Cache content placement and resource allocation optimization method for low-orbit multi-beam communication satellites

Abstract

The present invention provides a cache content placement and resource allocation optimization method for low-orbit multi-beam communication satellites, comprising the following steps: resource allocation step: firstly calculating the cumulative remaining request amount of each cell in time slot t within the TTL time A(j,t) , select the cell with the largest value, allocate the required bandwidth and power resources, update the remaining resources, and then repeat this process until the number of allocated cells is the same as the number of beams of a single satellite or the resources are allocated; cache update step: when updating the cache, the cache space is C, each file has (C+1) states, 0 state means that the file is not cached, 1 to C states are all cache states, file i is in state C means that it is the latest requested file at time t, and the hit rate of file i at time t is H i (t) That is, it refers to the probability that file i is in one of these cache states at time t. The beneficial effect of the present invention is that it can better improve the utilization rate and user experience effect.

Description

Technical Field

[0001] The present invention relates to the field of satellite technology, and in particular to a cache content placement and resource allocation optimization method for a low-orbit multi-beam communication satellite. Background Art

[0002] In the 5G wireless era, satellite communication networks and ground networks complement each other, and the satellite-ground integrated network that combines satellite networks with ground networks is an inevitable trend in technological development. With the increasing demand for various emerging applications, traditional satellite communications are difficult to support the current communication capacity requirements, and multi-beam satellites have emerged. A multi-beam satellite communication system refers to a satellite communication system that can generate multiple spot beams at the same time to cover a specific area on the ground. Compared with a single-beam satellite communication system, the advantages of multi-beam satellites are: improving power spectrum density, miniaturizing ground terminals, and using the idea of ​​MIMO systems, that is, jointly processing spatially separated beams to achieve flexible resource scheduling. For beams with heavier loads, more power and bandwidth resources can be allocated to improve the system throughput.

[0003] According to the characteristics of data traffic, which maintains a certain popularity over a period of time, most of the mobile data traffic is generated by the repeated transmission of a small number of the most popular files. The wireless communication network repeatedly sends these contents to each user, which leads to low transmission efficiency. By caching these highly popular contents in the network nodes in advance, when users request these files, they can directly obtain them from the local cache without any transmission, which can reduce the transmission load of the wireless communication network and save resource overhead.

[0004] In-network caching can also be used to improve the performance of satellite-ground networks. There have been some works on this issue, but most of them mainly take advantage of the wide coverage of satellites and improve the technology of ground networks, ignoring the role of caching in satellites. In fact, the storage capacity of the ground is limited, on the other hand, with the rich resource configuration, satellites have become more powerful than before. In this context, caching part of the data flowing on the satellite is necessary and practical, especially for content delivery services.

[0005] Wireless resource management technology refers to the technology that improves the established system utility by allocating, scheduling and optimizing the limited resources on the satellite under the condition of considering the uncertain factors such as system traffic volume and channels. It is one of the key technologies in the multi-beam satellite communication system and is related to the service quality and system performance of the entire multi-beam satellite communication system. Due to the diversity of current business types, the spatial unevenness and time-varying nature of business distribution, the traditional fixed power and frequency resource allocation strategy can no longer meet the on-demand coverage and efficient transmission requirements of current diversified tasks. Therefore, researchers have proposed a beam-hopping technology system to flexibly allocate on-board resources according to the changing distribution of traffic demand. The ACTS satellite launched in 1993 was the first experimental satellite to demonstrate beam-hopping technology. Five beams were responsible for serving 51 wave positions, which not only reduced the satellite's transmission power, but also increased the system capacity by nearly 3 times.

