39 results about "Queueing theory" patented technology

The present application belongs to the technical field of parallel task scheduling. Specifically, disclosed are a heterogeneous computing resource scheduling method and apparatus based on multi-objective optimization. The method comprises: selecting at least two performance indicators from both a task dimension and a resource dimension as optimization objectives, and establishing a multi-objective optimization model for heterogeneous computing resource scheduling; converting a computation task request process into a computation task waiting model on the basis of a queuing theory; on the basis of an observed resource state, constructing a multi-task adaptive scheduling model based on reinforcement learning; on the basis of the multi-objective optimization model and the computation task waiting model, constructing a Markov decision process model from a multi-task-oriented heterogeneous computing resource scheduling problem and a resource mapping process; and on the basis of the Markov decision process model and the multi-task adaptive scheduling model, realizing adaptive multi-task heterogeneous computing resource scheduling. The embodiments of the present application solve the problem of it being difficult for a homogeneous computing resource scheduling method to adapt to heterogeneous computing resource scheduling.

Aspects of the present invention include setting task management and project management parameters; utilizing a queuing theory calculator for a task and a project management based on various types of multi-dimensional visualization; generating a plurality of results of the task management and the project management based on utilizing the queueing theory calculator; analyzing the results with at least one machine learning algorithm; outputting a multi-dimensional analysis of the results based on the analysis of the results with the at least one machine learning algorithm; analyzing the results using a magic square analysis; comparing the results with a plurality of requirements based on the analysis of the results with the at least one machine learning algorithm and the analysis of the results using the magic square analysis; and determining a subset of the plurality of results that meet a threshold of a comparison of the requirements.

The invention, which belongs to the technical field of traffic organization optimization, discloses a double-end queuing theory-based simulation and optimization method for a passenger area in a transportation junction online-hailing vehicle, comprising the following steps: acquiring a passenger arrival rate, a vehicle arrival rate, a service rate and the number of service desks used in the passenger area in the transportation junction online-hailing vehicle; according to the passenger arrival rate and the vehicle arrival rate, simulating a passenger area on the transportation junction online hailing vehicle through a Markov chain to obtain steady-state probability distribution, and according to the steady-state probability distribution, calculating to obtain an average queuing length of the vehicle and the passenger; the service intensity is calculated according to the dynamic parameters, the service intensity is judged, and the service rate and the number of service desks are adjusted according to the judgment result; the adjusted service rate and the number of service desks are restrained according to the passenger arrival rate and the vehicle arrival rate; and carrying out threshold judgment on the service intensity, and carrying out lag control on the number of the constrained service desks to be started according to a threshold judgment result so as to realize simulation optimization of the passenger area in the transportation junction online hailing car.

The invention discloses a performance prediction method for an intelligent computing system, and the method achieves the efficient and explainable performance prediction through mixed analysis modeling. The method mainly comprises the steps of performing tracking and programmed synthesis on a target deep learning model, and generating a final operator graph containing a complete calculation and communication operation sequence; a dynamic overlap analysis model based on available parallelism is adopted, independent calculation and memory access efficiency is predicted in combination with a machine learning method, and GPU operator execution delay is accurately estimated; based on a configurable network topology, applying a queuing theory network model, and carrying out white box type modeling on queuing and transmission delay of communication operation in each node in a network path; and the calculation and communication prediction delays of the operators are aggregated according to the scheduling characteristics of the distributed strategy to obtain the final system end-to-end delay. The method provided by the invention can ensure the precision while remarkably shortening the prediction time, and has the advantages of high efficiency, high interpretability and the like.

The invention provides a computing power resource scheduling method based on a queuing theory, and the scheduling method specifically comprises the steps: building a waiting system queuing model based on the queuing theory according to the execution condition of a historical computing task; receiving a task request of a calculation task, and carrying out idle hardware computing power resource judgment; when it is judged that the idle hardware computing power resources exist, the idle hardware computing power resources are allocated, and the current computing task is executed; when it is judged that the idle hardware computing power resources do not exist, a queue type is selected, the current computing task is added into a queue of the corresponding queue type, and the hardware computing power resources are waited to be released; and distributing the released hardware computing power resources according to the corresponding queuing sequence, and executing the current computing task. According to the method, computing power resource scheduling is carried out on the computing tasks through the waiting system queuing model, efficient utilization of resources is guaranteed, when the computing tasks cannot be processed immediately, the waiting time of the computing tasks can be shortened through reasonable queue setting and management, and the computing efficiency is improved.

