedback value.

[0044] Optionally, the method further comprises:

[0045] When the current grouping method does not satisfy the constraint conditions of the optimization problem and the number of the second service users, a grouping method is randomly selected from the experience pool.

[0046] Optionally, grouping users in the user training set includes:

[0047] Constructing a 1×K vector corresponding to the users in the user training set; wherein the first M1 elements of the vector correspond to uplink users, and the elements after the M1th element in the vector correspond to downlink users, and the user training set includes M1 uplink users;

[0048] The users in the user training set are grouped according to the number of the second service users and the vector.

[0049] In a second aspect, in order to achieve the above-mentioned purpose, an embodiment of the present application provides a user scheduling device, including:

[0050] A first determination module is used to determine the transmission error probability of each downlink user to be served when a limited block length is used for data transmission;

[0051] A second determination module is used to determine the first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes the first number of uplink users and the first number of downlink users scheduled in a time-frequency domain resource;

[0052] The grouping module is used to group the user set to be served according to the first number of service users and the transmission error probability, wherein the user set to be served includes uplink users to be served and downlink users to be served.

[0053] In a third aspect, to achieve the above-mentioned purpose, an embodiment of the present application provides a user scheduling device, including a processor and a transceiver; the transceiver receives and sends data under the control of the processor, and the processor is used to:

[0054] Determining a transmission error probability for each downlink user to be served when a finite block length is used for data transmission;

[0055] Determine a first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes a first number of uplink users and a first number of downlink users scheduled in a time-frequency domain resource;

[0056] According to the first number of service users and the transmission error probability, a user set to be served is grouped into users, wherein the user set to be served includes uplink users to be served and downlink users to be served.

[0057] In the fourth aspect, in order to achieve the above-mentioned purpose, an embodiment of the present application provides a user scheduling device, including a transceiver, a processor, a memory, and a program stored in the memory and executable on the processor; when the processor executes the program, the user scheduling method described in the first aspect is implemented.

[0058] In a fifth aspect, in order to achieve the above-mentioned purpose, an embodiment of the present application provides a readable storage medium on which a program or instruction is stored, and when the program or instruction is executed by a processor, the user scheduling method as described in the first aspect is implemented.

[0059] The beneficial effects of the above technical solution of the present application are as follows:

[0060] The user scheduling method of the embodiment of the present application first determines the transmission error probability of each downlink user to be served when a finite block length is used for data transmission; secondly, determines the first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes the first number of uplink users and the first number of downlink users scheduled within a time-frequency domain resource; finally, according to the first number of service users and the transmission error probability, groups the user set to be served, wherein the user set to be served includes the uplink users to be served and the downlink users to be served. The embodiment of the present application determines the first number of uplink users and the first number of downlink users scheduled within a time-frequency domain resource based on the transmission error probability, provides a basis for selecting the number of service users, and can effectively reduce the transmission error probability of the system, thereby meeting service requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A flowchart of a user scheduling method according to an embodiment of the present application;

[0062] Figure 2 A schematic diagram of the structure of a user scheduling device according to an embodiment of the present application;

[0063] Figure 3 This is one of the structural schematic diagrams of the user scheduling device in an embodiment of the present application;

[0064] Figure 4 This is the second structural diagram of the user scheduling device according to an embodiment of the present application. DETAILED DESCRIPTION

[0065] In order to make the technical problems, technical solutions and advantages to be solved by the present application clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0066] It should be understood that the references to "one embodiment" or "an embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present application. Therefore, the references to "in one embodiment" or "in an embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0067] In the various embodiments of the present application, it should be understood that the size of the serial numbers of the following processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0068] Additionally, the terms "system" and "network" are often used interchangeably herein.

[0069] In the embodiments provided in the present application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined according to A. However, it should also be understood that determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information.

[0070] Before describing the embodiments of the present application, the relevant technical points are first described:

[0071] The Network-Assisted Full-Duplex (NAFD) system based on the non-cellular architecture implements the uplink and downlink service mode selection technology in the spatial domain within the same time-frequency resources. Its main working principle includes: the uplink and downlink links are carried out simultaneously on the same time-frequency resources. In such a system, multiple access points (APs) are equipped with a large number of antennas and densely distributed in the area. These APs are connected to the same central processing unit, which is responsible for baseband processing. Each AP has two service modes: uplink transmission and downlink transmission. In each time slot, each AP can independently select the uplink and downlink service mode. In this way, uplink and downlink services can be carried out simultaneously in the same time-frequency resource block. Compared with the traditional full-duplex system, there is no self-interference inside the AP in the NAFD system based on the non-cellular architecture, which can reduce the complexity of the transceiver. In addition, the flexible adjustment of the number of uplink APs and downlink APs can effectively cope with the increasingly serious uplink and downlink transmission asymmetry problem in the 5G mobile communication system, reduce latency, and improve transmission reliability. NAFD has become a key technology to support massive terminal mURLLC.

[0072] Based on this, existing technical solutions realize that the CLI between users can be reduced by grouping and scheduling the massive number of users waiting for service and allocating different time and frequency resources. The published user scheduling technology searches for the best grouping scheme through a genetic algorithm (GA) to solve the problem of minimizing the total IUI. The total user set is divided into multiple subsets, and the IUI within each subset under the corresponding grouping conditions can be calculated, where the total IUI is defined as the sum of the IUI of each subset, so the IUI can be calculated for each grouping. Through the GA technology, elites are continuously selected from the parent population for inheritance, hybridization and mutation, and a grouping scheme that can minimize the system IUI can be found. However, the above user scheduling scheme still has the following shortcomings:

[0073] (1) Reasonable user scheduling should be divided into two steps: the first step is to determine the number of scheduled users in each group; the second step is to determine the user scheduling plan. The user scheduling plan in the existing NAFD system skips the first step and does not explore the instantaneous optimal number of scheduled users in combination with system performance, so it cannot be applied to the actual system.

[0074] (2) Existing user scheduling schemes do not consider the high reliability requirements of mURLLC scenarios and cannot meet user QoS requirements or support URLLC traffic.

[0075] (3) The existing user scheduling solution is solved by genetic algorithm, but genetic algorithm has many shortcomings in use. First, when there are many variables, a large range of values, or no given range, the convergence speed decreases; second, it can find the vicinity of the optimal solution, but it cannot accurately determine the location of the optimal solution and may fall into the local optimum.

