Spectrum renting method combining spectrum reservation and on-demand request
By combining early reservation and on-demand request joint spectrum reservation methods, the problem of insufficient or overreservation of resource reservation in radio spectrum virtualization is solved, and lower operating costs and higher spectrum resource utilization are achieved.
Patent Information
- Application Number
- CN202410836515.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-26
- Publication Date
- 2024-11-05
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the existing radio spectrum virtualization methods, both pre-order and on-demand requests have shortcomings, making it difficult to effectively predict and respond to real-time traffic demands, resulting in insufficient or over-reservation of resource reservations, increasing operational costs and complexity.
Using three sets of time scales, the combined spectrum reservation and on-demand request methods include advance reservation, on-demand request and dynamic resource allocation. By optimizing user traffic and on-demand prices, dynamically allocating sub-channels by using the combination of early reservation and on-demand requests. To maximize spectrum resource utilization.
It reduces the operating costs and risks of mobile virtual network operators, improves the flexibility and utilization of spectrum resources, adapts to changes in user traffic at different levels, and achieves higher surpluses.
Smart Images

Figure CN118900435A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of joint spectrum reservation and on-demand spectrum leasing solutions for mobile virtual network operators, and in particular to a joint spectrum reservation and on-demand spectrum leasing method. Background Art
[0002] Wireless network virtualization (or network slicing in 5G systems) is an emerging technology that provides unprecedented opportunities for mobile virtual network operators to develop new services by leasing infrastructure and radio resources to mobile network operators. Compared with the upper-layer network function virtualization, the virtualization of the physical layer radio spectrum is a unique problem in wireless network virtualization due to the broadcast and random nature of wireless channels. Unlike other network resources, the radio spectrum can be dynamically reused by different links based on the geographical distance between transmission nodes, transmission power, interference cancellation capabilities, and service quality requirements, so spectrum virtualization is more complicated.
[0003] Existing radio spectrum virtualization methods can be roughly divided into two categories: advance reservation and on-demand request. In advance reservation, each mobile virtual network operator reserves a certain amount of spectrum resources to a mobile network operator for a long period of time. Generally speaking, advance reservation is beneficial for mobile network operators to pre-plan resource slices and allocations. However, due to the uncertainty of traffic realization and wireless channel fading, under-reservation (over-reservation) may occur when the reserved resources are less (more) than the real-time demand. In wireless network virtualization, random wireless channel fading introduces a new dimension of uncertainty, making it difficult for mobile virtual network operators to predict their demand for spectrum resources in advance.
[0004] In contrast, on-demand requests retain the flexibility of MVNOs to order spectrum resources in real time based on observed traffic demand. However, the flexibility of on-demand requests comes with high operational costs of frequently calculating and implementing leasing decisions. In addition, there is no guarantee that MVNOs will provide sufficient spectrum in real time. To this end, we propose a spectrum leasing method that combines spectrum reservation and on-demand requests. Summary of the invention
[0005] The object of the present invention is to provide a spectrum leasing method combining spectrum reservation and on-demand request to solve the problems raised in the above background technology.
[0006] To achieve the above object, the present invention provides the following technical solution: a spectrum leasing method combining spectrum reservation and on-demand request, comprising:
[0007] It includes three groups of time scales, namely advance booking, on-demand request, and dynamic resource allocation;
[0008] Advance reservation: MVNOs reserve multiple sub-channels for a long period of time. This decision is optimized based on estimated statistics of user traffic and on-demand prices during that period, ensuring that MVNOs obtain a certain amount of spectrum resources at a relatively low cost, allowing MVNOs to pre-plan resource slicing and system operations at an early stage.
[0009] On-demand request: The mobile virtual network operator decides whether to request additional sub-channels after observing the actual location of the user who needs service and the current on-demand price. The on-demand request changes with the arrival, departure and movement of users, changes within a fixed time range, and does not change with fast fading channels. The mobile virtual network operator can obtain additional spectrum resources based on real-time traffic demand. The mobile network operator can determine the real-time competition intensity between multiple mobile virtual network operators;
[0010] Dynamic resource allocation,On a millisecond time scale, the MVNO dynamically allocates subchannels obtained from advance reservations and on-demand requests to users based on instantaneous wireless channel fading.
[0011] Preferably, advance reservation, on-demand request and dynamic resource allocation constitute a set of three-level nested optimization problems. The nested structure makes the three sub-problems closely intertwined. The decisions made on a large time scale will affect the optimization on a small time scale. The optimal value of the problem on a small time scale is embedded in the objective function of the problem on a large time scale.
[0012] Preferably, the optimal channel-aware dynamic resource allocation strategy is derived under the optimal utility function in advance reservation and on-demand request, the closed-form expression is used to apply the optimal number of rentable subchannels in the on-demand request stage, and the stochastic gradient descent algorithm is used to find the number of subchannels to be rented in the advance reservation stage. For the special case of using the proportional fairness utility function, the closed-form expression of the optimal number of rentable subchannels in the two stages is used.
[0013] Preferably, the internal data collection of the proportional fairness utility function includes using a broadband wireless receiver to collect signals, making a rough estimate of the symbol rate of the signals collected by the receiver, and setting the initial value S of the upper and lower cutoff frequencies of the bandpass filter, continuously optimizing the upper and lower cutoff frequencies of the bandpass filter based on the measurement signal and the data-driven method, and generating the optimal cutoff frequency, using the optimal upper and lower cutoff frequencies to generate the optimal filter processing signal complex envelope, extracting the symbol rate spectrum, and completing the symbol rate estimation.