The invention relates to a traffic signal control system and method, and relates to the technical field of traffic control, and the method comprises the steps: collecting traffic flow data through multi-source sensing equipment and V2X communication, and carrying out the preprocessing to construct a traffic feature matrix; a hybrid modeling framework is adopted, a queue theory, a cellular automaton and deep reinforcement learning are integrated, a traffic state is evaluated, and a signal control strategy is optimized. Through a double-closed-loop feedback mechanism, the inner loop uses a PID controller to finely adjust the phase, and the outer loop learns online to update reinforcement learning proxy parameters. And finally, an optimization strategy is executed, the variable lane control equipment is linked, and a signal state and a vehicle speed guide instruction are pushed to the vehicle-mounted terminal, so that remarkable beneficial effects can be generated in the aspects of improving traffic efficiency, reducing congestion, improving safety and the like.

The invention discloses an intelligent load balancing method based on dynamic scheduling, and the method comprises the following steps: 1, determining the weight of a server through employing a weighted polling algorithm, distributing different weights according to the load capability of the server, and guaranteeing the load balancing; step 2, dynamically adjusting request distribution by adopting a response time weighting algorithm so as to dynamically adjust the request distribution according to the response time of the server and optimize the system performance; step 3, performing hash calculation through an IP hash algorithm; and 4, carrying out real-time monitoring and dynamic scheduling. According to the intelligent load balancing method based on dynamic scheduling, multiple load balancing algorithms such as weighted polling, shortest connection priority, response time weighting and IP hash are utilized, the queue theory and real-time data analysis are combined, the defects in a traditional load balancing method are effectively overcome, and the load balancing efficiency and the resource utilization rate of the system are improved.

The invention relates to the technical field of elevator module updating, and discloses an AI-based elevator updating control method and system.The method comprises the steps that prediction is conducted on the basis of operation state data to obtain an adjustment factor, elevator waiting time is calculated in combination with a pre-built waiting time model, the elevator service state can dynamically reflect operation differences, and the elevator updating control accuracy is improved. And then the elevator coverage area of the target elevator group is divided according to the elevator waiting time, the migration load value is evaluated based on a queuing theory model, a dynamic upgrading path diagram is constructed, the recognition capacity of task transfer capacity differences among the elevators is enhanced, and finally the elevator upgrading sequence is determined with the migration load value minimization as the target based on a graph search algorithm. Upgrading tasks are distributed according to the sequence with the minimum system overall service influence when a plurality of elevator units need to be updated, and therefore the problems of overload, task overstock or service blind areas caused by improper scheduling are avoided.

The invention discloses a bank cloud counter teller scheduling method and device, electronic equipment and a computer program product. The method comprises the steps of obtaining historical customer arrival data; according to the historical customer arrival data, a customer arrival quantity prediction algorithm is utilized to determine a customer arrival quantity estimated value in a future time period; based on a queuing theory model, an integer programming problem is constructed according to the customer arrival quantity estimated value and cloud counter teller service data; the integer programming problem is converted into a QUBO model; and solving the QUBO model by using a quantum solving algorithm to obtain the number of tellers in each scheduling time period in the future. According to the method, a cloud counter seat scheduling problem is abstracted into a nonlinear programming problem based on a queuing theory model, a QUBO model and a quantum computing technology are applied to bank cloud teller scheduling optimization for solving the programming problem, the powerful computing power of the quantum computing technology is utilized, the problem that a classical algorithm falls into a local optimal solution is avoided, and the scheduling efficiency is improved. And a high-quality solution is provided for an integer programming problem of NP-difficulty.

The invention relates to a large-throughput low-delay access and scheduling method for industrial internet platform equipment data, which comprises the following steps of: separately processing real-time data and historical data reported by edge equipment, and respectively processing the real-time data and the historical data through a real-time data processor and a historical data processor according to the data volume characteristics of the two types of data; and dividing different computing resources to the real-time data processor and the historical data processor. For a real-time data processor, a multi-stage feedback queue task scheduling mechanism is adopted, so that real-time data with smaller data packets can be processed preferentially, real-time data with larger data packets can be processed in time, and data loss caused by overlong waiting time of the smaller data packets is avoided. For a historical data processor, an M / M / C / infinity / infinity / FCFS queuing model in a queuing theory (stochastic service system theory) is adopted to dynamically adjust the size of computing resources, and it is ensured that when queuing is too long, historical data can be processed in time.

