WO2012167498A1 - 基于模拟退火算法的频率优化方法和装置 - Google Patents
基于模拟退火算法的频率优化方法和装置 Download PDFInfo
- Publication number
- WO2012167498A1 WO2012167498A1 PCT/CN2011/077958 CN2011077958W WO2012167498A1 WO 2012167498 A1 WO2012167498 A1 WO 2012167498A1 CN 2011077958 W CN2011077958 W CN 2011077958W WO 2012167498 A1 WO2012167498 A1 WO 2012167498A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- optimization
- thread
- state solution
- new state
- frequency
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/02—Resource partitioning among network components, e.g. reuse partitioning
- H04W16/04—Traffic adaptive resource partitioning
Definitions
- the present invention relates to the field of communications technologies, and in particular, to a frequency optimization method and apparatus based on a simulated annealing algorithm.
- Frequency optimization is a process of redistributing the planned carrier frequencies of each cell in a wireless network to reduce network interference (same frequency/inter-frequency interference).
- a cell in a wireless network generally plans a carrier frequency group including multiple carrier frequencies. Therefore, for one cell, frequency optimization is a search optimization process of a carrier frequency group, which belongs to a combination optimization problem.
- the frequency optimization process for each cell, search optimization is continuously performed within a carrier frequency set (called a solution space) available to the cell, and each available carrier frequency group (referred to as an available solution) is searched.
- the cost evaluation criterion is "the smallest interference in the whole network (can be the same frequency / inter-frequency interference)".
- the carrier frequency group currently allocated by each cell is called the current state solution, and the available and better carrier frequency group obtained by the optimization is called the proximity state solution.
- Frequency optimization is to continuously transform the current state solution into a neighboring state solution, and finally to make the whole-network cell carrier frequency planning reach the optimal state solution process.
- the simulated annealing algorithm is derived from the solid annealing principle and is a stochastic optimization algorithm based on the Monte Carlo iterative solution strategy. It has been widely used in many fields and has been introduced into the field of frequency optimization. For example, "Fixed Frequency Allocation Problem Based on Simulated Annealing Algorithm" (Microcomputer Information, 2007, issue 26) discloses a method for using a simulated annealing algorithm for frequency optimization.
- the better neighbor state solution must be based on the current state solution. Therefore, the simulated annealing algorithm itself is a serial process, in which the advantages and disadvantages of the initial state solution will directly affect the simulation.
- the efficiency of the annealing algorithm in the case of extremely harsh initial conditions, frequency optimization will be a long process with low efficiency.
- Embodiments of the present invention provide a frequency optimization method based on a simulated annealing algorithm to improve the efficiency of frequency optimization.
- a frequency optimization method based on simulated annealing algorithm including:
- the method includes: selecting, by each of the two or more threads, a cell group, and performing frequency optimization based on the simulated annealing algorithm on the selected cell group;
- a frequency optimization device based on simulated annealing algorithm comprising:
- the optimization module is configured to perform multi-thread optimization based on the current state solution at the current annealing temperature, and the one-time multi-thread optimization comprises: selecting one cell group for each of the two or more threads, and performing the selected cell group A frequency optimization based on simulated annealing algorithm;
- the judging module judges whether to accept the new state solution obtained by the optimization module in the multi-thread optimization according to the constraint condition between the multi-threads.
- the technical solution of the embodiment of the present invention adopts a multi-threading mechanism, and multiple threads respectively perform frequency optimization based on simulated annealing algorithm on respective selected cell groups, and judge whether to accept each multi-thread optimization according to constraints between multiple threads.
- the new state solution can effectively improve the efficiency of frequency optimization compared to the existing frequency optimization method.
- FIG. 1 is a flow chart of a frequency optimization method based on a simulated annealing algorithm according to an embodiment of the present invention
- FIG. 2 is a trend diagram of a frequency optimization process based on a simulated annealing algorithm
- FIG. 3 is a flow chart of steps for determining constraints between multiple threads according to frequency optimized progress;
- FIG. 4 is a flow chart of a frequency optimization method based on simulated annealing algorithm according to another embodiment of the present invention;
- FIG. 5 is an embodiment of the present invention;
- FIG. 6 is a schematic diagram of a frequency optimization device based on a simulated annealing algorithm according to another embodiment of the present invention.
- the embodiment of the invention provides a frequency optimization method based on a simulated annealing algorithm, adopts a multi-threading mechanism, and sets constraints between multiple threads to improve optimization efficiency.
- Embodiments of the present invention also provide corresponding devices. The details are described below separately.
- FIG. 1 an embodiment of the present invention provides a frequency optimization method based on a simulated annealing algorithm.
- the multi-threading mechanism can improve the efficiency of frequency optimization, and the method can be implemented by a frequency optimization apparatus.
- a complete frequency optimization process based on simulated annealing algorithm can be described in three aspects: initial state solution (initial scheme), new state solution (new scheme) and its acceptance, and optimization time.
- the initial state solution is the carrier frequency/carrier frequency group currently planned by each cell in the live network;
- the new state solution indicates the available and better new carrier frequency/carrier frequency group obtained by the current cell through search optimization, and the acceptance degree
- the cost of the new state solution is evaluated according to the cost function in the simulated annealing algorithm to determine whether to accept the new state solution;
- the optimization time refers to the time taken from the initial state solution to the completion of the frequency optimization.
- Figure 2 shows the trend of the frequency optimization process based on the simulated annealing algorithm.
- the horizontal axis represents the optimization time t in seconds (s); the vertical axis represents the current annealing temperature T and the search for each annealing temperature.
- the number of acceptances of the new program that is excellent is N. From Figure 2, the following conclusions can be drawn: 1. The annealing temperature and the acceptance rate of the new scheme at each annealing temperature are gradually reduced with the optimization time, and the variation trend is generally consistent; 2. The annealing temperature and the acceptance of the new scheme at each annealing temperature The trend of the number of times varies greatly in different time periods.
- the annealing temperature decreases sharply in a short time, and a large number of acceptable new solutions can be generated at each annealing temperature, which is called a highly active state; Between the thresholds 1 and 2, the annealing temperature decreases gradually with the optimization time, and the probability of producing an acceptable new solution at each annealing temperature is also gradually reduced, which is called normal active state; after threshold 2, annealing The temperature changes little with time, and the probability of producing an acceptable new solution at each annealing temperature is very low, approximately reaching the optimal solution, and the system state tends to be stable.
- the frequency optimization based on the simulated annealing algorithm begins with determining the initial annealing temperature.
- the number of times the new state solution is accepted whether the acceptance degree (ie, the ratio of the number of times the new state solution is accepted to the frequency optimization number) reaches a preset value (for example, 0.9), and if so, the current annealing temperature is the initial annealing temperature; If not, increase the annealing temperature (for example, double the original) and continue to optimize.