[0014] Preferably, the utilization of sub-channels obtained from advance reservations and on-demand requests is maximized according to the fast channel fading in each time slot.
[0015] Preferably, analytical results for deriving optimal solutions to the on-demand request and advance reservation subproblems are applied to proportional fairness utility and demonstrate unique properties of the two-stage rental scheme that are not reflected in general utility functions.
[0016] Preferably, the performance of the derived two-stage leasing scheme is numerically evaluated.
[0017] The present invention has at least the following beneficial effects:
[0018] Existing radio spectrum virtualization methods can be roughly divided into two categories: advance reservation and on-demand request. In advance reservation, each mobile virtual network operator reserves a certain amount of spectrum resources to a mobile network operator for a long time. Generally speaking, advance reservation is conducive to mobile network operators to pre-plan resource slices and allocation. However, due to the uncertainty of traffic realization and wireless channel fading, when the reserved resources are less than or less than the real-time demand, insufficient or excessive reservation may occur. In wireless network virtualization, random wireless channel fading introduces a new dimension of uncertainty, making it difficult for mobile virtual network operators to predict their demand for spectrum resources in advance. Compared with the existing technology, the present invention proposes a two-stage spectrum leasing scheme, that is, mobile virtual network operators obtain spectrum from two ways of advance reservation and on-demand request. The two-stage leasing scheme has the characteristics of complementary advantages. In the first stage, advance reservation reduces the risks and operational complexity of mobile virtual network operators and mobile network operators. On the one hand, advance reservation ensures that mobile virtual network operators obtain a certain amount of spectrum resources at a relatively low cost. On the other hand, it allows mobile network operators to pre-plan resource slices and system operations at an early stage.
[0019] The second phase, on-demand request, retains flexibility and competitiveness. On the one hand, MVNOs can obtain additional spectrum resources based on real-time traffic demand, thus avoiding being too conservative in the first phase. On the other hand, MVNOs can make more profits by setting higher on-demand prices based on the real-time competition intensity among multiple MVNOs.
[0020] The present invention aims to find the optimal two-stage spectrum leasing scheme for mobile virtual network operators. Specifically, the system operation involves three time scales.
[0021] The optimal channel-aware dynamic resource allocation strategy is derived under a wide range of utility functions. Through rigorous analysis, the optimal utility function is described as a function of the total number of subchannels obtained from advance reservations and on-demand requests. Based on the results of dynamic resource allocation, the on-demand request and advance reservation subproblems of the general utility function are solved. A closed-form expression for the optimal number of subchannels that can be rented in the on-demand request phase is derived, and a stochastic gradient descent algorithm is also developed to find the number of subchannels to be rented in the advance reservation phase.
[0022] For the special case of adopting proportional fairness utility function, a closed-form expression for the optimal number of subchannels that can be leased in two stages is derived, that is, the mobile virtual network operator can find the optimal operation on all three time scales through analytical calculation. Numerical results show that the proposed two-stage leasing scheme can utilize on-demand price changes to reduce the operating costs of mobile virtual network operators, can adapt to changes in user traffic at different levels, and the two-stage spectrum leasing plan achieves more surplus than the traditional single-stage spectrum leasing plan. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 It is a schematic diagram of the operation flow of the system of the present invention;
[0024] Figure 2 It is a schematic diagram of the structure of the system model of the present invention;
[0025] Figure 3 This is a schematic diagram of the convergence process structure of Algorithm 1 of the present invention;
[0026] Figure 4 This is a schematic diagram of the structure of the on-demand price change effect of the present invention;
[0027] Figure 5 It is a schematic diagram of the average total number of sub-channels and the average rental cost structure of each sub-channel of the present invention;
[0028] Figure 6 This is a schematic diagram of the structure of the impact of flow changes in the present invention;
[0029] Figure 7 This is a schematic diagram of average surplus of the present invention;
[0030] Figure 8 It is a schematic diagram of the overall workflow structure of the present invention. DETAILED DESCRIPTION
[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0032] See also Figure 1-8 , the present invention provides a technical solution:
[0033] Embodiment 1,
[0034] like Figure 8 A spectrum leasing method combining spectrum reservation and on-demand request is shown, comprising:
[0035] It includes three groups of time scales, namely advance booking, on-demand request, and dynamic resource allocation;
[0036] Advance reservation: MVNOs reserve multiple sub-channels for a long period of time. This decision is optimized based on estimated statistics of user traffic and on-demand prices during that period, ensuring that MVNOs obtain a certain amount of spectrum resources at a relatively low cost, allowing MVNOs to pre-plan resource slicing and system operations at an early stage.
[0037] On-demand request: The mobile virtual network operator decides whether to request additional sub-channels after observing the actual location of the user who needs service and the current on-demand price. The on-demand request changes with the arrival, departure and movement of users, changes within a fixed time range, and does not change with fast fading channels. The mobile virtual network operator can obtain additional spectrum resources based on real-time traffic demand. The mobile network operator can determine the real-time competition intensity between multiple mobile virtual network operators;
[0038] Dynamic resource allocation,On a millisecond time scale, the MVNO dynamically allocates subchannels obtained from advance reservations and on-demand requests to users based on instantaneous wireless channel fading.
[0039] Advance reservation, on-demand request and dynamic resource allocation constitute a set of three-level nested optimization problems. The nested structure makes the three sub-problems closely intertwined. The decisions made on a large time scale will affect the optimization on a small time scale. The optimal value of the problem on a small time scale is embedded in the objective function of the problem on a large time scale.