[0076] In view of the above-mentioned problems existing in the existing user scheduling method in the scenario of massive terminal mURLLC, the embodiment of the present application provides a user scheduling method, apparatus, device and readable storage medium. The implementation process of the embodiment of the present application is described in detail below in conjunction with the accompanying drawings.

[0077] like Figure 1 As shown, the embodiment of the present application provides a user scheduling method, including:

[0078] Step 101, when a finite block length (FBL) is used for data transmission, the transmission error probability of each downlink user to be served is determined; here, the downlink user to be served is a downlink user that needs to be served by a NAFD system based on a non-cellular network; wherein the finite block length can also be called a coding block length.

[0079] Step 102, determining the first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes the first number of uplink users and the first number of downlink users scheduled within a time-frequency domain resource; here, the first number of service users is the optimal number of scheduled users in the NAFD system based on the non-cellular architecture that maximizes the reliable transmission rate.

[0080] Step 103: grouping the user set to be served according to the number of first service users and the transmission error probability, wherein the user set to be served includes uplink users to be served and downlink users to be served.

[0081] In the embodiment of the present application, first, in the case of limited block length data transmission, the transmission error probability of each downlink user to be served is determined; secondly, the number of first service users is determined according to the transmission error probability, the limited block length and the transmitter transmission rate; and thirdly, the user set to be served is grouped according to the first number of service users and the transmission error probability. In this way, on the one hand, it is realized to determine the first number of uplink users and the first number of downlink users scheduled in a time-frequency domain resource based on the transmission error probability, which provides a basis for the selection of the number of service users. In this way, the high reliability requirements of the mURLLC scenario are taken into account in the user scheduling process, which can effectively reduce the transmission error probability of the system, thereby meeting service requirements or supporting URLLC traffic; on the other hand, a reasonable user scheduling process is realized, that is, firstly, the number of scheduled users for each group is determined, and then the user scheduling plan is determined.

[0082] As an optional implementation, step 101, in the case of limited block length data transmission, determines the transmission error probability of each downlink user to be served, including:

[0083] For a first downlink user, determining a first relationship between a transmitter transmission rate of the first downlink user, a transmission error probability of the first downlink user, and a finite block length; wherein the first downlink user is any downlink user among the downlink users to be served;

[0084] According to the first relational expression, the transmission error probability of the first downlink user is determined by using the random block length equivalent capacity and the first-order Taylor approximation.

[0085] Here, it should be noted that the process of determining the transmission error probability is an analysis of the reliability of the NAFD system based on a non-cellular architecture (also known as a non-cellular network-assisted full-duplex mURLLC system). In order to achieve the above reliability analysis, it is necessary to introduce the auxiliary variables shown in Table 1 below in the embodiment of the present application. The process of performing system reliability analysis based on these auxiliary variables will be described in detail later.

[0086] Table 1 Auxiliary variable identification and meaning

[0087] As a specific implementation method, according to the first relation, using the random block length equivalent capacity and the first-order Taylor approximation, the transmission error probability of the first downlink user is determined, including:

[0088] According to the formula:

[0089]

[0090] Determine the transmission error probability of the first downlink user; here, the first downlink user may be defined as user k involved in the above Table 1;

[0091] Among them, R k (·) represents the transmitter transmission rate of the first downlink user, n represents the finite block length, γ k represents the signal-to-interference-noise ratio of the first downlink user, ε k represents the transmission error probability of the first downlink user, represents the effective signal power of the first downlink user, γ R represents the channel capacity corresponding to the transmitter rate, σ IF,k represents the variance of the channel estimation error of the first downlink user and the finite block length error, λ c,k represents the total interference signal power received by the first downlink user, μ IF,k Represents the mean of the channel estimation error and the finite block length error of the first downlink user.

[0092] Here, it should be noted that ε k The specific calculation process is:

[0093]

[0094] Among them, Indicates the downlink noise transmission power, which is a preset value;

[0095] Among them, is the aforementioned first relationship; that is, the aforementioned first relationship can be understood as the relationship between the transmitter rate and the finite block length, the transmission error probability and the signal to interference and noise ratio.

[0096] In the above formula In order to equate the first relational expression based on the concept of equivalent capacity of random block length, and to obtain the formula by first-order Taylor approximation of the equivalent formula, the specific equivalence process and approximation process can be implemented based on the existing method and will not be described in detail here.

[0097] In addition, Q(·) in the above formula represents the right tail function of the standard normal distribution, and Γ(·) represents the gamma function.

[0098] In this specific implementation method, by calculating the transmission error probability of each downlink user, it can provide a basis for subsequently determining the optimal number of service users (the first number of service users), so as to effectively reduce the transmission error probability of the system, thereby meeting the high reliability requirements and user service quality (Quality of Service, QoS) requirements of the mURLLC scenario.

[0099] Here, it should be noted that the advantage of a smaller number of service users is that it can reduce collisions during transmission, reduce pilot length so that more time slots are used for data transmission, and reduce CLI between users, thereby reducing the probability of transmission errors; the advantage of a larger number of service users is that it can serve all users in a shorter time, allowing more retransmissions in the same time, which may also reduce the probability of transmission errors. Therefore, there must be an appropriate number of service users that can minimize the probability of transmission errors.

[0100] Based on this, an optional implementation in the embodiment of the present application is: Step 102, determining the number of first service users according to the transmission error probability, the finite block length and the transmitter transmission rate, includes:

[0101] According to the transmission error probability of each downlink user to be served, the average service rate of the downlink users to be served is determined; here, the average service rate can be expressed as: E[S] = n×R×E[1-ε k ]; where E[S] represents the average service rate, n represents the finite block length, R represents the transmitter transmission rate, ε k represents the transmission error probability of user k; where E[·] is the expected function.

[0102] Under the constraint that the ratio of the total number of uplink users in a time domain resource is greater than or equal to a preset ratio, and with the average service rate reaching a maximum value as a goal, the number of uplink users and the number of downlink users are traversed to solve the first number of service users. Here, since the number of service users is a discrete integer and the range is not large, the number of users that minimizes the overall transmission error probability can be obtained by traversing.