- the initial temperature is determined, it is also necessary to obtain the current state solution (currently planned carrier/carrier group).
- the current state solution can be manually input to the frequency optimization device or can be input to the frequency optimization device by the network element in the wireless network. The above steps of determining the initial temperature and obtaining the current state solution may be performed in a conventional manner.
- the frequency optimization based on the simulated annealing algorithm can be officially started, including: 101.
- the one-time multi-thread optimization includes: selecting one cell group for each of the two or more threads, and performing a simulation based on the selected cell group. Frequency optimization of the annealing algorithm.
- a frequency optimization process of searching for a new state solution for a set number of times is required, for example, 100 times.
- multiple threads are simultaneously optimized, and the number of optimizations per thread can be reduced, thereby improving the optimization frequency. Assuming 5 threads are optimized at the same time, each thread optimization needs only 20 optimizations at each annealing temperature.
- each of the two or more threads selects a plurality of cells as one cell group, and each thread only optimizes the cell group, and does not have to optimize all the cell groups. Further improve the optimization efficiency. For example, each of the five threads selects two cells as their own cell group, and only needs to search for two carrier frequencies for each optimization of each thread. It should be noted that each thread needs to reselect the cell for each optimization, and different threads must not select the same cell.
- the constraint determines whether to receive the new state solution obtained by this multi-thread optimization.
- the constraint can be defined as: Accept or not accept this time when the ratio of the number of threads accepting the new state solution obtained by the optimization to the total number of threads is higher than, equal to, or lower than a predetermined value. New state solution obtained by multi-thread optimization.
- the frequency optimization method based on the simulated annealing algorithm provided by the embodiment of the present invention adopts a multi-threading mechanism, and multiple threads respectively perform frequency optimization on respective selected cell groups, and a constraint condition between multiple threads is set for Determine whether to accept the new state solution obtained by this multi-thread optimization, as opposed to the present Some frequency optimization methods can effectively improve the efficiency of frequency optimization.
- an embodiment may further include the following steps:
- the constraints between multiple threads are determined based on the progress of the frequency optimization.
- This step can be performed prior to step 101 to employ different constraints at different frequency optimization stages to further improve optimization efficiency.
- this embodiment can use threshold 2 as the conversion threshold and divide the whole frequency optimization process into two phases. Before the transition threshold (the threshold 1 is merged into one phase), it is called “optimistic optimization.” Phase “, after the transition threshold is called “pessimistic optimization phase”, in the “optimistic optimization phase” and “pessimistic optimization phase” respectively set different constraints for determining whether to receive the new state solution obtained by this multi-thread optimization. At this time, as shown in FIG. 3, the step of determining the constraint condition between the multiple threads according to the progress of the frequency optimization may specifically include:
- the conversion threshold that is, the threshold 2 in FIG. 2, can be calculated according to the cell size of the wireless network (the number of cells N) and the annealing algorithm complexity coefficient M (recommended value 36), according to a strategy of a single order,
- the conversion threshold can be set equal to the product N*M of the number of cells N and the annealing algorithm complexity coefficient M.
- the conversion threshold is related to the number of times the new state solution obtained by the previous multi-thread optimization is not accepted. By determining whether the unaccepted number reaches the conversion threshold, it is determined whether the current is an "optimistic optimization phase" or a "pessimistic optimization phase.”".
- the new scheme accepts more times at each annealing temperature, that is, the acceptance probability is higher.
- the following constraints are determined: if each thread in the multi-thread receives its own optimized new If the state is solved, it is judged to accept the new state solution obtained by the current multi-thread optimization. Otherwise, it is judged as not accepting.
- the constraint condition is called an optimistic constraint condition.
- the conversion threshold determines that the pessimistic constraint is used, including: if one or more threads in the multi-thread accept the new state solution obtained by the optimization, determine to accept the new state solution obtained by the multi-thread optimization, otherwise judge Not accepted.
- the number of unacceptable times reaches the transition threshold, it indicates that the pessimistic optimization phase has been entered.
- the number of acceptances of the new scheme at each annealing temperature is small, that is, the acceptance probability is very low.
- the following constraints are determined: if one and only one thread in the multi-thread accepts the new optimization If the state is solved, it is judged to accept the new state solution obtained by the current multi-thread optimization, otherwise it is judged as not accepting.
- the constraint condition is called a pessimistic constraint condition.
- the optimistic optimization phase, the pessimistic optimization phase, and the pessimistic optimization phase are divided, and according to the actual situation of different phases, the constraint conditions matching the phase can be set, which can be further effectively improved. Optimize the frequency.
- the judgment result of step 102 is as follows:
- the new state solution obtained by accepting the multi-thread optimization means all thread optimization
- the new state solution that does not accept this multi-threaded optimization means that: The new state solution obtained by all thread optimization is not accepted.
- the following examples illustrate:
- each thread does not accept the new state solution obtained by itself, or it may be just one or two The thread does not accept the new state solution obtained by itself, but according to the optimistic constraint, the new state solution obtained by this multi-thread optimization is not accepted.
- the new state solution obtained by accepting this multi-thread optimization means: Only the new state solution obtained by one thread optimization is accepted; the new state obtained by this multi-thread optimization is not accepted.
- the solution means that the new state solution obtained by all thread optimization is not accepted.
- each thread randomly selects two cells as one cell group to perform frequency optimization.
- the available carrier frequency is 9 and from F1. To F9.
- the acceptance of each thread is as follows (constraints have not been considered):
- the frequency optimization process may be further divided according to other methods.
- the optimistic optimization phase is further divided into two phases according to the threshold 1, and different constraints are set.
- the following steps may be further included: 103. Count and determine whether the number of optimizations at the current annealing temperature reaches a set value.
- the number of optimizations at the current annealing temperature does not reach the set value, it is necessary to continue to optimize the frequency at the current temperature. Then, based on whether the new state solution obtained by this multi-thread optimization is accepted, it is determined how the next frequency optimization is performed. If the new state solution obtained by this multi-thread optimization is accepted, correspondingly, the new state solution obtained by the multi-thread optimization is used as the current state solution, and the next multi-thread optimization can be performed. 1042. If the set value is not reached, and the new state solution obtained by the multi-thread optimization is not accepted, the multi-thread optimization at the current temperature is ended, and the single-thread optimization is performed based on the finally accepted new state solution.
- the multi-thread optimization at the current temperature is ended, and the finally accepted new state solution is taken as the current state solution.
- next annealing temperature needs to be determined.
- the next annealing temperature can be determined according to a preset cooling mechanism, for example, reducing a fixed degree (for example, 5 degrees) or reducing a fixed ratio (for example, reducing 5%) based on the previous annealing temperature to obtain a new one.