[0040] The optimal channel-aware dynamic resource allocation strategy is derived under the optimal utility function in advance reservation and on-demand request. In the on-demand request stage, a closed-form expression is used to apply the optimal number of rentable subchannels. In the advance reservation stage, the stochastic gradient descent algorithm is used to find the number of subchannels to be rented. For the special case of using proportional fairness utility function, a closed-form expression for the optimal number of rentable subchannels in the two stages is used.
[0041] The internal data collection of the proportional fairness utility function includes using a broadband wireless receiver to collect signals, making a rough estimate of the symbol rate of the receiver collected signals, and setting the initial value S of the upper and lower cutoff frequencies of the bandpass filter. Based on the measurement signal and data-driven method, the upper and lower cutoff frequencies of the bandpass filter are continuously optimized, and the optimal cutoff frequency is generated. The optimal filter is used to process the signal complex envelope with the optimal upper and lower cutoff frequencies, and the symbol rate spectrum is extracted to complete the symbol rate estimation.
[0042] The utilization of subchannels obtained from advance reservations and on-demand requests is maximized according to the fast channel fading in each time slot.
[0043] The analytical results for deriving the optimal solutions to the on-demand request and advance reservation subproblems are applied to the proportional fairness utility and demonstrate the unique properties of the two-stage rental scheme that are not reflected in the general utility function.
[0044] The performance of the derived two-stage leasing scheme is numerically evaluated.
[0045] How it works
[0046] Compared with the traditional single-stage leasing scheme, i.e., a single advance reservation or a single on-demand request, the decision made in the present invention on a larger time scale will affect the optimization within a smaller time scale. Conversely, the optimal value of the smaller time scale problem is embedded in the objective function of the larger time scale problem. The smaller time scale problem is characterized analytically so that the larger time scale problem is suitable for an effective solution algorithm. A two-stage spectrum leasing scheme is proposed in a nested manner, i.e., the mobile virtual network operator obtains spectrum from both advance reservation and on-demand request. After careful optimization, the two-stage leasing scheme has the characteristics of complementary advantages. In the first stage, advance reservation reduces the risks and operational complexity of mobile virtual network operators and mobile network operators. On the one hand, advance reservation ensures that mobile virtual network operators obtain a certain amount of spectrum resources at a relatively low cost. On the other hand, it allows mobile network operators to pre-plan resource slices and system operations in the early stages. In the second stage, on-demand request, flexibility and competitiveness are retained. On the one hand, mobile virtual network operators can obtain additional spectrum resources according to real-time traffic demand, thereby avoiding being too conservative in the first stage. On the other hand, mobile network operators can obtain more profits by setting higher on-demand prices according to the real-time competition intensity among multiple mobile virtual network operators.
[0047] The present invention aims to find the optimal two-stage spectrum leasing scheme for mobile virtual network operators. Specifically, the system operation involves three time scales, such as Figure 1 shown.
[0048] In the first phase, advance booking, MVNOs reserve multiple subchannels for a long period of time, covering hours or days. The decision is optimized based on estimated statistics of user traffic and on-demand prices during that time period.
[0049] In the second stage, i.e. on-demand request, the MVNO decides whether to request additional sub-channels after observing the actual location of the user who needs service and the current on-demand price. On-demand request varies with the arrival, departure and movement of users, usually within a time frame of a few seconds. The on-demand request here does not vary with fast fading channels.
[0050] Then, on a millisecond time scale, the MVNO dynamically allocates subchannels obtained from advance reservations and on-demand requests to users based on instantaneous wireless channel fading. In order to maximize the profit of the MVNO, operations need to be optimized jointly on all three time scales.
[0051] The two-stage spectrum leasing problem is formulated as a three-level nested optimization problem, which consists of dynamic resource allocation, on-demand request, and advance reservation subproblems. Solving the problem is challenging in two aspects. The nested structure makes the three sub-optimization problems stochastic, as decisions on larger time scales must be optimized for stochastic network realizations on smaller time scales. As a first attempt, we attempt to solve the two-stage spectrum leasing problem. The invention assumes an α-fair utility function and constant on-demand prices. In this invention, the stochasticity of on-demand prices is considered and the analysis is extended to more general utility functions.
[0052] In this invention, we address the following challenges:
[0053] We derive optimal channel-aware dynamic resource allocation strategies under a wide range of utility functions. Through rigorous analysis, we describe the optimal utility function as a function of the total number of subchannels obtained from advance reservations and on-demand requests. Based on the results of dynamic resource allocation, we solve the on-demand request and advance reservation subproblems of the general utility function and derive a closed-form expression for the optimal number of subchannels that can be rented during the on-demand request phase. We develop a stochastic gradient descent algorithm to find the number of subchannels to rent during the advance reservation phase.
[0054] For the special case of adopting a proportional fairness utility function, a closed-form expression for the optimal number of subchannels that can be leased in both phases is derived. In other words, MVNOs can find the optimal operation on all three time scales by analytical calculation with negligible computational complexity.
[0055] Numerical results show that the proposed two-stage leasing scheme can reduce the operating costs of mobile virtual network operators by utilizing on-demand price changes to adapt to different levels of user traffic changes, and the two-stage spectrum leasing plan achieves more surplus than the traditional single-stage spectrum leasing plan.