[0103] In short, the implementation process of this optional implementation is as follows:

[0104] First, establish the problem of maximizing the reliable transmission rate. In a full-duplex mURLLC system without cellular network assistance, the reliability index is affected by the number of scheduled users. In order to optimize the system performance, it is necessary to explore the optimal number of service users within a time-frequency resource. In order to reflect the impact of transmission reliability, as mentioned above, the average service rate E[S] = n×R×E[1-ε k ], and solve the optimal number of scheduled users (the number of first service users) with the goal of maximizing E[·]. The optimization problem can be described as:

[0105]

[0106] ikB u ≥ξ u K

[0107] Among them, ξ u represents the minimum proportion of uplink users, K u Indicates the number of uplink users, K d represents the number of downlink users, and K represents the total number of users.

[0108] Second, the optimization problem is K u and K d By traversing and solving, we can obtain the optimal number of service users within a time-frequency resource of the system. In this way, a system only needs to be solved once, and its complexity can be ignored.

[0109] As an optional implementation, step 103, grouping users according to the number of first service users and the transmission error probability, includes:

[0110] The number of first service users, the transmission error probability of each downlink user to be served, and the user set to be served are input into a pre-trained deep Q network model, the user set to be served is grouped, and a grouping result is obtained.

[0111] In this optional implementation, a deep Q network model is used for grouping, which has the characteristics of low complexity and can be effectively applied to the mURLLC scenario with massive terminals, solving many shortcomings of the existing user scheduling scheme using the aforementioned genetic algorithm for grouping.

[0112] Further, as an optional implementation, the method further includes: training the deep Q network model based on the following steps:

[0113] (1) Establishing an optimization problem, wherein the optimization objective of the optimization problem is to minimize the inter-user interference IUI, the optimization variable of the optimization problem is the grouping method, and the constraints of the optimization problem include: the transmission error probability of the downlink user is less than or equal to the transmission error probability threshold and each downlink user to be served or each uplink user to be served is in only one user group;

[0114] Here, it should be noted that the goal of the user grouping algorithm in the embodiment of the present application is to find the optimal user grouping method to minimize IUI. Compared with the existing method, the advancedness lies in taking the reliability QoS requirement (i.e., the transmission error probability) into consideration as a constraint. Therefore, the training process of the deep Q network model that solves the grouping problem should also take the transmission error probability as a constraint. In this way, the optimization problem established in this step can be described as:

[0115]

[0116]

[0117]

[0118] 0≤ε k ≤ε max ,k∈[1,K d ]

[0119] Where Δ represents the grouping method, for example: Δ={A1,…,A M}, is defined in the total user set K tot The user families on the (user training set), M represents the number of groups, K u Indicates the number of uplink users in each group, K d represents the number of downlink users in each group, ζ m,i,k represents the symbolic function, γ m,i,k A represents the signal-to-interference-noise ratio between the i-th uplink user and the k-th downlink user in the m-th group of users. m represents the mth group of users, A n represents the nth group of users, K tot represents the total number of users, ε max represents the transmission error rate threshold, for example, ε max For 10 -3 .

[0120] (2) Determine the number of second service users corresponding to the user training set, wherein the number of second service users includes the number of second uplink users and the number of second downlink users scheduled within a time-frequency domain resource; in this step, the number of second service users can be determined in the same manner as the first service user number.

[0121] (3) Grouping the users in the user training set; for example, the users in the user training set may be randomly grouped.

[0122] (4) Iteratively train the deep Q network model according to the current grouping method, optimization problem, and the number of users of the second service. Here, the parameters of the deep Q network model can be iteratively adjusted through iterative training, so that the deep Q network model tends to be stable, and the trained deep Q network model can output the optimal user grouping solution.

[0123] In this optional implementation, during the training of the deep Q network model, value feedback is obtained based on the grouping action and the current environmental state (the system state corresponding to the grouping action); further, the optimal selection strategy is derived based on the accumulated rewards and the current environmental state. In successive training, the value network and decision network in the deep Q network are gradually trained to achieve an overall better strategy choice.

[0124] As an optional implementation method, the deep Q network model is iteratively trained according to the current grouping method, optimization problem and the number of second service users, including:

[0125] Calculate the inter-user interference IUI corresponding to the current grouping mode; wherein IUI is related to the signal to interference plus noise ratio (SINR) between users;

[0126] When the current grouping method satisfies the constraints of the optimization problem and the number of second service users, the users in the user training set are regrouped according to the current grouping method, and the process returns to the step of calculating the IUI corresponding to the current grouping method; that is, the user training set is continuously grouped by self-learning, and the IUI corresponding to each grouping method is calculated.

[0127] After obtaining the IUIs corresponding to the multiple grouping methods, the relative training error of the deep Q network model is calculated according to the minimum IUI; here, after the number of groupings reaches a preset number, the grouping method corresponding to the minimum IUI is selected to calculate the relative training error;

[0128] According to the relative training error, the deep Q network model is iteratively learned until the IUI converges to a preset feedback value.

[0129] As a specific implementation method, calculating the inter-user interference IUI corresponding to the current grouping method includes:

[0130] (1) For each user group, the signal-to-interference-and-noise ratio between each downlink user and each uplink user is calculated. In this step, the signal-to-interference-and-noise ratio between the downlink user and the uplink user can be expressed by the following formula:

[0131]

[0132] Among them, γ m,i,k represents the signal-to-interference-noise ratio between the i-th uplink user and the k-th downlink user in the m-th group of users, p u,m,i represents the transmission power of the i-th uplink user in the m-th group, g m,i,k represents the channel between the i-th uplink user and the k-th downlink user in the m-th group, ρ j represents the transmission power of the jth downlink user, represents the channel vector from the kth downlink user to all downlink base stations, w j represents the j-th downlink user precoding matrix, Indicates the downlink transmission noise power.

[0133] (2) for each of the user groups, calculating a sum of target signal to interference plus noise ratios; wherein the target signal to interference plus noise ratio is less than or equal to a transmission signal to noise ratio of an uplink user of the user group;

[0134] Here, it should be noted that according to the principle of continuous interference cancellation, when the signal-to-interference-noise ratio γ between the i-th uplink user and the k-th downlink user in the m-th group of users is m,i,k Greater than the transmission signal-to-noise ratio γ of the uplink user u,m,i When m,i,k can be completely eliminated. At this time, the inter-user signal to noise ratio will not cause interference to the user. That is to say, only the inter-user signal to noise ratio that is less than or equal to the transmission signal to noise ratio will contribute to the inter-user interference. Therefore, this step only calculates the inter-user signal to noise ratio that is less than or equal to the transmission signal to noise ratio.