[0006] However, most of the current research on beam hopping technology is focused on GEO satellite communication systems. Although some studies have been conducted on beam hopping of LEO satellites, only the direction of the beam is considered, which is an ideal assumption. At the same time, current research on LEO satellite caching issues generally only considers the cache location and what content to cache in order to achieve the best expected effect, and there is little research on cache updates. Summary of the invention

[0007] The present invention provides a method for optimizing cache content placement and resource allocation for a low-orbit multi-beam communication satellite, comprising the following steps:

[0008] Resource allocation steps: First, calculate the cumulative remaining request amount A(j,t) of each cell in time slot t within the TTL time, select the cell with the largest value, allocate the required bandwidth and power resources to it, and update the remaining resources at the same time, and then repeat this process until the number of allocated cells is the same as the number of beams of a single satellite or the resources are allocated.

[0009] Cache update step: When updating the cache, the cache space is C. Each file has (C+1) states. State 0 means that the file is not cached. States 1 to C are all cached states. File i in state C means that it is the latest requested file at time t. The hit rate of file i at time t is H. i (t) refers to the probability that file i is in one of these cache states at time t.

[0010] As a further improvement of the present invention, in the resource allocation step, it also includes:

[0011] Step 1: Calculate the cumulative remaining request amount of each cell in time slot t. The cumulative remaining request amount of cell j in time slot t is:

[0012]

[0013] Where d(j,t) is the request amount of cell j in time slot t:

[0014]

[0015] q i Indicates the user's access to file f i The request probability, λ j represents the mathematical expectation of the number of users in cell j, V(j,t) represents the beam coverage status of cell j in time slot t, and s i Represents file f i size.

[0016] Step 2: Select the cell with the largest value, assign it a beam V(j,t)=1, bandwidth B(j,t) and power P(j,t), and update B at the same time. max =B max -B(j,t) and P max =P max -P(j,t); then select the cell with the largest cumulative remaining request amount from the remaining cells and allocate beams, bandwidth and power to it.

[0017] Step 3: Repeat step 2 until the number of allocated cells is the same as the number of beams of a single satellite or the resource allocation is completed.

[0018] As a further improvement of the present invention, in step 1, there are a total of N files in the file library that can be requested by ground users, and the N files are sorted according to their popularity. The request probability of the N files forms a vector q=[q1,q2...,q N ], the request probabilities of N files are sorted from high to low, and the more popular files are ranked higher.

[0019] As a further improvement of the present invention, in the cache update step, first initialize the iteration counter z=1, and then propose two assumptions, Assumption 1: Assume that a request for any file k (k≠i) will cause file i to transfer from state j to state (j-1); Assumption 2: If files k and i are both in the cache, if k (k≠i) expires, it will not affect the state of file i; Under these two assumptions, a state transition intensity matrix is ​​generated, and the probability matrix of file i in time slot t is calculated, thereby obtaining the initial value of the hit rate of file i, and the initial values ​​of the hit rates of N files are calculated, and then the above two assumptions are corrected to generate a corrected probability transfer intensity matrix and probability matrix, and then generate the hit rate values ​​of N files in the first iteration, and repeat this iterative process until the hit rate values ​​of the N files generated in this iteration are the same as the hit rate values ​​generated in the previous iteration, and the required precise value is obtained.

[0020] As a further improvement of the invention, in the cache update step, the state transition intensity matrix is ​​determined by the file request rate and the failure rate.

[0021] As a further improvement of the present invention, in the cache update step, the probability matrix of file i in time slot t is obtained by performing exponential operation on the state transition intensity matrix.

[0022] As a further improvement of the present invention, the cache updating step includes:

[0023] Modification of Assumption 1: Exclude requests for objects currently occupying cache state w>j from the probability transition intensity of file i from state j to state (j-1) when the assumption is not modified;

[0024] As a further improvement of the present invention, the cache updating step includes:

[0025] Modification to Assumption 2: Combine the expiration rates of objects that currently occupy cache state w>j;

[0026] As a further improvement of the present invention, in the cache update step, the hit rates of N files in time slot t are defined as the weighted sum of the hit rates of each file in time slot t.