- the temperature is taken as the next annealing temperature.
- the frequency optimization at the next annealing temperature is continued, and the steps are the same as those described in the foregoing, and will not be described in detail.
- the frequency optimization exit condition is a preset condition for judging whether to end the entire frequency optimization process, and the condition may be set in various manners, one way is to set according to the annealing temperature, if the annealing temperature is lowered to some
- the preset value (for example, an annealing temperature close to 0) ends the frequency optimization; the other method is set according to the number of times the new state solution obtained by the previous multi-thread optimization is not accepted, if not The number of acceptances reaches a preset value (the preset value can be a value less than the conversion threshold to ensure that the optimistic phase will be entered during the frequency optimization process), then the frequency optimization is ended; another way is to follow the network interference.
- the purpose of frequency optimization is to reduce network interference, so you can set a network interference expected value (such as X% of the original network interference quantization value, X is less than 100), when the network interference is calculated according to the current state solution
- a network interference expected value such as X% of the original network interference quantization value, X is less than 100
- the frequency optimization is ended. If the frequency optimization exit condition is reached at an annealing temperature, the frequency optimization is ended, and the accepted current state solution obtained as the last time is used as the optimal state to de-allocate the cells, and thus the entire frequency optimization process is completed.
- the frequency optimization method based on the simulated annealing algorithm provided by the embodiment of the present invention adopts a multi-threading mechanism, which is equivalent to simultaneously performing multiple sets of frequency optimization based on simulated annealing algorithm on the same frequency optimization device (or server) to achieve parallel Searching for the running effect of the new state solution can effectively improve the efficiency of frequency optimization, and by setting reasonable multi-thread constraints to determine whether to accept the new state solution, it can guarantee high quality optimization results and further improve Frequency optimization efficiency.
- an embodiment of the present invention further provides a frequency optimization apparatus based on a simulated annealing algorithm, including:
- the optimization module 501 is configured to perform multi-thread optimization based on the current state solution at the current annealing temperature.
- the one-time multi-thread optimization includes: selecting one cell group for each of the two or more threads, and selecting the cell group Perform a frequency optimization based on simulated annealing algorithm;
- the determining module 502 determines, according to the constraint condition between the multiple threads, whether to accept the new state solution obtained by the optimization module in the current multi-thread optimization.
- the method may further include:
- the determining module 503 is configured to determine a constraint between the multiple threads according to the frequency optimized progress.
- the determining module 503 may be specifically configured to count the number of times that the new state solution obtained by the previous multi-thread optimization is not accepted, and determine whether the unaccepted number reaches the conversion threshold; if the conversion threshold is not reached.
- Determine the use of optimistic constraints including: If each thread in the multi-thread accepts the new state solution obtained by itself, it is judged to accept the new state solution obtained by the multi-thread optimization, otherwise it is judged as not accepting;
- the conversion threshold is determined, and the pessimistic constraint is determined, including: If one and only one thread in the multi-thread accepts the new state solution obtained by the optimization, it is determined to accept the new state solution obtained by the multi-thread optimization, otherwise it is judged as not accepting.
- the processing module 504 may be further included.
- the determining module 502 is further configured to count and determine whether the optimized number of times at the current annealing temperature reaches a set value.
- the processing module 504 is configured to, if the determining module determines that the set value has been reached, end the multi-thread optimization at the current temperature, and use the finally obtained accepted new state solution as the current state solution. Further, the processing module 504 may be further configured to: if the determining module determines that the set value is not reached, and the new state solution obtained by the multi-thread optimization is accepted, update the new state to the current state. Solving, the optimization module is notified to perform the next multi-thread optimization; if the judging module judges that the set value is not reached, and the new state solution obtained by the multi-thread optimization is not accepted, the multi-thread optimization at the current temperature is ended. Notifying the optimization module to perform single-thread optimization based on the last accepted new state solution.
- the determining module 502 may be further configured to determine whether the frequency optimization exit condition is reached.
- the processing module 504 may be further configured to determine the next annealing if the determining module determines that the frequency optimization exit condition is not met. Temperature; if the judging module judges that the frequency optimization exit condition has been reached, the frequency optimization is ended.
- the central processing unit includes more than two cores, so as to provide more than two threads for implementing the frequency optimization method based on the simulated annealing algorithm provided by the embodiments of the present invention.
- the device provided by the embodiment of the present invention may use multiple threads to perform frequency optimization on each selected cell group, and set a constraint condition between multiple threads to determine whether to accept the new state solution obtained by the current multi-thread optimization. Compared with the existing frequency optimization method, the efficiency of frequency optimization can be effectively improved.
- a person of ordinary skill in the art may understand that all or part of the steps of the foregoing embodiments may be completed by hardware, or may be instructed by a program to execute related hardware, and the program may be stored in a computer readable storage medium, and stored.
- the medium may include: a read only memory, a random access memory, a magnetic disk or an optical disk, and the like.