The present invention discloses a method for calculating the rapid distribution of traffic in the throat area of ​​railways, and belongs to the field of railway transportation technology. A solution is provided for the problem that traditional methods rely on manual experience and the existing model calculation is complex. The steps are as follows: input relevant data; construct a traffic distribution optimization model, balance the goals such as the difference between the plan and the actual train, and set constraints; discretize and segment the passenger train ratio parameters, and use the "big M method" to convert the nonlinear equation into a linear constraint; construct a capacity simulation model based on queuing theory, pre-calculate the capacity value to generate a capacity table; combine the linearized model and the capacity table, solve with the CPLEX solver, and obtain the optimized distribution result. The method provided by the present invention realizes rapid and accurate distribution of traffic, balances the difference between the plan and the actual, can quickly respond to dynamic demands, and provides a scientific basis for the management of traffic in the throat area.

This invention provides a dynamic network routing optimization method and system based on deep learning prediction, comprising: constructing a graph network with heterogeneous nodes based on network topology and traffic; generating a large number of network traffic performance indicators based on queuing theory algorithms based on the heterogeneous graph information; pre-training a deep learning model based on the heterogeneous graph network and network traffic data with pseudo-labels; fine-tuning the model by combining a small amount of network traffic data with real labels; selecting a better-performing routing optimization scheme using the trained network prediction model; and a dynamic network routing system in which a global monitoring device calculates and distributes routing optimization results to different forwarding devices based on the current network topology, traffic data of different forwarding devices, and the performance prediction model. This invention uses weakly supervised learning and pre-training fine-tuning methods to build a reliable network performance prediction model with low data cost and designs a routing optimization method to meet real-time dynamic routing strategies.

The present invention belongs to the technical field of network function virtualization, and specifically relates to a VNF deployment method based on coupled bandwidth allocation and with delay QoS guarantee; abstracting a physical network, obtaining a SFC to be deployed, and pre-allocating resources according to the characteristics of arriving traffic based on queuing theory and large deviation theory; sorting physical nodes in descending order according to the size of remaining CPU resources, screening candidate physical nodes and candidate physical links respectively according to pre-allocated CPU resources and bandwidth resources, estimating the coupled bandwidth after coupling new traffic with existing traffic of the candidate link based on martingale theory, and solving the coupled bandwidth increment of the candidate link; solving the joint optimization problem of satisfying delay QoS and minimum resource occupancy rate, and obtaining the optimal deployment plan; the method can allocate resources according to the arrival characteristics of data packets requested by the service, and achieve delay QoS guarantee while minimizing resource occupancy rate; in addition, the method can also fully save network resources and achieve the same SFC carrying capacity with less consumption.

The application discloses a kind of based on vacation queuing application service dynamic deployment and updating method.The method considers an edge computing system consisting of mobile device, edge server, cloud server, the key performance indicators of system are analyzed by queuing theory related model, deduce closed form expression including but not limited to edge server average delay, the probability that application is deployed on edge server, busy period probability and other related indexes, efficient dynamic application deployment and updating strategy is proposed.Optimization problem is constructed with the goal of minimizing average system delay, where storage space and energy consumption are constrained.The method described in the application constructs an interior point convex approximation optimization method to determine user task offloading probability, application response threshold, application waiting duration, edge server computing resource allocation, so as to minimize the average delay of the entire system.

The present invention discloses a high-efficiency transmission mode selection method and device for a dual-queue short packet system, the method comprising the following steps: according to queuing theory and short packet transmission theory, respectively obtaining the average peak information age (PAoI) gain of the short packet domain multiplexing mode and the short packet domain diversity mode compared to the single-queue system; according to the average peak information age (PAoI) gain and the transfer probability of the data packet in the dual-queue multiplexing mode, respectively deriving the signal-to-noise ratio threshold for switching the high-efficiency transmission mode of the data packet update under high-rate and low-rate conditions; and selecting the transmission mode according to the signal-to-noise ratio threshold. The present invention selects the high-efficiency transmission mode through the signal-to-noise ratio threshold, and reflects the effectiveness of the high-efficiency transmission mode selection method according to the average peak information age (PAoI).