[0056] Section 2. System Model and Problem Description
[0057] like Figure 2 As shown, the circular area of radius is represented by D(o,D). A group of users who simultaneously request services from a mobile virtual network operator is represented by a random point process Among them, each x k ∈D is the location of the kth user, and K∈Z + is the total number of users. We assume that each x k are uniformly distributed in the cell, and the distribution of K can be accurately estimated from historical data. The user set χ changes as users arrive, leave or move. In this way, the characteristics of traffic statistics are distributed as χ. It is assumed that both the base station and the user are equipped with a single antenna, and it is assumed that the transmitted signal is affected by large-scale path loss and fast Rayleigh fading.
[0058] The instantaneous data rate of user k on subchannel i is:
[0059]
[0060] Where B represents the bandwidth of each subchannel, P represents the fixed transmit power of the base station, l(.) represents the path loss function, ||.|| represents the Euclidean norm, gKi represents the power gain of independent and identically distributed Rayleigh fading, N o represents the power spectral density of Gaussian white noise at the receiver, and Γ represents the capacity surplus.
[0061] The two-stage spectrum leasing scheme involves operations on three time scales, such as Figure 3 As shown in Figure 2, a time period, a session, and a time slot are defined as three basic time units, in which traffic statistics (i.e., the distribution of χ), user locations (i.e., the realization of χ), and channel fading (i.e., ) remain unchanged.
[0062] Typically, time periods are measured in hours, sessions are measured in seconds, and time slots are measured in milliseconds.
[0063] At the beginning of a time period, the MVNO reserves a certain number of subchannels for the entire time period according to the distribution of χ. Then, at the beginning of each session, the MVNO decides whether to request more subchannels for the session based on the observed χ realization. The subchannels obtained from advance reservations and on-demand requests are dynamically allocated to users according to the fast fading channels in each time slot.
[0064] The profit maximization problem of mobile virtual network operators can be formulated as a three-layer nested optimization problem, which consists of the following three sub-problems.
[0065] Book in advance:
[0066] Let c r is the subscription price of the subchannel, n r is the number of reserved subchannels. The problem of finding the optimal advance reservation is formulated as:
[0067]
[0068] The first term in the objective function is the reservation cost, and the second term is the expected surplus per session. The expectation E contains all possible user realizations χ and the on-demand price c in the session. s , where Q(χ,c s ,n r ) represents the number of cells from the group containing χ, c s and n rThe surplus obtained in the session of , and will be explicitly defined in the on-demand reservation subproblem, assuming that the reserved subchannel cannot be released back to the MNO before the end of this period. The advance reservation of the MVNO should take into account the traffic uncertainty on the user side, as well as the price uncertainty on the MNO side, i.e., the user set χ varies between sessions due to random user arrivals, departures and movements. In addition, the on-demand price c s It will also vary depending on the intensity of competition between MVNOs in each time period. We assume that MVNOs estimate the on-demand price c for a period of time based on historical information. s The distribution of and Respectively represent c s The cumulative distribution function and probability density function of c r Set to below c s , to encourage advance booking.
[0069] The impact of price differences is discussed in detail through rigorous analysis.
[0070] On-demand request:
[0071] At the beginning of each session, the MVNO sets the on-demand price c according to the observed user set χ and the on-demand price c announced by the MNO. s Decide whether to request additional subchannels. Let n s is the number of subchannels requested in the session. The on-demand request problem is stated as:
[0072]
[0073] Its subchannel u g is a fixed scaler that converts utility into monetary units. G(χ,n) represents the maximum utility obtained by providing n users with χ. The first term in the objective function (3) is the cost of the on-demand subchannel, and the second term is the utility-based income of the service user. The optimal value Q(χ,c s ,n r ) is called session surplus, which becomes part of the MVNO’s total profit. Note that the number of subchannels in an on-demand request varies between sessions.
[0074] 3 Dynamic resource allocation:
[0075] The maximum system utility G(χ,n) of a session is achieved by dynamic resource allocation at the physical layer, where the mobile virtual network operator dynamically allocates n = n according to the fast channel fading in each time slot. r +n ssubchannels to k users in x. Thanks to the independent channel variations between users and subchannels, dynamic resource allocation can greatly improve the utility of the system due to multi-user diversity. represents the utility function of each user, where is the average throughput of the session. We assume is a continuously differentiable increasing concave function that captures the user's received throughput. By deploying different utility functions, MVNOs can achieve a balance between user throughput maximization and service fairness.
[0076] Assume g ki [t] represents the fast fading channel gain between the base station and user k in subchannel i at time slot t, b ki [t] represents the corresponding achievable data rate calculated by (1). Let a ki [t]∈[0,1] represents the proportion of talk time allocated to user k in subchannel i at time slot t. represents all allocation decisions in the session. Then, the average throughput of user k (for k = 1, ..., K) is given by
[0077]
[0078] Where T is the total number of time slots in the session.
[0079] The problem of finding the optimal subchannel allocation strategy for all time slots can be formulated as:
[0080]
[0081] In particular, when This means that if the user set is empty, the utility is 0, and if represents the penalty value, then G(χ,n) may be negative, for example, when The dynamic resource allocation problem in (5) attempts to minimize the overall transmission delay.
[0082] From the above description, we can see that the three optimization problems are nested, and the smaller time scale Q(χ,c s ,n r ) and the optimal values of G(χ,n) are respectively embedded into the objective functions of the larger time scale (2) and (3). Conversely, the larger time scale n r and n s The decisions in are parameters of the problems in (3) and (5) at small time scales.
[0083] Next, we first study the dynamic resource allocation problem (5). Based on the results of dynamic resource allocation, the optimal on-demand request and advance reservation of the general utility function are derived in Section 4. Section 5 further analyzes the special properties of the two-stage leasing scheme when the proportional fairness utility function is adopted.