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
一种基于模拟退火算法的频率优化方法,包括:在当前退火温度下,基于当前状态解进行一次多线程优化,一次多线程优化包括:两个以上线程中的每个线程分别选择一个小区组,对选取的所述小区组进行一次基于模拟退火算法的频率优化;根据多线程间的约束条件判断是否接受本次多线程优化得到的新状态解。本发明实施例还提供相应的装置。本发明技术方案相对于现有的频率优化方法,可以有效提高频率优化的效率。
Description
基于模拟退火算法的频率优化方法和装置
技术领域
本发明涉及通信技术领域,具体涉及一种基于模拟退火算法的频率优化方 法和装置。
背景技术
频率优化是对无线网络中各小区已规划的载频进行重分配优化,以达到降 低网络干扰 (同频 /异频干扰)的过程。无线网络中一个小区一般会规划一个包括 多个载频的载频组, 因此针对一个小区来说, 频率优化是一个载频组的搜索寻 优过程, 属于一种组合优化问题。
在频率优化过程中, 针对每个小区, 要在该小区可用的载频集合内(称为 解空间)不断进行搜索寻优, 对搜索到的每个可用载频组 (称为一个可用方案) 进行代价评估, 代价评估准则为 "全网干扰最小 (可以是同频 /异频干扰)"。 其 中每个小区当前分配的载频组称为当前状态解,寻优得到的可用的、 更优的载 频组称为邻近状态解。频率优化就是不断把当前状态解转变为邻近状态解, 最 终使得全网小区载频规划达到最优状态解的过程。
模拟退火算法来源于固体退火原理, 是基于蒙特卡罗( Monte Carlo )迭代 求解策略的一种随机寻优算法, 目前已在多个领域取得广泛应用, 并已被引入 频率优化领域。 例如, 《基于模拟退火算法的固定频率分配问题》(《微计算机 信息》, 2007年 26期)就公开了一种将模拟退火算法用于频率优化的方法。
基于模拟退火算法的频率优化方法中,更优的邻近状态解的产生必须以当 前状态解为基础, 因此模拟退火算法本身是一个串行的过程, 其中, 初始状态 解的优劣将直接影响模拟退火算法的效率,在初始状态极度恶劣的情况下, 频 率优化将是一个漫长的过程, 效率很低。
发明内容
本发明实施例提供一种基于模拟退火算法的频率优化方法,以提高频率优 化的效率。
一种基于模拟退火算法的频率优化方法, 包括:
在当前退火温度下,基于当前状态解进行一次多线程优化, 一次多线程优
化包括: 两个以上线程中的每个线程分别选择一个小区组,对选取的所述小区 组进行一次基于模拟退火算法的频率优化;
根据多线程间的约束条件判断是否接受本次多线程优化得到的新状态解。 一种基于模拟退火算法的频率优化装置, 包括:
优化模块,用于在当前退火温度下,基于当前状态解进行一次多线程优化, 一次多线程优化包括: 两个以上线程中的每个线程分别选择一个小区组,对选 取的所述小区组进行一次基于模拟退火算法的频率优化;
判断模块,根据多线程间的约束条件判断是否接受所述优化模块本次多线 程优化得到的新状态解。
本发明实施例的技术方案采用多线程机制,由多个线程分别对各自选取的 小区组进行基于模拟退火算法的频率优化,并根据多线程间的约束条件判断是 否接受每次多线程优化得到的新状态解,相对于现有的频率优化方法, 可以有 效提高频率优化的效率。
附图说明
图 1是本发明一个实施例的基于模拟退火算法的频率优化方法的流程图; 图 2是基于模拟退火算法的频率优化过程的趋势图;
图 3是根据频率优化的进度确定多线程间的约束条件的步骤的流程图; 图 4是本发明另一实施例的基于模拟退火算法的频率优化方法的流程图; 图 5是本发明一个实施例的基于模拟退火算法的频率优化装置的示意图; 图 6是本发明另一实施例的基于模拟退火算法的频率优化装置的示意图。
具体实施方式
本发明实施例提供一种基于模拟退火算法的频率优化方法,采用多线程机 制, 并设定有多线程间的约束条件, 以提高优化效率。 本发明实施例还提供相 应的装置。 以下分别进行详细说明。 请参考图 1 , 本发明实施例提供一种基于模拟退火算法的频率优化方法, 采用多线程机制,可以提高频率优化的效率,该方法可以由频率优化装置实施。
一次完整的基于模拟退火算法的频率优化过程可以从三个方面进行描述, 即: 初始状态解(初始方案), 新状态解(新方案)及其接受度, 和优化时间。
其中, 初始状态解即是现网中各小区当前规划的载频 /载频组; 新状态解表示 当前小区通过搜索寻优得到的可用的、 更优的新载频 /载频组, 接受度是根据 模拟退火算法中的代价函数对新状态解进行代价评估, 判断是否接受新状态 解; 优化时间是指从初始状态解开始到频率优化完成所消耗的时间。
图 2示出了基于模拟退火算法的频率优化过程的趋势图, 图中横轴表示优 化时间 t, 以秒(s ) 为单位; 纵轴表示当前的退火温度 T和每个退火温度下搜 索寻优得到的新方案的接受次数 N。 从图 2可以得出以下结论: 1. 退火温度和 每个退火温度下新方案的接受次数随着优化时间逐渐减少, 变化趋势大体一 致; 2. 退火温度和每个退火温度下新方案的接受次数的变化趋势在不同的时 间段存在很大的差异性, 在门限 1之前退火温度在短时间内急剧降低, 每个退 火温度下可以产生大量可接受的新方案, 称之为高度活跃状态; 在门限 1和 2 之间, 退火温度的降温幅度随优化时间逐渐减緩,每个退火温度下可以产生可 接受新方案的概率也逐渐降低, 称之为正常活跃状态; 在门限 2之后, 退火温 度随时间变化很小,每个退火温度下可以产生可接受新方案的概率很低, 近似 达到最优解, 系统状态趋于稳定态。
基于模拟退火算法的频率优化,从确定初始退火温度开始。初始退火温度 的计算实际上是模拟退火算法的逆过程, 即: 从一个较低的退火温度开始(例 如可以从退火温度 t=l度开始 ), 逐渐升高退火温度(例如按照 t=t*2的指数速率 提升), 在每个退火温度下进行设定次数(例如 100次)的基于当前状态解(当 前规划的载频 /载频组) 的频率优化, 统计该退火温度下优化得到的新状态解 被接受的次数,计算接受度(即新状态解被接受的次数与频率优化次数的比值 ) 是否达到预设值(例如 0.9 ), 若达到, 则以当前退火温度为初始退火温度; 若 没有达到, 则提高退火温度(例如提高为原来的两倍), 继续优化。 初始温度 确定后, 还需要获取当前状态解(当前规划的载频 /载频组)。 当前状态解可以 由人工输入给频率优化装置, 也可以由无线网络中的网元输入给频率优化装 置。 以上确定初始温度以及获取当前状态解的步骤, 可以采用常用的方法。
在确定初始温度和获取当前状态解后,即可正式开始基于模拟退火算法的 频率优化, 包括:
101、 在当前退火温度下, 基于当前状态解进行一次多线程优化, 一次多 线程优化包括: 两个以上线程中的每个线程分别选择一个小区组,对选取的所 述小区组进行一次基于模拟退火算法的频率优化。
基于模拟退火算法的频率优化过程中,在每个退火温度下, 需要进行设定 次数的搜寻新状态解的频率优化过程, 例如 100次。 本实施例中, 则由多个线 程同时进行优化, 每个线程的优化次数便可以减少, 从而提高优化频率。 假设 采用 5个线程同时进行优化, 则在每个退火温度下, 每个线程优化只需要优化 20次即可。
进一步, 假设无线网络中有 10个小区, 每个小区规划一个载频。 则, 如果 采用单线程优化, 每次优化时需要搜寻 10个载频。 而本实施例中, 则由两个以 上线程中的每个线程分别选择若干个小区作为一个小区组,每个线程仅对该小 区组进行优化即可,而不必对全部小区组进行优化,可以进一步提高优化效率。 例如, 5个线程中的每个线程分别选择两个小区作为自己的小区组, 在每个线 程每次优化时仅需要搜寻 2个载频即可。 需要说明的是, 每次优化时各个线程 都需要重新选择小区, 并且, 不同的线程不得选择相同的小区。