The invention discloses a strip mine transportation scheduling model optimization method, system, equipment and medium considering a large key equipment fault, and the method comprises the steps: building an M / M / m queuing model according to a queuing theory principle when the mine large key equipment has a fault, taking the queuing length of a whole transportation system truck as an optimization index, solving the model through employing an iteration method, and solving the model according to the optimization index; determining the optimal number of trucks capable of meeting the production requirements during the equipment failure period; an available truck set is determined again, and a rescheduling model with the maximum yield in the fault time, the minimum idle time of the electric shovel or the crusher and the minimum grade fluctuation deviation is constructed; according to the characteristics of the established model, a differential evolution algorithm is improved by adopting a two-stage optimization mechanism, and a multi-objective solving algorithm is designed on the basis of the improved differential evolution algorithm; solving the rescheduling model to obtain an optimal scheduling scheme; the system, the equipment and the medium are used for implementing the method. The operation continuity of the mine transportation system is guaranteed, the transportation efficiency is improved, and the transportation cost is reduced.

This invention relates to the field of text processing technology and discloses a method for structuring unstructured information based on multi-agent systems. The method includes: outputting candidate data definition languages ​​and multi-table distribution samples from a generator, statistically projecting them to obtain discrete statistics; constructing a query plan cost phase transition surface based on the statistics, and calculating the steady-state probability of each plan using a Markov steady-state model combined with a plan caching mechanism; calculating the high-quantile tail delay using a queuing theory model, and solving for the second derivative of the tail delay with respect to the distribution perturbation in the minimum transport direction in Wasserstein space to obtain the optimal transmission bifurcation buffer congestion curvature; finally, updating the generator with minimizing this curvature as a constraint and outputting the table structure. This invention effectively reduces the risk of performance avalanche caused by data perturbations and significantly improves database robustness.

The invention relates to the technical field of text processing, and discloses a multi-agent-based unstructured information structured extraction method, which comprises the following steps of: outputting a candidate data definition language and a multi-table distribution sample through a generator, and carrying out statistical projection on the candidate data definition language and the multi-table distribution sample to obtain discrete statistics; constructing a query plan cost phase change surface based on the statistics, and calculating the steady-state probability of each plan in combination with a Markov steady-state model of a plan cache mechanism; calculating a high-quantile tail delay by using a queuing theory model, and solving a second derivative of the tail delay about distributed perturbation in the minimum carrying direction of the Warisstein space to obtain an optimal transmission bifurcation cache congestion curvature; and finally, a generator is updated by taking the minimum curvature as a constraint, and a table structure is output. According to the method, the performance avalanche risk caused by data perturbation can be effectively reduced, and the robustness of the database is remarkably improved.

The present invention provides a method for calculating the frame collision rate of frames on a CAN bus and a method for minimizing CAN bus resources. The method for minimizing CAN bus resources includes: S1, setting an initial CAN bus communication rate; S2, calculating the frame collision rate of frames of each priority that operate systematically on the CAN bus based on the Erlang B formula in queuing theory at the preset rate in step S1; S3, determining whether the frame collision rate of frames of each priority calculated in step S2 exceeds the upper limit of the allowable frame collision rate for frames of that priority. If it exceeds the upper limit, increase the CAN bus communication rate in step S1 and proceed to step S2 to continue the cycle. If it does not exceed the upper limit, decrease the CAN bus communication rate in step S1 and proceed to step S2 to continue the cycle; S4, outputting the minimum communication rate corresponding to not exceeding the upper limit of the allowable frame collision rate for frames of each priority to complete the optimization. The present invention provides an algorithm for the frame collision rate, and further calculates the required minimum bus resources from the upper limit of the frame collision rate of frames of each priority to optimize the resources.