[0135] Based on the above content, this step can be expressed by the following formula: Among them, when γ m,i,j >γ u,m,i When m,i,j is 0; when γ m,i,j ≤γ u,m,i When m,i,j is 1; where the transmission signal-to-noise ratio γ u,m,i It can be expressed as follows:

[0136]

[0137] Among them, p u,m,i represents the transmission power of the i-th uplink user in the m-th group, g u,m,i is the channel vector from the i-th uplink user in the m-th group to all uplink APs, v m,i is the corresponding receiver vector. is the interference channel from each downlink AP to each uplink AP, w m,k is the precoding vector for the k-th downlink user in the m-th group.

[0138] (3) Determine the IUI based on the sum of multiple target signal-to-interference-and-noise ratios of each user group. In this step, the sum of the target signal-to-interference-and-noise ratios in each user group is accumulated to obtain the IUI.

[0139] As another specific implementation, users in the user training set are regrouped according to the current grouping method, including:

[0140] Select the first user whose user group needs to be adjusted; in this step, the first user can be randomly selected with a lower probability (ε=10%) according to the ε greedy strategy (i.e., randomly make a location modification request q), and the first user whose user group needs to be adjusted can be estimated with a higher probability (ε=90%) (i.e., estimate the specific modified user location based on the existing grouping method);

[0141] When the first target user group to which the first user belongs is a preset user group, the first user is randomly regrouped; here, the preset user group is the first user group or the last user group among all user groups. For example, the q position data is saturated (the top layer group, the uplink x q =4, downlink x q =3) or has reached the bottom (the lowest group x q =1), then choose to keep the current state and re-randomize the position.

[0142] In the case that the first target user group is not a preset user group, the first user is randomly adjusted with the second user in the second target user group; wherein the first target user group and the second target user group are two adjacent groups, and the second user is any user in the second target user group of the same type as the first user, which type includes an uplink user or a downlink user. In other words, if the group sequence number of the user at the modified position q is not saturated / bottomed out, the system will randomly select to increase x q +=1 (i.e., adjust the position of q to the target user group adjacent to and after the first target user group), or, reduce the group number x q - = 1 (i.e. adjust the q position to the target group adjacent to and before the first target user group)

[0143] `

[0144] User group), and select matching users (users of the same type) to reduce x accordingly p =x q +1,x p =x q (i.e., the matched users are the target users that are adjacent to the first target user group and located after the first target user group)

[0145] `

[0146] user group, adjust to the first target user group) or increase x p =x q -1,x p =x q (ie, the matched users are adjusted from the target user group adjacent to and before the first target user group to the first target user group) to maintain the user grouping constraint.

[0147] As another specific implementation method, the deep Q network model is iteratively learned according to the relative training error until the IUI converges to a preset feedback value, including:

[0148] Adjust the parameters of the deep Q network model based on the relative training error;

[0149] Iterative learning is performed based on the adjusted parameters; that is, after adjusting the parameters of the deep Q network model, multiple user groupings are performed based on the user training set, the IUI corresponding to each grouping method is calculated, and the relative training error of the deep Q network model is calculated based on the minimum IUI;

[0150] After iterative learning, when the state of the deep Q network model reaches the first state, the deep Q network model is retrained based on the grouping method corresponding to the minimum IUI until the IUI converges to the preset feedback value. Here, the first state is, for example, a state in which the deep Q network model tends to be stable, and the minimum IUI in this step is, for example, the smallest of the multiple minimum IUIs obtained in the above steps.

[0151] That is to say, after each user switching action (the action of the first user adjusting the user group), the IUI corresponding to the grouping method after the user switching is calculated. After obtaining the IUI, the new state information (channel state information, such as the values ​​of the aforementioned auxiliary parameters, transmission error probability, etc.) is merged, and the current state information (the state information corresponding to the currently obtained IUI), IUI and current action are recorded in the experience pool, and the minimum batch samples are randomly transmitted from the experience pool to calculate the relative training error of the model in the training round, and the corresponding loss function value is obtained to guide the corresponding neural network parameter training. After multiple rounds of learning iterations, the model training tends to be stable. Select the optimal allocation strategy bias to reset the current state (the state information corresponding to the minimum IUI), and train again, and repeat multiple rounds until the grouping scheme converges to the optimal feedback value.

[0152] Furthermore, as an optional implementation, the method further includes:

[0153] When the current grouping method does not meet the constraints of the optimization problem and the number of second service users, a grouping method is randomly selected from the experience pool. That is, when the grouping method does not meet the above constraints, the optimal grouping method and corresponding state information can be randomly selected as the current grouping method based on the grouping methods recorded in the experience pool.

[0154] As a specific implementation method, users in the user training set are grouped, including:

[0155] Constructing a 1×K vector corresponding to the users in the user training set; wherein the first M1 elements of the vector correspond to uplink users, and the elements after the M1th element in the vector correspond to downlink users, and the user training set includes M1 uplink users;

[0156] This step is to model the users in the user training set as a 1×K (total number of uplink and downlink users) vector s t=[x1,x2,L,x M ], where the first M1 variables correspond to uplink users, with a value range of {1,2,3,4,…U1}, indicating that each uplink user will be uniquely assigned to any group from 1 to U1, and the last M2 variables correspond to downlink users, with a value range of {1,2,3,4,…U2}, indicating that each downlink user will be uniquely assigned to any group from 1 to U2. It is worth noting that the number of values ​​of each number in the first M1 variables is uniquely determined, and the number of values ​​of each number in the last M2 variables is also uniquely determined. The introduction of this constraint greatly compresses the agent's searchable space, ensures the effectiveness of the agent's decision-making during the algorithm convergence process, and reduces error accumulation to accelerate network convergence.

[0157] According to the number of second service users and the vector, the users in the user training set are grouped. This step is to group the users in the user training set based on the vector, and the number of users in each user group meets the requirement of the number of second service users.