102、 根据多线程间的约束条件判断是否接受本次多线程优化得到的新状 态解。
在采用单线程优化时, 每次优化完毕后, 按照预设的接收条件(例如代价 函数)判断是否接受本次优化得到的新状态解即可。但是,采用多线程优化后, 情况有所变化,在各个线程判断完毕是否接受自己优化得到的新状态解后,还 需要预先设定的多线程间的约束条件,才能进一步根据该多线程间的约束条件 判断是否接收本次多线程优化得到的新状态解。 约束条件可以这样定义: 当多 个线程中接受自己优化得到的新状态解的线程的个数与线程总个数的比值高 于、等于或者低于一个预定的值时,接受或者不接受本次多线程优化得到的新 状态解。
综上, 本发明实施例提供的基于模拟退火算法的频率优化方法, 采用多线 程机制, 由多个线程分别对各自选取的小区组进行频率优化, 并且设定有多线 程间的约束条件用于判断是否接受本次多线程优化得到的新状态解,相对于现
有的频率优化方法, 可以有效提高频率优化的效率。
然而, 从图 2中可以看出, 在频率优化过程的不同阶段, 每个退火温度下 的新方案接受次数有相当大的不同。 为了进一步提高优化效率, 一个实施例实 施中还可以包括以下步骤:
根据频率优化的进度确定多线程间的约束条件。
该步骤可以在步骤 101之前执行, 以便在不同的频率优化阶段, 采用不同 的约束条件, 从而进一步提高优化效率。
从图 2中可以看出, 在门限 1之后、 门限 2之前, 虽然每个退火温度下的新 方案接受度有所降低,但是就优化性能而言仍然是一个重要环节。基于优化性 能和优化效率的综合考虑, 本实施例可以将门限 2作为转换门限, 把整个频率 优化过程划分为两个阶段, 转换门限之前(门限 1前后过程合并为一个阶段) 称为 "乐观优化阶段", 转换门限之后称为 "悲观优化阶段", 在 "乐观优化阶 段" 和 "悲观优化阶段" 分别设定不同的约束条件, 用于判断是否接收本次多 线程优化得到的新状态解。 此时, 如图 3所示, 上述根据频率优化的进度确定 多线程间的约束条件的步骤具体可以包括:
1001、统计已进行的历次多线程优化得到的新状态解不被接受的次数, 并 判断所述不被接受的次数是否达到转换门限。
所说的转换门限, 即图 2中的门限 2, 可以根据无线网络的小区规模 (小区 个数 N)和退火算法复杂度系数 M (建议值 36)计算得到, 按照一种筒单的策略, 可以设定转换门限等于小区个数 N和退火算法复杂度系数 M的乘积 N*M。 该转 换门限与历次多线程优化得到的新状态解不被接受的次数相关,通过判断所述 不被接受的次数是否达到转换门限, 来确定当前是属于 "乐观优化阶段"或者 是 "悲观优化阶段"。 其中, 历次多线程优化得到的新状态解不被接受的次数, 是指从初始退火温度开始的第一次频率优化开始,将每次多线程优化得到的新 状态解不被接受的次数进行累计求和得到的次数。例如,设当前已降温到第三 个退火温度, 假定初始退火温度下新状态解不被接受的次数为 5, 第二个退火 温度下新状态解不被接受的次数为 6, 则当前阶段, 所述历次多线程优化得到 的新状态解不被接受的次数为 5+6=11。 接下来判断 11是否超过 N*M。
1002、 若未达到转换门限, 确定采用乐观约束条件, 包括: 若多线程中的 每个线程都接受自己优化得到的新状态解,则判断为接受本次多线程优化得到 的新状态解, 否则判断为不接受。
若所述不被接受的次数未达到转换门限,说明还处于乐观优化阶段。在乐 观优化阶段, 每个退火温度下新方案的接受次数较多, 也就是接受概率较高, 本实施例中确定采用如下约束条件:若多线程中的每个线程都接受自己优化得 到的新状态解, 则判断为接受本次多线程优化得到的新状态解, 否则判断为不 接受, 本实施例中称此约束条件为乐观约束条件。
1003、 若达到转换门限, 确定采用悲观约束条件, 包括: 若多线程中有且 只有一个线程接受自己优化得到的新状态解,则判断为接受本次多线程优化得 到的新状态解, 否则判断为不接受。
若所述不被接受的次数达到转换门限,说明已进入悲观优化阶段。在悲观 优化阶段, 每个退火温度下新方案的接受次数很少, 也就是接受概率很低, 本 实施例中确定采用如下约束条件:若多线程中有且只有一个线程接受自己优化 得到的新状态解, 则判断为接受本次多线程优化得到的新状态解, 否则判断为 不接受, 本实施例中称此约束条件为悲观约束条件。
综上, 本实施例中通过对频率优化过程进行分析, 划分出乐观优化阶段和 悲观优化阶段和悲观优化阶段,根据不同阶段的实际情况,设定匹配该阶段的 约束条件, 可以进一步有效的提升优化频率。
按照上述的乐观约束条件和悲观约束条件, 步骤 102的判断结果如下: 在乐观优化阶段, 有两种判断结果, 其中, 接受本次多线程优化得到的新 状态解意味着: 全部线程优化得到的新状态解都被接受; 不接受本次多线程优 化得到的新状态解则意味着: 全部线程优化得到的新状态解都不被接受。 下面 举例说明:
假设无线网络中有 6个小区,频率优化装置中有三个线程用于多线程优化, 每个线程随机选取两个小区作为一个小区组, 进行频率优化; 设可用的载频有 9个, 从 F1至 F9。 则在乐观阶段, 一次多线程优化后, 只有表 1和表 2表示的以 下两种结果:
小区组编号 当前状态解 新状态解 接受情况
(原始规划载频) (新载频)
1 ( F1,F2 ) ( F3,F2 ) 接受
2 ( F3,F4 ) ( F5,F8 ) 接受
3 ( F5,F6 ) ( F1,F7 ) 接受
表 2
小区组编号 当前状态解 新状态解 接受情况
(原始规划载频) (新载频)
1 ( F1,F2 ) ( F3,F2 ) 不接受
2 ( F3,F4 ) ( F5,F8 ) 不接受
3 ( F5,F6 ) ( F1,F7 ) 不接受 其中, 表 2所示的不接受的结果中, 未必每个线程都不接受自己优化得到 的新状态解,也可能是只是其中一个或两个线程不接受自己优化得到的新状态 解,但根据乐观约束条件 ,最终本次多线程优化得到的新状态解的都不被接受。
在悲观优化阶段, 有两种判断结果, 其中, 接受本次多线程优化得到的新 状态解意味着: 只有一个线程优化得到的新状态解被接受; 不接受本次多线程 优化得到的新状态解则意味着: 全部线程优化得到的新状态解都不被接受。 下 面举例说明:
殳无线网络中有 6个小区,频率优化装置中有三个线程用于多线程优化, 每个线程随机选取两个小区作为一个小区组, 进行频率优化; 设可用的载频有 9个, 从 F1至 F9。 则在悲观阶段, 假设一次多线程优化后, 各个线程的接受情 况如下 (尚未考虑约束条件 ):
小区组编号 当前状态解 新状态解 接受情况
(原始规划载频) (新载频)
1 ( F1,F2 ) ( F3,F2 ) 不接受
2 ( F3,F4 ) ( F6,F8 ) 接受
3 ( F5,F6 ) ( F1,F7 ) 接受 由于其中有两个线程的判断结果为接受自己优化得到的新状态解,按照悲 观约束条件, 本次多线程优化得到的新状态解的都不能被接受, 最终结果如表
4所示:
表 4
值得说明的是,在其它实施例中,还可以按照其它方式对频率优化过程进 行阶段划分, 例如, 按照门限 1将乐观优化阶段进一步划分为两个阶段, 并设 定不同的约束条件。 请参考图 4, 本发明实施例中, 在步骤 102之后还可以包括以下步骤: 103、 统计并判断当前退火温度下的优化次数是否达到设定值。