[0084] Dynamic resource allocation
[0085] We first study the dynamic resource allocation problem in (5), which attempts to maximize the utilization of subchannels obtained from advance reservations and on-demand requests according to the fast channel fading in each time slot.
[0086] Optimal dynamic resource allocation strategy
[0087] let and are the Lagrangian multipliers of the two constraints in (5). Then the Lagrangian function is:
[0088]
[0089] Let A * represents the optimal solution and denotes the corresponding optimal average throughput for all users, the following conditions hold:
[0090]
[0091] here, yes exist We can infer from (7) that if Then v ki [t] = 0. Therefore, we have
[0092]
[0093] In other words, subchannel i is assigned to a channel that has The maximum value of the user. Because b ki [t] comes from a continuous distribution, so two or more users have the same The probability of the value is zero, so each subchannel is only allocated to a single user in each time slot. The optimal subchannel allocation strategy is described as:
[0094]
[0095] The dynamic resource allocation strategy in (9) states that the subchannels in each time slot are allocated to users that can obtain a relatively high bit rate as a function of their average throughput. Note that the subchannel allocation strategy in (9) requires knowledge of the optimal average throughput This will be derived in the next subsection.
[0096] B. Optimal Average Throughput
[0097] Using the subchannel allocation strategy in (9), the average throughput per user is given by
[0098]
[0099] 1 of them {.} is an indicator function, which is equal to 1 if the parameter is true, otherwise it is equal to 0. Assuming that T is large enough and the channel fading process is ergodic, the time-averaged throughput in (10) can be transformed into the expectation of independent fast fading, which is given by
[0100]
[0101] The last step is the fast fading independence between users. Note that Division requires
[0102] This is true for the continuous increase of the utility function. Since fast fading is an independent and unidirectional distribution, b ki The same distribution is followed for all i, where the cumulative distribution function F bk (.) and the probability density function f bk (.) can be calculated based on the Rayleigh fading distribution of (1). Note that F bk (.) and f bk (.) Depends only on the user's location x k Then, (11) can be written as
[0103]
[0104] From (12), we can see that the optimal throughput is the root of the following nonlinear equation system
[0105]
[0106] in,
[0107]
[0108] According to Brouwer's fixed point theorem, system (13) has at least one root. From (14), we can see that Φ k (r) is r k is a decreasing function, because Yes k is a decreasing function, which means that if all are all fixed, then satisfy r k =nΦ k (r) k There is at most one,system of equations in (13) with a unique root.
[0109] The average throughput of all users achieved by the optimal dynamic resource allocation strategy can be calculated by solving the nonlinear system (13). Substituting into (9), we can obtain the optimal dynamic resource allocation strategy for each group of users in the case of n subchannels to maximize the total system utility. The dynamic resource allocation strategy is calculated at the beginning of each session based on the distribution of fast fading. Then, at the beginning of each time slot, subchannels are allocated to serve users based on the estimated instantaneous channel fading.
[0110] C. Optimal System Utility
[0111] Using the optimal dynamic resource allocation strategy, we now derive an analytical expression for the optimal value G(χ,n) that is used to solve the on-demand request problem in (3). We use To represent the optimal throughput vector obtained by using n subchannels, by solving equation (13), we can get and the corresponding optimal utility In general, there is no closed-form expression for G(χ,n), and to keep the analysis tractable, we consider a broad class of utility functions where the average throughput achieved increases linearly with the number of subchannels.
[0112] Lemma 1:
[0113] The average throughput per user achieved in the optimal dynamic resource allocation strategy (9) increases linearly with the number of subchannels, i.e.
[0114]
[0115] The above formula requires that the utility function satisfies the following conditions
[0116] For any non-negative rates r1, r2>0 and positive integers n=1, 2, ..., we have
[0117]
[0118] prove:
[0119] Using the properties in (16), the average rate in (12) can be written as
[0120]
[0121] This means For all n, the system of equations r k=Φ k (r), for k = 1, ..., K. Since the system has a unique root So we have
[0122]
[0123] Thus we get the result (15).
[0124] The most famous class of utility functions that satisfy condition (16) is the α-fair utility function, which is defined as
[0125]
[0126] where α is the fairness and is the average throughput of the user. Specifically, the dynamic resource allocation problem based on α-fair utility is transformed into a throughput maximization problem when α = 0 and a delay minimization problem when α = 2. In addition, when α = 1, proportional fairness is achieved among users through the logarithmic utility function. Note that not all increasing concave utility functions satisfy condition (16). For example, the exponential utility function and diminishing returns None of them can satisfy condition (16).
[0127] When (16) holds, the average user throughput obtained by the optimal dynamic resource allocation strategy is a linear function of the number of subchannels. Therefore, when n=1, the nonlinear system (13) only needs to be solved once to obtain the average throughput for general n.
[0128] The optimal utility function is given by
[0129]
[0130] It will be used in the next section to derive the optimal on-demand requests and advance reservations. Note that is uniquely determined by the user location χ and the distribution of fast channel fading.
[0131] Section 4. Optimal two-stage leasing scheme
[0132] In this section, we study the optimal on-demand request and advance reservation for a class of utility functions that satisfy the conditions of Lemma 1.
[0133] Best on-demand request
[0134] Using the optimal utility obtained from the dynamic resource allocation problem, the on-demand request problem in (3) becomes
[0135]
[0136] According to the property of (16), we can calculate the objective function for n s The first-order derivative of is
[0137]
[0138] in,
[0139]
[0140] Note that Θ(χ) is uniquely determined by the user position χ and can be numerically computed by solving the nonlinear system (13) with n = 1. Since the target in (21) is n s The optimal solution is obtained by setting the first-order derivative (22) to zero.