[0027] The beneficial effects of the present invention are: 1. The present invention studies common on-board resource allocation methods, combines the development of beam-hopping technology, and proposes a resource joint optimization method that matches the request amount, which can better improve the utilization rate of high-throughput multi-beam satellite communication resources and user experience effects; 2. The resource allocation method and cache update method proposed in the present invention increase the effectiveness of system throughput. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of performance comparison between two resource allocation modes and different cache strategies under different satellite cache spaces of the present invention;

[0029] Figure 2 A schematic diagram showing the performance comparison of three cache strategies under different resource allocation modes under different numbers of single satellite beams of the present invention;

[0030] Figure 3 A schematic diagram showing the performance comparison of three cache strategies under different resource allocation modes under different satellite number parameters of the present invention;

[0031] Figure 4 Schematic diagram of the multi-beam low-orbit satellite communication system set forth in the invention.

[0032] Figure 5 Cache update flow chart. DETAILED DESCRIPTION

[0033] Through the study of Markov chains and user file requests, it is found that continuous-time Markov chains are a type of Markov process with continuous time and discrete states. This process has the property of "no memory", that is, the probability of reaching each state in the next step is only determined by the current state, and has nothing to do with the state of the previous time steps. The update of the cache probability for each file is also only related to the failure rate of the current file and the user request probability, so it is possible to use Markov chains to study the cache update problem.

[0034] When allocating resources, first calculate the cumulative remaining request amount A(j,t) of each cell in time slot t within TTL (TTL belongs to time period l), select the cell with the largest value, allocate the required bandwidth and power resources to it, update the remaining resources, and then repeat this process until the number of cells allocated resources is equal to the number of beams of a single satellite or there are no resources remaining.

[0035] When updating the cache, the cache space is C, and each file has (C+1) states. State 0 means that the file is not cached, and states 1 to C are all cached states. File i is in state C when it is the most recently requested file at time t. The hit rate of file i at time t is the probability that file i is in one of these cache states at time t.

[0036] First, there are two assumptions. ① Assume that a request for any file k (k≠i) will cause file i to transfer from state j to state (j-1). ② Ignore the situation when object k (k≠i) expires while objects k and i are in the cache. If files k and i are both in the cache, if k (k≠i) expires, it will not affect the state of file i. Under these two assumptions, a state transition intensity matrix is ​​generated, and the probability matrix of file i in time slot t is calculated, thereby obtaining an initial hit rate value, that is, an initial cache probability of file i, and calculating the initial hit rate values ​​of N files. Then, by correcting the above two assumptions, ① exclude requests for objects that currently occupy cache states w>j, and ② merge the expiration rates of objects that currently occupy cache states w>j. According to these two corrections, the hit rate value of the first iteration is generated, and this iterative process is repeated until the hit rate values ​​of the N files generated in this iteration are the same as the hit rate values ​​generated in the previous iteration, and the required accurate value is obtained. The relevant parameters of resource allocation and cache update are brought into the derived system throughput function, and compared with other resource allocation methods and cache schemes.

[0037] The specific approach of the present invention is:

[0038] At present, for most users, the communication service needs are all in the forward link (the forward link refers to the process of multimedia services being uploaded from the gateway to the satellite for transfer and then sent to ground users), and the resource control of the forward link is centralized and processed on the satellite, so here we mainly discuss the downlink segment of the forward link, that is, the process of multimedia communication services being transmitted to ground users via the satellite through the beam-hopping downlink. Here, the downlink adopts time division multiplexing, and the coverage area of ​​each spot beam is called a beam position. There are a total of H spot beams jumping between M cells to meet ground service requests.

[0039] Let V be the beam hopping matrix, V(j,t) represents the beam coverage status of cell j in time slot t. If cell j has beam coverage in time slot t, V(j,t) = 1, otherwise V(j,t) = 0. Let B be the bandwidth allocation matrix, B(j,t) is the bandwidth allocation of cell j in time slot t. Let P be the power allocation matrix, P(j,t) is the power allocation of cell j in time slot t. If V(j,t) = 0, then B(j,t) = 0 and P(j,t) = 0.