由于在每个退火温度下, 需要进行设定次数的频率优化过程。 因此, 在步 骤 102判断完是否接受本次多线程优化得到的新状态解之后, 还需要统计当前 退火温度下的优化次数,通过判断是否达到设定次数, 决定是否结束当前温度 下的多线程优化。
1041、 若未达到设定值, 且本次多线程优化得到的新状态解被接受, 则将 所述新状态接更新为当前状态解, 进行下一次多线程优化。
若当前退火温度下的优化次数没有达到设定值,说明还需要继续在当前温 度下进行频率优化。 然后, 可以根据本次多线程优化得到的新状态解是否被接 受, 决定下一次频率优化如何进行。如果本次多线程优化得到的新状态解被接 受, 相应的, 将本次多线程优化得到的新状态解作为当前状态解, 进行下一次 多线程优化即可。
1042、 若未达到设定值, 且本次多线程优化得到的新状态解不被接受, 则 结束当前温度下的多线程优化,基于最后得到的被接受的新状态解进行单线程 优化。
如果本次多线程优化得到的新状态解不被接受,并且当前温度下的优化次 数尚未达到, 则后续的频率可以按照如下方式进行: 结束当前温度下的多线程 优化,基于最后得到的被接受的新状态解进行单线程优化, 直到达到设定的次 数。 假设, 当前退火温度下需要进行 100次频率优化, 有 5个线程, 则每个线程 需要进行 20次频率优化,假设进行到第 14次多线程频率优化时,新状态解不被 接受, 则将该 5个线程都结束, 然后基于第 13次多线程频率优化的得到的被接 受的新状态解,由该 5个线程之外的主线程再进行 100-13=87次单线程的频率优 化, 至此, 100次频率优化完成, 退出当前温度。
1043、 若已达到设定值, 则结束当前温度下的多线程优化, 将最后得到的 被接受的新状态解作为当前状态解。
假如某次多线程优化后, 当前温度下的优化次数已达到设定值, 则结束当 前温度下的多线程优化, 将最后得到的被接受的新状态解作为当前状态解。
然后, 可以进行以下步骤:
105、 判断是否达到频率优化退出条件; 若没有达到, 则确定下一个退火 温度; 若已达到, 则结束频率优化。
若没有达到频率优化退出条件,说明还需要继续优化, 则需要确定下一个 退火温度。可以按照预设的降温机制确定下一个退火温度, 例如在前一个退火 温度的基础上降低一个固定的度数(例如 5度)或者降低一个固定的比例 (例 如降低 5% )等, 以得到的新温度作为下一个退火温度。 然后, 继续进行下一 个退火温度下的频率优化, 步骤与前文描述的各步骤相同, 不再详述。
若达到了频率优化退出条件,说明优化完成,可以结束整个频率优化过程。 其中,频率优化退出条件是预先设定的用来判断是否结束整个频率优化过程的 条件, 该条件可以有多种设定方式, 一种方式是按照退火温度设定, 如果退火 温度降低到某个预设值(例如一个接近 0的退火温度), 则结束频率优化; 另一 种方式是按照历次多线程优化得到的新状态解不被接受的次数设定,如果不被
接受的次数达到某个预设值(该预设值可以是一个小于转换门限的值, 以保证 频率优化过程中会进入悲观优化阶段), 则结束频率优化; 再一种方式是按照 网络干扰的大小设定, 频率优化的目的即是降低网络干扰, 因而可以设定一个 网络干扰的期望值(例如原始网络干扰量化值的 X% , X小于 100 ), 当根据当 前状态解计算得到的网络干扰量化值达到该网络干扰的期望值时,结束频率优 化。 如果某个退火温度下, 达到了频率优化退出条件, 则结束频率优化, 以最 后一次得到的被接受的现状解作为最优状态解分配各个小区, 至此, 整个频率 优化过程完成。 综上, 本发明实施例提供的基于模拟退火算法的频率优化方法, 采用多线 程机制, 等效于同一台频率优化装置(或服务器)上同时执行多组基于模拟退 火算法的频率优化, 达到并行搜寻新状态解的运行效果, 可以有效提升频率优 化的效率, 并且,通过设定合理的多线程间的约束条件用于判断是否接受新状 态解, 可以保证得到高质量的优化结果, 并进一步提高频率优化效率。
发明人对本发明实施例提供的基于模拟退火算法的频率优化方法进行了 实际的测试验证, 结果如表 5所示:
优化方案个数
1600 3000 3000
(优化次数) 电月 配置 2核 4核 8核 是否多线程 否 疋 否 疋 否 疋 疋 转换门限 - 10000 - 10000 - 100000 100000 每个线程的优
- 1000 - 1000 - 1000 1000 化次数 线程数 - 2 - 2 - 4 8
所用时长
683.85 403.97 780.4 609.58 1162.2 447.88 359.5
C分钟) 速度提高 - 40.93% - 21.98% - 61.46% 69.07% 从表 5中可以看出, 采用本发明实施例的多线程(2个、 4个和 8个)优化, 相对于单线程优化, 优化速度提高了 21.98%到 69.07%不等, 频率优化的效率 明显提高。 请参考图 5 ,本发明实施例还提供一种基于模拟退火算法的频率优化装置, 包括:
优化模块 501 , 用于在当前退火温度下, 基于当前状态解进行一次多线程 优化,一次多线程优化包括:两个以上线程中的每个线程分别选择一个小区组, 对选取的所述小区组进行一次基于模拟退火算法的频率优化;
判断模块 502 , 根据多线程间的约束条件判断是否接受所述优化模块本次 多线程优化得到的新状态解。
一种实施方式中, 还可以包括:
确定模块 503 , 用于根据频率优化的进度确定多线程间的约束条件。
进一步的, 所述确定模块 503 , 可以具体用于统计已进行的历次多线程优 化得到的新状态解不被接受的次数,判断所述不被接受的次数是否达到转换门 限; 若未达到转换门限, 确定采用乐观约束条件, 包括: 若多线程中的每个线 程都接受自己优化得到的新状态解,则判断为接受本次多线程优化得到的新状 态解, 否则判断为不接受; 若达到转换门限, 确定采用悲观约束条件, 包括: 若多线程中有且只有一个线程接受自己优化得到的新状态解,则判断为接受本 次多线程优化得到的新状态解, 否则判断为不接受。
如图 6所示,另一种实施方式中,还可以包括处理模块 504;本实施方式中: 所述判断模块 502, 还用于统计并判断当前退火温度下的优化次数是否达 到设定值;
所述处理模块 504, 用于若所述判断模块判断已达到设定值, 则结束当前 温度下的多线程优化, 将最后得到的被接受的新状态解作为当前状态解。
进一步的, 所述处理模块 504, 还可以用于若所述判断模块判断未达到设 定值,且本次多线程优化得到的新状态解被接受, 则将所述新状态接更新为当 前状态解,通知所述优化模块进行下一次多线程优化; 若所述判断模块判断未 达到设定值,且本次多线程优化得到的新状态解不被接受, 则结束当前温度下 的多线程优化,通知所述优化模块基于最后得到的被接受的新状态解进行单线 程优化。
更进一步的, 所述判断模块 502, 还可以用于判断是否达到频率优化退出 条件; 所述处理模块 504, 还可以用于若所述判断模块判断没有达到频率优化 退出条件, 则确定下一个退火温度; 若所述判断模块判断已达到频率优化退出 条件, 则结束频率优化。
本发明实施例的装置, 其中央处理器(CPU ) 包括 2个以上内核, 以便提 供 2个以上线程, 用于实施本发明实施例提供的基于模拟退火算法的频率优化 方法。