[0141]
[0142] yes The inverse function of From (24), we can see that the mobile virtual network operator uses the reserved n r subchannels to serve the baseline traffic and Request additional subchannels when . Due to the concave property of U(.), is a decreasing function, so With c s This means that the mobile virtual network operator requests fewer sub-channels when observing higher on-demand prices or user realizations with smaller utility metric Θ(χ).
[0143] A special case is the linear utility function, where in this case, is zero or infinite, depending on This is because the rental cost and the user's utility function both increase with n. s Increases linearly, and the slope of the linear increase depends on Θ(χ). In practice, the number of certificates sold is usually limited. Therefore, the optimal strategy of the mobile virtual network operator is to rent as many sub-channels as possible. If Otherwise or there is no on-demand request.
[0144] For the optimal on-demand request (24), we can calculate the corresponding surplus in the session as
[0145]
[0146] in,
[0147]
[0148] Specifically, we have This is equivalent to zero payoff in the idle session, where we use the real-valued solution As the calculation of Q(χ,c s ,n r ), the approximation error in computing the optimal advance reservation in the next subsection can be neglected because the approximation error can be found in E[Q(χ,c s ,n r )] to find the average value.
[0149] Best Advance Booking
[0150] Using Q(χ,c in (25) s ,n r ), solve the advance booking problem in (2), let J(n r )and Denote the objective function and its first-order derivative of (2) respectively. Combined with (16), we can calculate for
[0151]
[0152] From (27), we have
[0153]
[0154] because When E[c s ]≤c r , That is, J(n r ) is n r A non-increasing function of . Therefore, the optimal non-negative solution is Therefore, when E[c s ]≤c r MVNOs do not make advance reservations, i.e. In other words, all subchannels are obtained from on-demand requests in each session.
[0155] The above shows that price discounts are crucial to motivating MVNOs to make advance bookings. From the perspective of MVNOs, advance bookings have the advantages of no risk in returns and simple operation. Therefore, MVNOs usually set discounts to encourage MVNOs to reserve resources for a long time.
[0156] For E[c s ]>c r In the case of r ) is n r is a concave function, so the optimal real-valued solution is Satisfy the first-order condition
[0157]
[0158] From (29), we can see that when c r Lower or c s When it is higher, This matches the intuition that the MVNO reserves more subchannels in advance for discounted subscription prices so that it can avoid expensive on-demand requests.
[0159] Due to the lack of a closed expression for Θ(χ), even if there is c s and χ distribution functions, nor can we calculate J(n r )and Therefore, there is no optimal retention Analytical expression. Since J(n r ) is concave. We propose a stochastic gradient descent algorithm, as shown in Algorithm 1, to find the optimal number of subchannel leases in the advance booking phase. The key idea is to approximate the true gradient with the gradient of the sampled user set χ in each iteration, which is given by
[0160]
[0161] The last step is as follows:
[0162]
[0163] Using c s The CDF of , the approximate gradient can be efficiently calculated by (30). For the i-th iteration, when the step size is set to In L = O (1 / ∈ 2 ) After iteration, the stochastic gradient descent algorithm can solve an ∈-approximate solution, and the integer solution can be obtained by rounding to the optimal real-valued solution.
[0164]
[0165] Section 5 Special Cases: Proportional Fairness Utility
[0166] In this section, we apply the analytical results of the previous section to proportional fairness utility and show the unique properties of the two-stage leasing scheme that are not reflected in the general utility function. Formally, the proportional fairness utility function is defined as The first-order derivative is It satisfies the conditions in Lemma 1.
[0167] A. Optimal on-demand request
[0168] Will Substituting into (23), we obtain Θ(χ) = K. Then, the optimal on-demand request in (24) can be simplified to
[0169]
[0170] Compared with the general case (24), the optimal on-demand request of proportional fairness utility in (33) is only related to the number of users in the session and has nothing to do with the user location. This is because in the total proportional fairness utility The information about the user's location χ only appears in the second additional term that is irrelevant to the rental decision n. The physical meaning of (33) can be explained as follows: if there are only a small number of active users or the on-demand price is very high, that is, u g K≤n r c s , then the MVNO does not need additional subchannels. Otherwise, the MVNO will request additional subchannels. The number of subchannels increases linearly with the number of users. The on-demand request of the MVNO is related to c s This is the same demand curve widely used in telecommunication systems.
[0171] B. Best Advance Booking
[0172] Will Substituting Θ(χ)=K into (29), we have
[0173]
[0174] As shown in Proposition 1, for E[c s ]>c r , In this case, (34) can be rewritten as
[0175]
[0176] In the first step, for a non-negative random variable Y, we have E(Y) = ∫Pr(Y≥η)dη, and F K (.) is the CDF function of K. Using c s The CDF of and K can be obtained by solving equation (35): Due to the proportional fairness utility function Only relying on the distribution of K, rather than the distribution of user locations χ as in the general utility function case, i.e., without using Algorithm 1, MVNOs can easily calculate the optimal number of subchannel leases in the advance booking stage.
[0177] Proposition 2:
[0178] In the case of proportional fairness utility function, the average cost of the optimal two-stage rental scheme increases linearly with the average number of users per session, i.e.
[0179]
[0180] Proof: From (34), the cost of reserving a subchannel can be calculated as follows:
[0181]
[0182] Furthermore, from (33), the average cost of on-demand requests can be calculated as:
[0183]
[0184] Adding (37) to (38), we can get the total average cost as:
[0185]
[0186] This leads to the proof.