[0158] Here, it should be noted that a NAFD massive MIMO system is considered, which includes L u Uplink AP and L d There are a large number of uplink and downlink users waiting for service in the area. All users are half-duplex single antenna users. The total number of uplink users is K. tot,u , the total number of downlink users is K tot,d Since it is impossible to provide services to a large number of users at the same time, a service can only be provided to a subset of all users. Assume that in this subset, the number of uplink users is K. u , number of downlink users K d . All L u The uplink APs are combined into K u Uplink UE services, all L d The number of downstream APs combined is K d downlink UE services.

[0159] Among them, in order to select the best number of service users, in the embodiment of the present application, the following qualitative analysis is first performed: the advantage of a smaller number of service users is that it can reduce collisions during transmission, reduce the pilot length so that more time slots can be used for data transmission, and reduce the CLI between users, thereby reducing the probability of transmission errors; the advantage of a larger number of service users is that it can serve all users in a shorter time, allowing more retransmissions in the same time, which may also reduce the probability of transmission errors. Therefore, there must be a suitable number of service users that can minimize the probability of transmission errors. In order to explore this value, it is necessary to model the transmission error probability of the system under the conditions of finite block length and non-ideal channel state information. Since the number of service users is a discrete integer and the range is not large, the number of users that minimizes the overall probability of transmission errors can be obtained by traversal.

[0160] The purpose of user grouping is to minimize the CLI between users, so the grouping is obtained by solving the optimization problem of minimizing IUI. The DQN algorithm is innovatively used to solve the user scheduling problem of minimizing IUI in this scenario. Compared with existing technical solutions, this algorithm solves the problem that GA's inefficient exploration may fall into local optimality, and uses internal driving force to improve the exploration ability of the intelligent agent.

[0161] That is to say, the embodiment of the present application proposes a massive service user scheduling scheme in a cellular-free NAFD system in a massive indoor terminal mURLLC scenario. The proposed user scheduling scheme determines the optimal number of service users by analyzing the transmission error probability of the system, and then obtains the user scheduling method according to the system settings by training the deep Q network to minimize the interference between users that meet the URLLC service requirements.

[0162] like Figure 2 As shown, an embodiment of the present application provides a user scheduling device, including:

[0163] A first determination module 201 is used to determine the transmission error probability of each downlink user to be served when a limited block length is used for data transmission;

[0164] A second determination module 202 is used to determine a first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes a first number of uplink users and a first number of downlink users scheduled in a time-frequency domain resource;

[0165] The grouping module 203 is configured to group the user set to be served according to the first number of service users and the transmission error probability, wherein the user set to be served includes uplink users to be served and downlink users to be served.

[0166] Optionally, the first determining module 201 includes:

[0167] A first determination submodule, configured to determine, for a first downlink user, a first relationship between a transmitter transmission rate of the first downlink user and a transmission error probability of the first downlink user;

[0168] A second determination submodule is used to determine the transmission error probability of the first downlink user according to the first relationship using a random block length equivalent capacity and a first-order Taylor approximation;

[0169] The first downlink user is any downlink user among the downlink users to be served.

[0170] Optionally, the second determining submodule is specifically configured to:

[0171] According to the formula:

[0172]

[0173] determining a transmission error probability of the first downlink user;

[0174] Among them, R k (·) represents the transmitter transmission rate of the first downlink user, n represents the finite block length, γ k represents the signal-to-interference-noise ratio of the first downlink user, ε k represents the transmission error probability of the first downlink user, represents the effective signal power of the first downlink user, γ R represents the channel capacity corresponding to the transmitter rate, σ IF,k represents the variance of the channel estimation error of the first downlink user and the finite block length error, λ c,k represents the total interference signal power received by the first downlink user, μ IF,k Represents the mean of the channel estimation error and the finite block length error of the first downlink user.

[0175] Optionally, the second determining module 202 includes:

[0176] A third determining submodule, configured to determine an average service rate of the downlink users to be served according to the transmission error probability of each of the downlink users to be served;

[0177] The fourth determination submodule is used to traverse the number of uplink users and the number of downlink users to solve the first service user number with the goal of achieving the maximum value of the average service rate when the constraint condition is that the ratio of the total number of uplink users in a time domain resource is greater than or equal to a preset ratio.

[0178] Optionally, the grouping module 203 is specifically used to input the first number of service users, the transmission error probability of each downlink user to be served, and the user set to be served into a pre-trained deep Q network model, group the user set to be served, and obtain a grouping result.

[0179] Optionally, the device further includes a training module for training the deep Q network model based on the following steps:

[0180] Establishing an optimization problem, wherein the optimization objective of the optimization problem is to minimize the inter-user interference IUI, the optimization variable of the optimization problem is the grouping mode, and the constraints of the optimization problem include: the transmission error probability of the downlink user is less than or equal to the transmission error probability threshold and each of the downlink users to be served or each of the uplink users to be served is located in only one user group;

[0181] Determine the number of second service users corresponding to the user training set, wherein the number of second service users includes the number of second uplink users and the number of second downlink users scheduled in one time-frequency domain resource;

[0182] Grouping users in the user training set;

[0183] The deep Q network model is iteratively trained according to the current grouping method, the optimization problem and the number of the second service users.

[0184] Optionally, when the training module is used to iteratively train the deep Q network model according to the current grouping mode, the optimization problem and the second number of service users, it is specifically used to:

[0185] Calculate the inter-user interference IUI corresponding to the current grouping mode;

[0186] When the current grouping method satisfies the constraint conditions of the optimization problem and the number of users of the second service, regroup the users in the user training set according to the current grouping method, and return to the step of calculating the IUI corresponding to the current grouping method;

[0187] After obtaining the IUIs corresponding to the multiple grouping modes, calculating the relative training error of the deep Q network model according to the minimum IUI;

[0188] According to the relative training error, the deep Q network model is iteratively learned until the IUI converges to a preset feedback value.

[0189] Optionally, when the training module is used to calculate the inter-user interference IUI corresponding to the current grouping mode, it is specifically used to:

[0190] For each user group, calculating the signal to interference and noise ratio between each downlink user and each uplink user;

[0191] For each of the user groups, a sum of target signal to interference plus noise ratios is calculated; wherein the target signal to interference plus noise ratio is less than or equal to a transmission signal to noise ratio of an uplink user of the user group;

[0192] The IUI is determined according to the sum of the target signal to interference and noise ratios of the user groups.