本发明实施例提供的装置,可以采用多个线程分别对各自选取的小区组进 行频率优化,并设定有多线程间的约束条件用于判断是否接受本次多线程优化 得到的新状态解,相对于现有的频率优化方法,可以有效提高频率优化的效率。 本领域普通技术人员可以理解上述实施例的各种方法中的全部或部分步 骤可以通过硬件完成,也可以通过程序来指令相关的硬件完成, 该程序可以存 储于一计算机可读存储介质中, 存储介质可以包括: 只读存储器、 随机存取存 储器、 磁盘或光盘等。
以上对本发明实施例所提供的基于模拟退火算法的频率优化方法以及装 阐述, 以上实施例的说明只是用于帮助理解本发明的方法及其核心思想, 不应 理解为对本发明的限制。
Claims
1、 一种基于模拟退火算法的频率优化方法, 其特征在于, 包括: 在当前退火温度下,基于当前状态解进行一次多线程优化, 一次多线程优 化包括: 两个以上线程中的每个线程分别选择一个小区组,对选取的所述小区 组进行一次基于模拟退火算法的频率优化;
根据多线程间的约束条件判断是否接受本次多线程优化得到的新状态解。
2、根据权利要求 1所述的方法, 其特征在于, 所述根据约束条件判断是否 接受本次多线程优化得到的新状态解之前还包括:
根据频率优化的进度确定多线程间的约束条件。
3、根据权利要求 2所述的方法, 其特征在于, 所述根据频率优化的进度确 定多线程间的约束条件包括:
统计已进行的历次多线程优化得到的新状态解不被接受的次数,判断所述 不被接受的次数是否达到转换门限;
若未达到转换门限, 确定采用乐观约束条件, 包括: 若多线程中的每个线 程都接受自己优化得到的新状态解,则判断为接受本次多线程优化得到的新状 态解, 否则判断为不接受;
若达到转换门限, 确定采用悲观约束条件, 包括: 若多线程中有且只有一 个线程接受自己优化得到的新状态解,则判断为接受本次多线程优化得到的新 状态解, 否则判断为不接受。
4、 根据权利要求 1至 3中任一项所述的方法, 其特征在于, 所述根据约束 条件判断是否接受本次多线程优化得到的新状态解之后还包括:
统计并判断当前退火温度下的优化次数是否达到设定值;
若已达到设定值, 则结束当前温度下的多线程优化,将最后得到的被接受 的新状态解作为当前状态解。
5、根据权利要求 4所述的方法, 其特征在于, 所述统计并判断当前退火温 度下的优化次数是否达到设定值之后还包括:
若未达到设定值,且本次多线程优化得到的新状态解被接受, 则将所述新 状态接更新为当前状态解, 进行下一次多线程优化; 若未达到设定值,且本次多线程优化得到的新状态解不被接受, 则结束当 前温度下的多线程优化, 基于最后得到的被接受的新状态解进行单线程优化。
6、 根据权利要求 4所述的方法, 其特征在于, 所述若已达到设定值, 则结 束当前温度下的多线程优化,将最后得到的被接受的新状态解作为当前状态解 之后还包括:
判断是否达到频率优化退出条件;
若没有达到, 则确定下一个退火温度;
若已达到, 则结束频率优化。
7、 一种基于模拟退火算法的频率优化装置, 其特征在于, 包括: 优化模块,用于在当前退火温度下,基于当前状态解进行一次多线程优化, 一次多线程优化包括: 两个以上线程中的每个线程分别选择一个小区组,对选 取的所述小区组进行一次基于模拟退火算法的频率优化;
判断模块,根据多线程间的约束条件判断是否接受所述优化模块本次多线 程优化得到的新状态解。
8、 根据权利要求 7所述的装置, 其特征在于, 还包括:
确定模块, 用于根据频率优化的进度确定多线程间的约束条件。
9、 根据权利要求 8所述的装置, 其特征在于:
所述确定模块,具体用于统计已进行的历次多线程优化得到的新状态解不 被接受的次数, 判断所述不被接受的次数是否达到转换门限; 若未达到转换门 限, 确定采用乐观约束条件, 包括: 若多线程中的每个线程都接受自己优化得 到的新状态解, 则判断为接受本次多线程优化得到的新状态解, 否则判断为不 接受; 若达到转换门限, 确定采用悲观约束条件, 包括: 若多线程中有且只有 一个线程接受自己优化得到的新状态解,则判断为接受本次多线程优化得到的 新状态解, 否则判断为不接受。
10、 根据权利要求 7至 9中任一项所述的装置, 其特征在于, 还包括: 处理 模块;
所述判断模块,还用于统计并判断当前退火温度下的优化次数是否达到设 定值; 所述处理模块, 用于若所述判断模块判断已达到设定值, 则结束当前温度 下的多线程优化, 将最后得到的被接受的新状态解作为当前状态解。
11、 根据权利要求 10所述的装置, 其特征在于:
所述处理模块,还用于若所述判断模块判断未达到设定值,且本次多线程 优化得到的新状态解被接受, 则将所述新状态接更新为当前状态解,通知所述 优化模块进行下一次多线程优化; 若所述判断模块判断未达到设定值,且本次 多线程优化得到的新状态解不被接受, 则结束当前温度下的多线程优化,通知 所述优化模块基于最后得到的被接受的新状态解进行单线程优化。
12、 根据权利要求 10所述的装置, 其特征在于:
所述判断模块, 还用于判断是否达到频率优化退出条件;
所述处理模块, 还用于若所述判断模块判断没有达到频率优化退出条件, 则确定下一个退火温度; 若所述判断模块判断已达到频率优化退出条件, 则结 束频率优化。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201180001710.0A CN103039100B (zh) | 2011-08-03 | 2011-08-03 | 基于模拟退火算法的频率优化方法和装置 |
PCT/CN2011/077958 WO2012167498A1 (zh) | 2011-08-03 | 2011-08-03 | 基于模拟退火算法的频率优化方法和装置 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2011/077958 WO2012167498A1 (zh) | 2011-08-03 | 2011-08-03 | 基于模拟退火算法的频率优化方法和装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012167498A1 true WO2012167498A1 (zh) | 2012-12-13 |
Family
ID=47295382
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2011/077958 WO2012167498A1 (zh) | 2011-08-03 | 2011-08-03 | 基于模拟退火算法的频率优化方法和装置 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN103039100B (zh) |
WO (1) | WO2012167498A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014089974A1 (zh) * | 2012-12-14 | 2014-06-19 | 上海大唐移动通信设备有限公司 | 一种用于通信系统的频率优化方法 |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP7108185B2 (ja) * | 2018-11-22 | 2022-07-28 | 富士通株式会社 | 最適化装置および最適化装置の制御方法 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030050067A1 (en) * | 2001-09-10 | 2003-03-13 | Jack Rozmaryn | Wireless systems frequency reuse planning using simulated annealing |