[0187] Proposition VB shows that during busy periods with many users, more investment is needed, and the allocation of investment to advance bookings and on-demand requests depends on K and c s This will be discussed in the next section using numerical results.
[0188] Section 6. Numerical results and discussions
[0189] In this section, the performance of the derived two-stage leasing scheme is numerically evaluated. The users of the MVNOs with labels are randomly placed in circular cells with a radius of D = 1 km. The number of users in each session K is a random integer uniformly distributed in the interval [0,16]. The wireless channel is a Rayleigh fading channel. The standard path loss model with a path loss exponent of 3.67 is adopted. The average signal-to-noise ratio received at the cell edge is -6 dB. The reservation price c is assumed. r Normalized to 1 and u g = 5. Unless otherwise stated, the on-demand prices are uniformly distributed in [0.8, 1.8]. The general α-fair utility functions with α = 1 and α = 0.8 are considered as two examples. Note that α = 1 corresponds to proportional fairness. The results are given in Section V. Extensive model experimental results show that the use of other distributions or utility functions does not change the conclusions of the present invention, so they are omitted for brevity. We use MATLAB software to run system-level simulations.
[0190] Figure 3 The convergence process of Algorithm 1 is shown. r[l] is plotted as a function of the iteration step l. The main complexity of Algorithm 1 lies in sampling enough realizations of the user set, which can be obtained from the historical records. Since the sampling is performed offline, the complexity of Algorithm 1 is not a problem for practical implementation. With optimal advance reservation, the mobile virtual network operator can calculate the optimal on-demand request in each session through the closed-form expression in (24). The overhead of obtaining the user location and on-demand price at the beginning of each session is very low compared to the session duration. The leased subchannels from advance reservations and on-demand requests are allocated to users according to the optimal dynamic resource allocation strategy in (9), where the channel estimation has been standardized in the physical layer protocol. According to our analysis results, the mobile virtual network operator can implement two-stage leasing and dynamic resource allocation with low operational complexity.
[0191] As we show in Section 4, the optimal rental decision in both phases depends on the on-demand price c s and the changes of user set χ, in this section, we will study c through numerical simulation s and χ changes. Specifically, the level of variation of the random variable z is measured by the coefficient of variation, which is defined as:
[0192]
[0193] Among them, σ z and μ z are the standard deviation and mean of the random variable z, respectively. z , we can compare the level of variation of z at different scales.
[0194] A. Impact of On-Demand Price Changes
[0195] exist Figure 4 , fixed on-demand price μ cs and change ξ cs The coefficient of variation of the on-demand price is used to study the impact of the optimal two-stage rental solution. For uniform distribution, c s Supported by Give, especially The corresponding period is c s For constant cases, focus on the case As shown in Proposition 1, c r Avoid trivial solutions Specifically, In the figure it is set to 1.2 and 1.3.
[0196] exist Figure 4 In 4a and 4b, we plotted the optimal advance booking and optimal on-demand requests Expectations as As can be seen from the function Get bigger, Reduce, This means that when the on-demand price changes greatly, the number of advance reservations by the mobile virtual network operator decreases and the number of on-demand requests increases. This trend also holds true for the utility function of The larger it is, the more sub-channels are reserved in advance and the fewer sub-channels are leased on demand due to the higher cost of on-demand leasing.
[0197] Further in Figure 5 The average total number of subchannels used per session and the average rental cost of each subchannel are shown in Figures 5a and 5b, respectively. Figure 5 In a, we can see that with The larger the change, the more sub-channels are leased overall. This is because when the change is large, c s The low percentile of will become lower, sometimes, c s The realization of r This provides MVNOs with the opportunity to lease sub-channels on demand at low cost. Figure 5 As shown in b, the average rental cost of each subchannel increases with This indicates that the derived two-stage leasing scheme can reduce the cost of MVNOs by taking advantage of short-term low on-demand prices. As a result, the larger The total ordering of subchannels is more when When is lower, the total number of leased sub-channels increases and the average cost decreases, i.e., the MVNO consumes more sub-channels at a lower on-demand price.
[0198] B. Impact of traffic changes
[0199] One advantage of the two-stage leasing scheme is that MVNOs can flexibly adjust their leasing requests based on the real-time traffic in the on-demand request phase. Figure 6 In the above example, we fixed the flow intensity μ K The average value of K The coefficient of variation is used to study the impact of traffic changes on the optimal leasing decision. Figure 6 a and 6b, when ξ K Increase, best booking Lowering expectations for optimal on-demand requests In some sense, this is intuitive, as MVNOs tend to rely more on on-demand requests and less on advance reservations in order to handle more fluctuating traffic demands. The observation holds for different traffic intensities μ K and different fairness factors α.
[0200] In order to obtain a larger μ K , whether booked in advance or requested on demand, more sub-channels will be purchased, that is, mobile virtual network operators need more sub-channels to serve more intensive traffic.
[0201] C. Comparison with Phase I Leasing Program
[0202] For comparison, the single-period spectrum leasing schemes of reservation-only and on-demand-only are used as benchmarks. In the reservation-only scheme used in [8], the MVNO specifies a number of subchannels at the beginning of a period and has no other opportunities to purchase spectrum resources during the entire period. The formulation of the reservation-only problem is similar to the stochastic programming in (2), except that Replaced by E x [G(χ,n r )], and E x [G(χ,n r )] Regardless of the on-demand stage, the best booking, use It can be expressed as:
[0203]
[0204] The average surplus can be calculated based on this. In the on-demand only scheme used in
[12] , the MVNO dynamically purchases spectrum resources based on on-demand users and their locations. The on-demand only problem is a special case of the optimization problem (3), where n r = 0, optimal on-demand request, expressed as Its expression is:
[0205]
[0206] The average surplus for the period can be calculated accordingly.