[0193] Optionally, when the training module is used to regroup the users in the user training set according to the current grouping method, it is specifically used to:

[0194] Select the first user whose user group needs to be adjusted;

[0195] When the first target user group to which the first user belongs is a preset user group, randomly regrouping the first users;

[0196] When the first target user group is not the preset user group, the first user is randomly adjusted with a second user in a second target user group; wherein the first target user group and the second target user group are two adjacent groups, and the second user is any user in the second target user group who is of the same type as the first user, and the type includes an uplink user or a downlink user.

[0197] Optionally, when the training module is used to iteratively learn the deep Q network model according to the relative training error until the IUI converges to a preset feedback value, it is specifically used to:

[0198] Adjusting parameters of the deep Q network model according to the relative training error;

[0199] Perform iterative learning based on the adjusted parameters;

[0200] When the state of the deep Q network model reaches the first state after iterative learning, the deep Q network model is retrained based on the grouping method corresponding to the minimum IUI until the IUI converges to the preset feedback value.

[0201] Optionally, when the training module is used to iteratively train the deep Q network model according to the current grouping mode, the optimization problem and the second service user number, it is specifically used to:

[0202] When the current grouping method does not satisfy the constraint conditions of the optimization problem and the number of the second service users, a grouping method is randomly selected from the experience pool.

[0203] Optionally, when the training module is used to group users in the user training set, it is specifically used to:

[0204] Constructing a 1×K vector corresponding to the users in the user training set; wherein the first M1 elements of the vector correspond to uplink users, and the elements after the M1th element in the vector correspond to downlink users, and the user training set includes M1 uplink users;

[0205] The users in the user training set are grouped according to the number of the second service users and the vector.

[0206] It should be noted here that the above-mentioned user scheduling device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned user scheduling method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0207] like Figure 3 As shown, the embodiment of the present application further provides a user scheduling device, including a processor 310 and a transceiver 320; the transceiver 320 receives and sends data under the control of the processor 310, and the processor 310 is used to:

[0208] Determining a transmission error probability for each downlink user to be served when a finite block length is used for data transmission;

[0209] Determine a first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes a first number of uplink users and a first number of downlink users scheduled in a time-frequency domain resource;

[0210] According to the first number of service users and the transmission error probability, a user set to be served is grouped into users, wherein the user set to be served includes uplink users to be served and downlink users to be served.

[0211] Optionally, when the processor 310 is used to determine the transmission error probability of each downlink user to be served in case of limited block length data transmission, it is specifically used to:

[0212] For a first downlink user, determining a first relationship between a transmitter transmission rate of the first downlink user and a transmission error probability of the first downlink user;

[0213] Determine the transmission error probability of the first downlink user according to the first relationship using random block length equivalent capacity and first-order Taylor approximation;

[0214] The first downlink user is any downlink user among the downlink users to be served.

[0215] Optionally, when the processor 310 is used to determine the transmission error probability of the first downlink user according to the first relationship by using a random block length equivalent capacity and a first-order Taylor approximation, it is specifically used to:

[0216] According to the formula:

[0217]

[0218] determining a transmission error probability of the first downlink user;

[0219] Among them, R k (·) represents the transmitter transmission rate of the first downlink user, n represents the finite block length, γ k represents the signal-to-interference-noise ratio of the first downlink user, ε k represents the transmission error probability of the first downlink user, represents the effective signal power of the first downlink user, γ R represents the channel capacity corresponding to the transmitter rate, σ IF,k represents the variance of the channel estimation error of the first downlink user and the finite block length error, λ c,k represents the total interference signal power received by the first downlink user, μ IF,k Represents the mean of the channel estimation error and the finite block length error of the first downlink user.

[0220] Optionally, when the processor 310 is used to determine the number of first service users according to the transmission error probability, the finite block length, and the transmitter transmission rate, it is specifically used to:

[0221] Determining an average service rate of the downlink users to be served according to the transmission error probability of each of the downlink users to be served;

[0222] When the constraint condition is that the ratio of the total number of uplink users in a time domain resource is greater than or equal to a preset ratio, the number of uplink users and the number of downlink users are traversed with the goal of achieving the maximum value of the average service rate to solve the first number of service users.

[0223] Optionally, when the processor 310 is used to group users according to the number of first service users and the transmission error probability, it is specifically used to:

[0224] The number of the first service users, the transmission error probability of each downlink user to be served, and the user set to be served are input into a pre-trained deep Q network model, the user set to be served is grouped, and a grouping result is obtained.

[0225] Optionally, the processor 310 is further configured to: train the deep Q network model based on the following steps:

[0226] Establishing an optimization problem, wherein the optimization objective of the optimization problem is to minimize the inter-user interference IUI, the optimization variable of the optimization problem is the grouping mode, and the constraints of the optimization problem include: the transmission error probability of the downlink user is less than or equal to the transmission error probability threshold and each of the downlink users to be served or each of the uplink users to be served is located in only one user group;

[0227] Determine the number of second service users corresponding to the user training set, wherein the number of second service users includes the number of second uplink users and the number of second downlink users scheduled in one time-frequency domain resource;

[0228] Grouping users in the user training set;

[0229] The deep Q network model is iteratively trained according to the current grouping method, the optimization problem and the number of the second service users.

[0230] Optionally, when the processor 310 is used to iteratively train the deep Q network model according to the current grouping mode, the optimization problem and the second service user number, it is specifically used to:

[0231] Calculate the inter-user interference IUI corresponding to the current grouping mode;

[0232] When the current grouping method satisfies the constraint conditions of the optimization problem and the number of users of the second service, regroup the users in the user training set according to the current grouping method, and return to the step of calculating the IUI corresponding to the current grouping method;

[0233] After obtaining the IUIs corresponding to the multiple grouping modes, calculating the relative training error of the deep Q network model according to the minimum IUI;

[0234] According to the relative training error, the deep Q network model is iteratively learned until the IUI converges to a preset feedback value.