CN1889752A (zh) * | 2006-06-26 | 2007-01-03 | 西安交通大学 | 认知无线电系统中授权用户检测的频率分配方法 |
CN101593132A (zh) * | 2009-06-25 | 2009-12-02 | 北京航空航天大学 | 基于线程构造模块的多核并行模拟退火方法 |
CN101800995A (zh) * | 2010-01-21 | 2010-08-11 | 华为技术有限公司 | 小区关系确定方法及设备和频率规划方法及设备 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5796625A (en) * | 1996-03-01 | 1998-08-18 | Lsi Logic Corporation | Physical design automation system and process for designing integrated circuit chip using simulated annealing with "chessboard and jiggle" optimization |
CN101826167B (zh) * | 2010-03-31 | 2012-09-05 | 北京航空航天大学 | 基于云控制器的自适应多核并行模拟退火遗传方法 |
-
2011
- 2011-08-03 WO PCT/CN2011/077958 patent/WO2012167498A1/zh active Application Filing
- 2011-08-03 CN CN201180001710.0A patent/CN103039100B/zh active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030050067A1 (en) * | 2001-09-10 | 2003-03-13 | Jack Rozmaryn | Wireless systems frequency reuse planning using simulated annealing |
CN1889752A (zh) * | 2006-06-26 | 2007-01-03 | 西安交通大学 | 认知无线电系统中授权用户检测的频率分配方法 |
CN101593132A (zh) * | 2009-06-25 | 2009-12-02 | 北京航空航天大学 | 基于线程构造模块的多核并行模拟退火方法 |
CN101800995A (zh) * | 2010-01-21 | 2010-08-11 | 华为技术有限公司 | 小区关系确定方法及设备和频率规划方法及设备 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014089974A1 (zh) * | 2012-12-14 | 2014-06-19 | 上海大唐移动通信设备有限公司 | 一种用于通信系统的频率优化方法 |
Also Published As
Publication number | Publication date |
---|---|
CN103039100B (zh) | 2015-11-25 |
CN103039100A (zh) | 2013-04-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Deepjs: Job scheduling based on deep reinforcement learning in cloud data center | |
Qin et al. | Mooncake: A kvcache-centric disaggregated architecture for llm serving | |
CN110727685B (zh) | 一种基于Cassandra数据库的数据压缩方法、设备以及存储介质 | |
CN108776897B (zh) | 数据处理方法、装置、服务器及计算机可读存储介质 | |
CN108366047B (zh) | 基于博弈论的有源配电网数据安全高效传输优化方法及装置 | |
WO2015043528A1 (zh) | 并行多线程报文处理的方法和装置 | |
JP6576324B2 (ja) | 非均質システム用の分散ストレージ配分 | |
US20120060146A1 (en) | Automatic Application Tuning | |
WO2019109352A1 (zh) | 对设备的性能数据进行采样的方法和装置 | |
JP2021035055A (ja) | 無線周波数リソースの割り当て方法、装置、デバイスおよびシステム、ならびに記憶媒体 | |
WO2010012196A1 (zh) | 一种数据读写的方法和装置 | |
WO2012167498A1 (zh) | 基于模拟退火算法的频率优化方法和装置 | |
Liu et al. | Effective protocols for kNN search on broadcast multi-dimensional index trees | |
CN104778088B (zh) | 一种基于减少进程间通信开销的并行i/o优化方法与系统 | |
Liu et al. | An online approach to dynamic channel access and transmission scheduling | |
JP2016208504A (ja) | ネットワークパラメータの決定方法及び決定装置 | |
CN110069719B (zh) | 一种面向互联网环境的行为预测方法及其预测系统 | |
CN104202757B (zh) | 一种认知无线电网络性能最优检测方法 | |
WO2025039321A1 (zh) | 基于cpe-bft算法的区块链共识机制性能优化方法 | |
CN101730255A (zh) | 一种选择prach的方法及装置 | |
Yu et al. | Staleness aware semi-asynchronous federated learning | |
CN107636636A (zh) | 调节处理器核操作 | |
JP2012048387A (ja) | 検索処理方法および検索処理装置 | |
Yang et al. | Demystifying and Enhancing the Efficiency of Large Language Model Based Search Agents | |
CN118295978A (zh) | 文件导入方法、计算机装置、存储介质及程序产品 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201180001710.0 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11867385 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 11867385 Country of ref document: EP Kind code of ref document: A1 |