This invention discloses an optimized scheduling method for electric vehicle (EV) charging based on queue theory. The EV charging process is divided into an input process, a queuing process, and a service process. When an EV connects to the power grid with a discrete Poisson probability distribution, it enters the queuing process. When the number of EVs connected to the grid is small or the grid load is low, EVs can charge directly without waiting. However, when the number of EVs connected to the grid is large or the grid load is at its peak, available grid power resources are limited, requiring queuing. This invention optimizes the queuing process, effectively balancing efficiency and fairness in the EV charging process, significantly reducing queue length and waiting time at charging stations.

The invention discloses a joint optimization method for DNN partition and resource allocation, which realizes collaborative reasoning and resource optimization of a deep neural network in an environment of a single edge server and a plurality of mobile devices, and comprises the following steps of: determining DNN hierarchy segmentation points of each device and the number of computing resource blocks and wireless resource blocks purchased from the edge server; decoupling modeling is performed on a DNN structure and an edge environment through a heterogeneous graph attention network, an optimal partition point and a resource deployment strategy are determined through Actor-Critic under the constraint of an M / M / 1 queuing theory model, and an edge-end cooperative reasoning scheme with price-demand-time delay closed-loop optimization is formed.

The invention relates to a charging station site selection decision method and device based on machine learning waiting time prediction, and a medium. The site selection decision method comprises the steps of obtaining candidate stations, construction cost, mixed traffic flow requirements and road network parameters; initializing a site selection scheme, completing preliminary distribution of electric vehicle flow in the mixed road network, and calculating driving time; calculating the arrival rate of each station by using a flow result, and calling a pre-training machine learning model to estimate the charging waiting time in combination with the pile number; overlapping the waiting time and the driving time into total travel time, and adjusting the traffic flow through user equalization iteration until the network is re-equalized; and with the purpose of minimizing the sum of the total travel time in the equilibrium state and the construction cost, synchronously optimizing the site selection and the pile number under the constraint of the demand and the capacity, and outputting a final decision. Compared with the prior art, the method has the advantages that machine learning and a traffic model are combined, the waiting time is estimated more accurately, estimation errors of a traditional queuing theory model are avoided, and the utilization efficiency of infrastructures and the travel experience of users are improved.

The invention discloses a server-free platform-oriented random workflow performance optimization method, which comprises a task random workflow sorting stage: according to a basic principle of a queuing theory model, performing arrival sorting on tasks of random workflows on the basis of an arrival sequence and a topological structure of the random workflows; in the instance parameter calculation stage, the blocking probability is calculated through an Erland B formula, the cold start probability is deduced, modeling is conducted through the semi-Markov process in the queuing theory, and performance indexes under the steady state condition are calculated through the parameters so as to evaluate the performance of the platform; and a system performance analysis stage: according to different task parameters, comprehensively analyzing the influence of the different task parameters on the performance of the function task, and based on the influence, giving out estimated resource overhead so as to realize cost minimization. According to the method, a processor distribution mode and constraint conditions are studied, and a demand processor combination prediction model is constructed by using a queuing theory model M / N / m / infinity for the problem of inaccurate prediction caused by strong uncertainty of a cloud center.

The invention relates to the technical field of edge computing and data distribution transmission, in particular to an information processing board card multi-channel data distribution transmission method based on edge computing. The system comprises a data portrait acquisition module, a queuing theory simulation module, a risk assessment module and a strategy optimization module. The system constructs prediction delay distribution by using a queuing theory simulation function by acquiring a static business portrait, predicting an arrival rate and a channel state; the core of the method is that a timeliness failure probability is calculated by combining prediction delay and a failure benchmark, and a quantitative task decision risk score is constructed; when the risk score exceeds a preset threshold value, the system starts a constraint optimization process, and generates and issues an optimal shunting strategy; according to the invention, through digital twinborn simulation, conversion from passive response to congestion to active prediction of delay distribution and risk decision is realized.

The application relates to the technical field of elevator module updating, and discloses an elevator updating control method and system based on AI, which comprises the following steps: obtaining an adjustment factor by prediction based on running state data, combining a pre-constructed waiting time model to calculate elevator waiting time, enabling the elevator service state to dynamically reflect running differences, dividing the elevator coverage area of a target elevator group through the elevator waiting time, evaluating a migration load value based on a queuing theory model, constructing a dynamic upgrading path graph, enhancing the identification capability of the task transfer capability difference between elevators, and finally determining an elevator upgrading sequence based on a graph search algorithm, taking minimizing the migration load value as the target, so that the upgrading tasks are distributed in the order of minimizing the overall system service influence when multiple elevator units need to be updated, thereby avoiding the problems of overload, task backlog or service blind area caused by improper scheduling.