[0207] We are respectively Figure 7 In Fig. 7a and 7b, the proposed two-stage leasing plan is compared with the one-stage leasing plan under different demand price changes. and different flow rate changes K The surplus obtained by the first-level rental scheme under Figure 7 Different in a The corresponding rows (lines) remain unchanged. From these two figures, we can see that for all and K , the two-stage leasing scheme achieves a higher surplus than the reservation-only scheme, as and K As the number of increases, the difference becomes more obvious. This is because the on-demand phase retains the flexibility to obtain additional subchannels to handle the emerging traffic. In this way, the two-phase leasing scheme effectively reduces the overbooking in the advance booking phase. In addition, we can see that for all and K The two-stage leasing scheme also outperforms the on-demand leasing scheme. and K The advantage is more obvious as the first phase enables mobile virtual network operators to reserve sub-channels at a lower cost. With reserved sub-channels to serve benchmark traffic, mobile virtual network operators can avoid purchasing expensive on-demand sub-channels in real time. In short, the two-phase leasing solution can take advantage of the low cost of advance booking and the flexibility of on-demand requests.
[0208] Section 7. Conclusion
[0209] In this invention, a two-stage leasing scheme is derived, which enables mobile virtual network operators to take advantage of both the low cost of pre-booking and the flexibility of on-demand booking. In order to find the optimal leasing decision in these two stages, I formulate the problem as a three-level nested optimization problem and solve it efficiently. Through analysis, it is concluded that mobile virtual network operators can easily calculate the optimal number of subchannels to be leased in each stage. Numerical results show that the proposed two-stage spectrum leasing scheme can adapt to different degrees of network changes and achieve more surplus than the traditional single-stage leasing scheme.
[0210] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0211] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A spectrum leasing method combining spectrum reservation and on-demand request, characterized in that: It includes three groups of time scales, namely advance booking, on-demand request, and dynamic resource allocation; Advance reservation: MVNOs reserve multiple sub-channels for a long period of time. This decision is optimized based on estimated statistics of user traffic and on-demand prices during that period, ensuring that MVNOs obtain a certain amount of spectrum resources at a relatively low cost, allowing MVNOs to pre-plan resource slicing and system operations at an early stage. On-demand request: The mobile virtual network operator decides whether to request additional sub-channels after observing the actual location of the user who needs service and the current on-demand price. The on-demand request changes with the arrival, departure and movement of users, changes within a fixed time range, and does not change with fast fading channels. The mobile virtual network operator can obtain additional spectrum resources based on real-time traffic demand. The mobile network operator can determine the real-time competition intensity between multiple mobile virtual network operators; Dynamic resource allocation,On a millisecond time scale, the mobile virtual network operator dynamically allocates,subchannels obtained from advance reservations and on-demand requests to users,based on the instantaneous wireless channel fading.
2. A spectrum leasing method for combined spectrum reservation and on-demand request according to claim 1, characterized in that: Advance reservation and on-demand request are two-stage leasing schemes. Advance reservation, on-demand request and dynamic resource allocation constitute a set of three-level nested optimization problems. The nested structure makes the three sub-problems closely intertwined. The decisions made on a large time scale will affect the optimization on a small time scale. The optimal value of the problem on a small time scale is embedded in the objective function of the problem on a large time scale.
3. A spectrum leasing method for combined spectrum reservation and on-demand request according to claim 1, characterized in that: The optimal channel-aware dynamic resource allocation strategy is derived under the optimal utility function in advance reservation and on-demand request. In the on-demand request stage, a closed-form expression is used to apply the optimal number of rentable subchannels. In the advance reservation stage, the stochastic gradient descent algorithm is used to find the number of subchannels to be rented. For the special case of using proportional fairness utility function, a closed-form expression for the optimal number of rentable subchannels in the two stages is used.
4. A spectrum leasing method for combined spectrum reservation and on-demand request according to claim 3, characterized in that: The internal data collection of the proportional fairness utility function includes using a broadband wireless receiver to collect signals, making a rough estimate of the symbol rate of the receiver collected signals, and setting the initial value S of the upper and lower cutoff frequencies of the bandpass filter. Based on the measurement signal and data-driven method, the upper and lower cutoff frequencies of the bandpass filter are continuously optimized, and the optimal cutoff frequency is generated. The optimal filter is used to process the signal complex envelope with the optimal upper and lower cutoff frequencies, and the symbol rate spectrum is extracted to complete the symbol rate estimation.
5. A spectrum leasing method for combined spectrum reservation and on-demand request according to claim 3, characterized in that: The utilization of subchannels obtained from advance reservations and on-demand requests is maximized according to the fast channel fading in each time slot.
6. A spectrum leasing method for combined spectrum reservation and on-demand request according to claim 1, characterized in that: The analytical results for deriving the optimal solutions to the on-demand request and advance reservation subproblems are applied to the proportional fairness utility and demonstrate the unique properties of the two-stage rental scheme that are not reflected in the general utility function.
7. A spectrum leasing method for combined spectrum reservation and on-demand request according to claim 2, characterized in that: The performance of the derived two-stage leasing scheme is numerically evaluated.
Citation Information
Cited By
Intelligent communication method and system integrating mobile communication and broadcast communication
CN119967600A