[0040] In order to accurately describe the service performance of the satellite, the parameters are set as follows: the total bandwidth is B max , the total power of the satellite is P max , the total window length is W, when there is beam coverage, the signal-to-noise ratio of cell j in time slot t is SNR(j,t), then the capacity allocation of cell j is as follows:

[0041] C(j,t)=B(j,t)·log2(1+SNR(j,t))

[0042]

[0043] B(j,t) is the bandwidth allocated to cell j in time slot t, P(j,t) is the power allocated to cell j in time slot t, P EIRP is the effective isotropic radiated power, L FS is the free space loss, G / T is the value of the receiver (G is the antenna gain, T is the antenna noise temperature), G Tx is the transmitting antenna gain of the payload on the spacecraft, G Rx is the antenna gain of the user terminal, λ is the wavelength of the communication link signal, k is the Boltzmann constant, T syst is the system noise temperature.

[0044] When allocating resources, some terminal information, channel information, and satellite information are known. Terminal information such as the maximum bandwidth B of a single satellite max and power P max , user density λ j0, Zipf distribution parameter α, number of cells in the ground area M; channel information includes the wavelength λ of the communication link signal, the satellite height d, and the transmit antenna gain G of the payload Tx , the antenna gain G of the user terminal Rx ; Satellite information includes the number of satellites L, the cache capacity of a single satellite I, and the number of beams of a single satellite H.

[0045] Through these known information, we can get the optimized resource allocation matrix B, P, V. The specific process is as follows:

[0046] Step 1: Calculate the cumulative remaining request amount of each cell in time slot t. The cumulative remaining request amount of cell j in time slot t is:

[0047]

[0048] Where d(j,t) is the request amount of cell j in time slot t:

[0049]

[0050] q i Indicates the user's access to file f i The request probability satisfies the Zipf distribution. There are N files in the file library that can be requested by ground users. The files are sorted according to their popularity. That is, file f1 is the most popular and most likely to be requested by users. The corresponding file size is s = {s1, s2, ..., s N}.

[0051] The request probability of N files forms a vector q=[q1,q2...,q N ], the request probabilities of N files are sorted from high to low, and the more popular files are ranked higher.

[0052]

[0053] in is the normalization factor A lower α means that more files have similar popularity, and a higher α means that a few files have very high popularity, while a large number of files have very low popularity.

[0054] Assume that the number of users in cell j follows an independent and identical Poisson distribution, that is

[0055]

[0056] where λ j It represents the mathematical expectation of the number of users in cell j and is related to the cell size. Assume that the mathematical expectation of the number of users in a unit area in cell j is λ j0 , then λj =πr 2 λ j0

[0057] Step 2: Select the cell with the largest value and assign it a beam V(j,t)=1, bandwidth B(j,t) and power P(j,t). At the same time, update B max =B max -B(j,t) and P max =P max -P(j,t); then select the cell with the largest cumulative remaining request amount from the remaining cells and allocate beam bandwidth and power to it.

[0058] Step 3: Repeat step 2 until the number of allocated cells is the same as the number of beams of a single satellite or the resource allocation is completed.

[0059] Through simulation analysis, the present invention obtains the performance comparison of the traditional fixed resource allocation method and the resource allocation method proposed by the present invention with several caching strategies in terms of system throughput under different satellite-related parameters, different user densities, and different file popularity parameters.

[0060] in, Figure 1 It shows the performance comparison of two resource allocation methods and different cache strategies under different satellite cache spaces. The present invention sets H = 5, λ j0 =0.003, α=0.6, B max =200MHz, P max =200W, M=15, d=600km.

[0061] It can be seen that for different satellite-related parameters, the resource allocation method proposed in the present invention can effectively improve the system performance in terms of throughput. At the same time, the performance of the most popular cache strategy is always better than random cache and uniform cache.

[0062] Regarding cache update, the model considered in the present invention includes L LEO multi-beam satellites, which jointly serve a large ground area Λ. Since LEO satellites and the ground are always in a state of relatively high-speed motion, after a satellite serves the ground area for a period of time, its coverage area will deviate from Λ, and the next satellite will take over the service, and this cycle will be repeated. The basic idea of ​​cache update here is to implement LRU (Least Recently Used) through Markov chain, and set the cache matrix of the second satellite according to the cache status of the previous satellite and the user request status.