The application relates to the field of airport management and aviation services, and discloses an airport check-in and consignment counter configuration planning method and device, which comprises the following steps: S1, collecting real-time or historical data to obtain a passenger arrival rate, a counter service rate and system limitation condition parameters; S2, constructing a queuing theory model based on the collected data, and preliminarily evaluating the check-in counter configuration through an M / M / c queuing model; and S3, using optimal control theory to establish a target function, optimizing the configuration scheme, and minimizing the comprehensive target among the queuing time, the passenger waiting time and the resource configuration cost. The dynamic counter configuration scheme based on the real-time data acquisition and feedback mechanism achieves the technical effect of dynamically adjusting the counter quantity during the peak period, thereby effectively avoiding the problem of long passenger queues during the peak period and reducing the resource waste during the trough period.

The invention discloses an RIS-assisted D2D cache network task unloading modeling method, and belongs to the technical field of queue modeling, a reconfigurable intelligent surface (RIS) technology is combined with a D2D cache network, content unloading comprises a self-unloading method and a D2D unloading method, the situation that an auxiliary cache technology combines the dual characteristics of user terminal cache and D2D communication is considered, and the content unloading efficiency is improved. In other words, content unloading comprises a self-unloading method and a D2D unloading method, and content transmission needs to meet double constraints of a receiver signal-to-interference ratio and a D2D distance. A randomly generated service model is established based on a queuing theory, and a queuing model is established by combining the performance advantages of RIS and the advantages of a D2D cache network. And establishing an M / G / 1 queuing model for the cache queue of the user, and analyzing the conversion of the queue state. Then, constructing a Markov chain containing a queue length state and a transmission state, and deriving a closed expression of a successful unloading probability, an average queue length of a cache network and average waiting time;

The invention discloses a queuing theory model-based futures counter front-end selection optimization method, relates to the field of futures counter front-end selection optimization, solves the problem that a futures counter front-end system cannot be effectively optimized in real time, and comprises the steps of determining distribution of an input process and a processing process according to historical futures orders; constructing a queuing rule, a processing rule and a queuing theory model in the futures counter front-end system; order parameters in the futures counter front-end system are calculated based on a queuing theory model; selecting an actual futures counter front-end system according to order parameters in the futures counter front-end system, and performing preliminary analysis on the actual futures counter front-end system; and calculating an order change rate threshold value corresponding to the real-time futures order change rate, and optimizing the actual futures counter front-end system. The real-time effective optimization of the futures counter front-end system is realized.

The invention provides a task-aware end-side cloud cooperative computing unloading method and device based on a queuing theory, and the method comprises the steps: obtaining target tasks, classifying the target tasks according to the time delay sensitivity, and obtaining the types of the target tasks; a cooperative computing model is obtained, the cooperative computing model comprises at least two cooperative function layers, the cooperative function layers are used for processing the target tasks of the corresponding categories, and the cooperative function layers are provided with computing nodes with different computing power; based on a queuing system of the collaborative function layer, determining a queuing state of a target task which is queued and executed in the collaborative function layer; according to the queuing state and the type of the target task, an unloading strategy of the target task is determined, and the unloading strategy is used for indicating an unloading path of the target task. The processing delay of the target task can be reduced.

This invention relates to a reliable transmission method for RDMA cross-WAN transmission, comprising the following steps: Step 1: Detecting and obtaining the basic physical packet loss rate and the average duration of micro-burst packet loss of the WAN link; Step 2: Constructing a probability-based lower bound model for packet loss recovery theoretical requirements; Step 3: Constructing a nonlinear congestion penalty model based on queuing theory; Step 4: Multiplying the ideal number of redundant packets by the congestion resistance penalty coefficient to obtain a deterministic analytical solution for the theoretically optimal number of redundant packets; Step 5: Deriving the safe time interval during which redundant packets must be delayed; Step 6: After the sending end continuously sends k original data packets on the underlying physical link, time-domain interleaving is achieved; This scheme enables the redundant packets to be physically isolated from the original data stream on the time axis, effectively avoiding the long-tail coverage effect of micro-burst packet loss.