[0235] Optionally, when the processor 310 is used to calculate the inter-user interference IUI corresponding to the current grouping mode, it is specifically used to:

[0236] For each user group, calculating the signal to interference and noise ratio between each downlink user and each uplink user;

[0237] For each of the user groups, a sum of target signal to interference plus noise ratios is calculated; wherein the target signal to interference plus noise ratio is less than or equal to a transmission signal to noise ratio of an uplink user of the user group;

[0238] The IUI is determined according to the sum of the target signal to interference and noise ratios of the user groups.

[0239] Optionally, when the processor 310 is used to regroup the users in the user training set according to the current grouping mode, it is specifically used to:

[0240] Select the first user whose user group needs to be adjusted;

[0241] When the first target user group to which the first user belongs is a preset user group, randomly regrouping the first users;

[0242] When the first target user group is not the preset user group, the first user is randomly adjusted with a second user in a second target user group; wherein the first target user group and the second target user group are two adjacent groups, and the second user is any user in the second target user group who is of the same type as the first user, and the type includes an uplink user or a downlink user.

[0243] Optionally, when the processor 310 is used to iteratively learn the deep Q network model according to the relative training error until the IUI converges to a preset feedback value, it is specifically used to:

[0244] Adjusting parameters of the deep Q network model according to the relative training error;

[0245] Perform iterative learning based on the adjusted parameters;

[0246] When the state of the deep Q network model reaches the first state after iterative learning, the deep Q network model is retrained based on the grouping method corresponding to the minimum IUI until the IUI converges to the preset feedback value.

[0247] Optionally, when the processor 310 is used to iteratively train the deep Q network model according to the current grouping mode, the optimization problem and the second service user number, it is further used to:

[0248] When the current grouping method does not satisfy the constraint conditions of the optimization problem and the number of the second service users, a grouping method is randomly selected from the experience pool.

[0249] Optionally, when the processor 310 is used to group users in the user training set, it is specifically used to:

[0250] Constructing a 1×K vector corresponding to the users in the user training set; wherein the first M1 elements of the vector correspond to uplink users, and the elements after the M1th element in the vector correspond to downlink users, and the user training set includes M1 uplink users;

[0251] The users in the user training set are grouped according to the number of the second service users and the vector.

[0252] It should be noted that the above-mentioned user scheduling device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned user scheduling method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those of the method embodiment will not be described in detail here.

[0253] like Figure 4 As shown, an embodiment of the present application also provides a user scheduling device, including a transceiver 410, a processor 400, a memory 420, and a program or instruction stored in the memory 420 and executable on the processor 400; when the processor 400 executes the program or instruction, the above-mentioned user scheduling method is implemented.

[0254] The transceiver 410 is used to receive and send data under the control of the processor 400 .

[0255] Among them, Figure 4 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 400 and various circuits of memory represented by memory 420 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 410 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. For different devices, the user interface 430 may also be an interface capable of externally and internally connecting required devices, and the connected devices include but are not limited to a keypad, a display, a speaker, a microphone, a joystick, and the like.

[0256] The processor 400 is responsible for managing the bus architecture and general processing, and the memory 420 can store data used by the processor 400 when performing operations.

[0257] A readable storage medium of an embodiment of the present application stores a program or instruction thereon. When the program or instruction is executed by a processor, the steps in the user scheduling method described above are implemented and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0258] The processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0259] It should be further explained that the electronic devices described in this specification include but are not limited to smart phones, tablet computers, etc., and many of the functional components described are called modules in order to more particularly emphasize the independence of their implementation methods.

[0260] In the embodiment of the present application, module can be implemented with software so that it can be executed by various types of processors. For example, an executable code module of an identification can include one or more physical or logical blocks of computer instructions, for example, it can be constructed as an object, process or function. Nevertheless, the executable code of the identified module does not need to be physically located together, but can include different instructions stored in different positions, and when these instructions are logically combined together, it constitutes a module and realizes the specified purpose of the module.

[0261] In fact, executable code module can be a single instruction or many instructions, and can even be distributed on a plurality of different code segments, distributed among different programs, and distributed across a plurality of memory devices. Similarly, operating data can be identified in the module, and can be implemented and organized in the data structure of any appropriate type according to any appropriate form. The operating data can be collected as a single data set, or can be distributed in different locations (including on different storage devices), and can only be present on a system or network as an electronic signal at least in part.

[0262] When a module can be implemented by software, considering the level of existing hardware technology, a person skilled in the art can build a corresponding hardware circuit to implement the corresponding function of the module that can be implemented by software without considering the cost. The hardware circuit includes a conventional very large scale integration (VLSI) circuit or gate array and existing semiconductors such as logic chips, transistors, or other discrete components. The module can also be implemented by a programmable hardware device, such as a field programmable gate array, a programmable array logic, a programmable logic device, etc.

[0263] The above exemplary embodiments are described with reference to the accompanying drawings, and many different forms and embodiments are feasible without departing from the spirit and teachings of the present application. Therefore, the present application should not be constructed as a limitation of the exemplary embodiments proposed herein. More specifically, these exemplary embodiments are provided so that the present application will be perfect and complete, and the scope of the present application will be conveyed to those who are familiar with the technology. In these figures, the component sizes and relative sizes may be exaggerated for clarity. The terms used here are only based on the purpose of describing specific exemplary embodiments and are not intended to be limiting. As used herein, unless the text clearly indicates otherwise, the singular forms "one", "an" and "the" are intended to include these multiple forms. It will be further understood that the terms "comprising" and / or "including" when used in this specification indicate the presence of the features, integers, steps, operations, components and / or components, but do not exclude the presence or increase of one or more other features, integers, steps, operations, components, components and / or their groups. Unless otherwise indicated, when stated, a range of values ​​includes the upper and lower limits of that range and any subranges therebetween.

[0264] The above is a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A user scheduling method, characterized in that: include: Determining a transmission error probability for each downlink user to be served when a finite block length is used for data transmission; Determine a first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes a first number of uplink users and a first number of downlink users scheduled in a time-frequency domain resource; According to the first number of service users and the transmission error probability, a user set to be served is grouped into users, wherein the user set to be served includes uplink users to be served and downlink users to be served.

2. The method according to claim 1, characterized in that In the case of limited block length data transmission, determining the transmission error probability of each downlink user to be served includes: For a first downlink user, determining a first relationship between a transmitter transmission rate of the first downlink user and a transmission error probability of the first downlink user; Determine the transmission error probability of the first downlink user according to the first relationship using random block length equivalent capacity and first-order Taylor approximation; The first downlink user is any downlink user among the downlink users to be served.