[0063] The basic idea of ​​the LRU algorithm is that data that has not been used for a long time is unlikely to be used in the future. Therefore, when the space occupied by the cached data reaches the threshold, the least recently used data is removed to introduce new data.

[0064] For a continuous-time Markov chain, there is a stateful transition intensity matrix Q:

[0065]

[0066] The elements of the Q matrix satisfy q ij ≥0, E={0,1,2,L} ​​is the state space.

[0067] The analysis assumes that the cached object i has a finite lifetime, which is exponentially distributed with a mean of 1 / μ. i . The lifetime of an object is independent of its size and popularity. The Independent Reference Model (IRM) is used to study how the file expiration rate and request rate affect the instantaneous hit rate. The total number of files in the file library is N. The requests for N files satisfy the Poisson distribution (β), and the requests for each file satisfy the Zipf distribution. The cache space is C, and each file has (C+1) states. State 0 means that the file is not cached, and states 1 to C are all cached states. File i is in state C, which means that it is the most recently requested file at time t. The hit rate of file i at time t refers to the probability that file i is in one of these cache states at time t. The hit rate of file i at time t after the final iteration is the cache probability of file i after update.

[0068] (1) Generate an initial hit rate value under two assumptions

[0069] ① Assume that a request for any file k (k≠i) will cause file i to move from state j to state (j-1)

[0070] ② Ignore the situation when object k (k≠i) expires while objects k and i are in the cache. If files k and i are both in the cache, if k (k≠i) expires, it will not affect the status of file i.

[0071] (2) The exact value of the hit rate is generated by iterating the corrections to the above two assumptions

[0072] ① Exclude requests for objects currently occupying cache state w>j from the probability transition intensity of file i transferring from state j to state (j-1) when it is assumed to be unmodified.

[0073] ② Combine the expiration rates of objects that currently occupy the cache state w>j.

[0074] The iteration is completed until the hit rate values ​​of the N files generated in this iteration are the same as the hit rate values ​​generated in the previous iteration, and the required accurate value is obtained.

[0075] Through simulation analysis, the present invention obtains a performance comparison of cache update and traditional most popular cache strategy and random cache strategy in terms of system throughput under different satellite-related parameters, different user densities, and different file popularity parameters.

[0076] Figure 2 The performance comparison of the three cache strategies under different resource allocation modes under different numbers of single satellite beams is shown. The present invention sets I = 1000MB, λ j0 =0.003, α=0.6, B max =200MHz, P max =200W, M=15, d=600km, L=8.

[0077] Figure 3 The performance comparison of the three cache strategies under different resource allocation modes under different satellite number parameters is shown. The present invention sets I = 1000MB, λ j0 =0.003, H=5, α=0.6, B max =200MHz, P max =200W, M=15, d=600km.

[0078] It can be seen that for different satellite-related parameters, the proposed cache update method can effectively improve the system performance in terms of throughput.

[0079] The present invention studies the problem of on-board resource allocation and on-board file cache matrix update of LEO satellites, considers the hierarchical heterogeneity of the system, combines beam hopping technology, jointly allocates on-board bandwidth and power resources of LEO satellites, and uses Markov chains to update the cache matrix of the satellite to maximize the system throughput:

[0080] 1. By studying the common on-board resource allocation methods and combining the development of beam-hopping technology, a resource joint optimization method that matches the request volume is proposed to better improve resource utilization.

[0081] 2. The present invention uses Markov chain to implement LRU algorithm, completes the update of file cache probability, and proposes the concept of throughput to represent the sum of the amount of data successfully transmitted within a certain period of time under this resource allocation method and cache update method, and obtains the function expression of system throughput, which verifies the effectiveness of the proposed resource allocation method and cache update method in increasing system throughput.

[0082] The beneficial effects of the present invention are as follows: 1. The present invention studies common on-board resource allocation methods, combines the development of beam-hopping technology, and proposes a resource joint optimization method that matches the request amount, which can better improve the utilization rate of high-throughput multi-beam satellite communication resources and user experience effects; 2. The resource allocation method and cache update method proposed in the present invention increase the effectiveness of system throughput.

[0083] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

Claims

1. A cache content placement and resource allocation optimization method for low-orbit multi-beam communication satellites, characterized in that: This involves performing the following steps: Resource allocation step: First, calculate the cumulative remaining request amount A(j,t) of each cell in time slot t within the TTL time, select the cell with the largest value, allocate the required bandwidth and power resources to it, and update the remaining resources at the same time, and then repeat this process until the number of allocated cells is the same as the number of beams of a single satellite or the resources are allocated; Cache update step: When updating the cache, the cache space is C. Each file has (C+1) states. State 0 means that the file is not cached. States 1 to C are all cached states. File i in state C means that it is the latest requested file at time t. The hit rate of file i at time t is H. i (t) refers to the probability that file i is in one of these cache states at time t; The resource allocation step further includes: Step 1: Calculate the cumulative remaining request amount of each cell in time slot t. The cumulative remaining request amount of cell j in time slot t is: Where d(j,t) is the request amount of cell j in time slot t: q i Indicates the user's access to file f i The request probability, λ j represents the mathematical expectation of the number of users in cell j, V(j,t) represents the beam coverage status of cell j in time slot t, and s i Represents file f i size; Step 2: Select the cell with the largest value, assign it a beam V(j,t)=1, bandwidth B(j,t) and power P(j,t), and update B at the same time. max =B max -B(j,t) and P max =P max -P(j,t); then select the cell with the largest cumulative remaining request amount from the remaining cells and allocate beams, bandwidth and power to it; Step 3: Repeat step 2 until the number of allocated cells is the same as the number of beams of a single satellite or the resource allocation is completed.

2. The cache content placement and resource allocation optimization method according to claim 1, characterized in that: In step 1, there are N files in the file library that can be requested by ground users. The N files are sorted according to their popularity. The request probability of the N files forms a vector q = [q1, q2..., q N ], the request probabilities of N files are sorted from high to low, and the more popular files are ranked higher.

3. The cache content placement and resource allocation optimization method according to claim 1, characterized in that: In the cache update step, first initialize the iteration counter z=1, and then propose two conditions, condition 1: a request for any file k (k≠i) will cause file i to transfer from state j to state (j-1); condition 2: if files k and i are both in the cache, if k (k≠i) expires, it will not affect the state of file i; under these two conditions, a state transition intensity matrix is ​​generated, and the probability matrix of file i in time slot t is calculated, thereby obtaining the initial hit rate of file i, and the initial hit rate values ​​of N files are calculated, and then the above two conditions are corrected to generate a corrected probability transfer intensity matrix and probability matrix, and then generate the hit rate values ​​of N files in the first iteration, and repeat this iterative process until the hit rate values ​​of N files generated in this iteration are the same as the hit rate values ​​generated in the previous iteration, and the required precise value is obtained.

4. The cache content placement and resource allocation optimization method according to claim 3, characterized in that: In the cache update step, the state transition intensity matrix is ​​determined by the file request rate and the failure rate.

5. The cache content placement and resource allocation optimization method according to claim 3, characterized in that: In the cache update step, the probability matrix of file i in time slot t is obtained by performing exponential operation on the state transition intensity matrix.

6. The cache content placement and resource allocation optimization method according to claim 3, characterized in that: The cache updating step includes: Modification to condition 1: exclude requests for objects currently occupying cache state w>j from the probability transition intensity of file i from state j to state (j-1) when it is not modified.

7. The cache content placement and resource allocation optimization method according to claim 3, characterized in that: The cache updating step includes: Modification to condition 2: Combine the expiration rates of objects that currently occupy cache status w>j.

8. The cache content placement and resource allocation optimization method according to claim 1, characterized in that: In the cache update step, the hit rates of N files in time slot t are defined as the weighted sum of the hit rates of each file in time slot t.

Images