3. The method according to claim 2, characterized in that Determining the transmission error probability of the first downlink user according to the first relationship by using the random block length equivalent capacity and the first-order Taylor approximation includes: According to the formula: determining a transmission error probability of the first downlink user; Among them, R k (·) represents the transmitter transmission rate of the first downlink user, n represents the finite block length, γ k represents the signal-to-interference-noise ratio of the first downlink user, ε k represents the transmission error probability of the first downlink user, represents the effective signal power of the first downlink user, γ R represents the channel capacity corresponding to the transmitter rate, σ IF,k represents the variance of the channel estimation error of the first downlink user and the finite block length error, λ c,k represents the total interference signal power received by the first downlink user, μ IF,k Represents the mean of the channel estimation error and the finite block length error of the first downlink user.

4. The method according to claim 1, characterized in that: Determining the number of first service users according to the transmission error probability, the finite block length and the transmitter transmission rate includes: Determining an average service rate of the downlink users to be served according to the transmission error probability of each of the downlink users to be served; When the constraint condition is that the ratio of the total number of uplink users in a time domain resource is greater than or equal to a preset ratio, the number of uplink users and the number of downlink users are traversed with the goal of achieving the maximum value of the average service rate to solve the first number of service users.

5. The method according to claim 1, characterized in that The method further comprises: grouping users according to the number of users of the first service and the transmission error probability, including: The number of the first service users, the transmission error probability of each downlink user to be served, and the user set to be served are input into a pre-trained deep Q network model, the user set to be served is grouped, and a grouping result is obtained.

6. The method according to claim 5, characterized in that: The method further comprises: The deep Q network model is trained based on the following steps: Establishing an optimization problem, wherein the optimization objective of the optimization problem is to minimize the inter-user interference IUI, the optimization variable of the optimization problem is the grouping mode, and the constraints of the optimization problem include: the transmission error probability of the downlink user is less than or equal to the transmission error probability threshold and each of the downlink users to be served or each of the uplink users to be served is located in only one user group; Determine the number of second service users corresponding to the user training set, wherein the number of second service users includes the number of second uplink users and the number of second downlink users scheduled in one time-frequency domain resource; Grouping users in the user training set; The deep Q network model is iteratively trained according to the current grouping method, the optimization problem and the number of the second service users.

7. The method according to claim 6, characterized in that According to the current grouping mode, the optimization problem and the number of users of the second service, the deep Q network model is iteratively trained, including: Calculate the inter-user interference IUI corresponding to the current grouping mode; When the current grouping method satisfies the constraint conditions of the optimization problem and the number of users of the second service, regroup the users in the user training set according to the current grouping method, and return to the step of calculating the IUI corresponding to the current grouping method; After obtaining the IUIs corresponding to the multiple grouping modes, calculating the relative training error of the deep Q network model according to the minimum IUI; According to the relative training error, the deep Q network model is iteratively learned until the IUI converges to a preset feedback value.

8. The method according to claim 7, characterized in that Calculating the inter-user interference IUI corresponding to the current grouping mode includes: For each user group, calculating the signal to interference and noise ratio between each downlink user and each uplink user; For each of the user groups, a sum of target signal to interference plus noise ratios is calculated; wherein the target signal to interference plus noise ratio is less than or equal to a transmission signal to noise ratio of an uplink user of the user group; The IUI is determined according to the sum of the target signal to interference and noise ratios of the user groups.

9. The method according to claim 7, characterized in that: Regrouping the users in the user training set according to the current grouping method includes: Select the first user whose user group needs to be adjusted; When the first target user group to which the first user belongs is a preset user group, randomly regrouping the first users; When the first target user group is not the preset user group, the first user is randomly adjusted with a second user in a second target user group; wherein the first target user group and the second target user group are two adjacent groups, and the second user is any user in the second target user group who is of the same type as the first user, and the type includes an uplink user or a downlink user.

10. The method according to claim 7, characterized in that According to the relative training error, iteratively learning the deep Q network model until the IUI converges to a preset feedback value, including: Adjusting parameters of the deep Q network model according to the relative training error; Perform iterative learning based on the adjusted parameters; When the state of the deep Q network model reaches the first state after iterative learning, the deep Q network model is retrained based on the grouping method corresponding to the minimum IUI until the IUI converges to the preset feedback value.

11. The method according to claim 7, characterized in that The method further comprises: When the current grouping method does not satisfy the constraint conditions of the optimization problem and the number of the second service users, a grouping method is randomly selected from the experience pool.

12. The method according to claim 6, characterized in that Grouping users in the user training set includes: Constructing a 1×K vector corresponding to the users in the user training set; wherein the first M1 elements of the vector correspond to uplink users, and the elements after the M1th element in the vector correspond to downlink users, and the user training set includes M1 uplink users; The users in the user training set are grouped according to the number of the second service users and the vector.

13. A user scheduling device, characterized in that: include: A first determination module is used to determine the transmission error probability of each downlink user to be served when a limited block length is used for data transmission; A second determination module is used to determine the first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes the first number of uplink users and the first number of downlink users scheduled in a time-frequency domain resource; The grouping module is used to group the user set to be served according to the first number of service users and the transmission error probability, wherein the user set to be served includes uplink users to be served and downlink users to be served.

14. A user scheduling device, comprising a processor and a transceiver; the transceiver receives and sends data under the control of the processor, characterized in that: The processor is used to: Determining a transmission error probability for each downlink user to be served when a finite block length is used for data transmission; Determine a first number of service users according to the transmission error probability, the finite block length and the transmitter transmission rate, wherein the first number of service users includes a first number of uplink users and a first number of downlink users scheduled in a time-frequency domain resource; According to the first number of service users and the transmission error probability, a user set to be served is grouped into users, wherein the user set to be served includes uplink users to be served and downlink users to be served.

15. A user scheduling device, comprising a transceiver, a processor, a memory, and a program stored in the memory and executable on the processor; characterized in that: When the processor executes the program, the user scheduling method according to any one of claims 1 to 12 is implemented.

16. A readable storage medium having a program or instruction stored thereon, characterized in that: When the program or instruction is executed by a processor, the user scheduling method according to any one of claims 1 to 12 is implemented.

Images