Task processing method and device and cloud computer terminal
Through improved standard linear Kalman filters to predict the terminal location of cloud computers and allocate tasks to multiple edge servers, the problem of low success rate of task processing in high dynamic scenarios is solved, and the overall time Delay and higher task processing success rate.
Patent Information
- Application Number
- CN202410995135.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-07-24
AI Technical Summary
Traditional MEC task management methods lead to a low success rate of task processing results in high dynamic scenarios of cloud computer terminals, mainly due to additional time delays caused by interruption of wireless communication links and task reallocation.
Improve task processing efficiency by using improved standard linear Kalman filters on cloud computer terminals, filtering the available edge servers around the target location and assigning tasks to multiple edge servers based on transmission delay and connectivity to improve task processing efficiency and success rate.
It effectively reduces the overall delay, reduces the impact of path loss and shadow effects on task allocation, and improves the success rate of cloud computer terminals to collect task processing results, especially in medium-speed and high-speed mobile scenarios.
Smart Images

Figure CN119030975A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of communication technology, and in particular to a task processing method, device and cloud computer terminal. Background Art
[0002] Compared with traditional computers, cloud computers do not have hardware such as central processing units, memory and hard disks. All these hardware are gathered in the cloud data center. Users only need a small terminal device to access the network anywhere there is a network, connect the keyboard, mouse and monitor, and access their personal desktop, data and various applications, just like using an ordinary personal computer.
[0003] Cloud computers are provided in the form of cloud services through the Internet. In cloud services, the computing process is performed by servers and databases on the cloud. This is called cloud computing. In cloud computing, the terminal is connected to a remote cloud server. Due to the long connection distance, this causes communication delays. Therefore, cloud computing is not suitable for situations that require real-time processing. In order to connect the server outside the terminal to achieve low-latency processing, Mobile Edge Computing (MEC) came into being. In MEC, a server with computing resources is called an edge server (ES), which is placed at the edge of the network such as base stations of cellular networks and access points of wireless LANs. ES can not only provide computing resources and reduce latency for cloud computer terminals, but also directly process tasks assigned by cloud computer terminals.
[0004] Traditional MEC task management methods can only adapt to static or low-dynamic scenarios. When the cloud computer terminal moves at medium or high speed, the wireless communication link between it and the ES is often interrupted during the ES task processing process, resulting in the cloud computer terminal being unable to collect task processing results from the ES. Traditional MEC task management methods generally reschedule tasks to the ES closest to the task request location for processing. This will take extra time to process, resulting in a low success rate for collecting task processing results. Summary of the invention
[0005] The purpose of the embodiments of the present application is to provide a task processing method, device and cloud computer terminal to solve the problem that the existing MEC task management method leads to a low success rate of the cloud computer terminal in collecting task processing results.
[0006] In order to solve the above problems, the present application provides a task processing method, which is applied to a cloud computer terminal, and the method includes:
[0007] In the case that the cloud computer terminal has a first task that needs to be processed, determining a plurality of available edge servers around a target location according to connectivity between the cloud computer terminal and the edge server; wherein the target location is the location of the cloud computer terminal;
[0008] Allocating the first task to at least two of the available edge servers according to a transmission delay between a cloud computer terminal and an available edge server and connectivity between the cloud computer terminal and the available edge server;
[0009] Receive task processing results returned by the at least two edge servers after processing the tasks.
[0010] Wherein, before determining a plurality of available edge servers around the target location according to the connectivity between the cloud computer terminal and the edge server, the method further comprises:
[0011] Detecting motion state information of the cloud computer terminal, wherein the motion state information includes at least one of motion speed, acceleration, and jerk;
[0012] When it is determined according to the motion state information that the cloud computer terminal is in a high dynamic scene, using an improved standard linear Kalman filter to predict the position of the cloud computer terminal to obtain a predicted position, and using the predicted position as the target position;
[0013] Alternatively, when it is determined according to the motion state information that the cloud computer terminal is in a static scene or a low-dynamic scene, the current position of the cloud computer terminal is used as the target position.
[0014] The process equation of the state space model of the improved standard linear Kalman filter is represented by a first formula, and the linear observation equation of the state space model of the improved standard linear Kalman filter is represented by a second formula;
[0015] The first formula is:
[0016] Where F represents the transition matrix; v k Represented as additive process noise, it follows a normal distribution Q k is represented as the covariance matrix of the process noise; x k represents the state vector at the current moment; x k-1 Represents the state vector at the previous moment; T s Indicates the sampling period;
[0017] The second formula is: k =Hx k +n k;
[0018] Among them, y k represents the observed value; H = [1 0 0 0] is the observation transformation matrix; x k Represents the state vector at the current moment; n k represents observation noise, which follows a normal distribution R k Represents the covariance matrix of the observation noise.
[0019] The improved standard linear Kalman filter adjusts the covariance matrix Q of the process noise k To control the Kalman gain K k slowly decreasing from a first steady-state value to a second steady-state value;
[0020] Among them, Q k =q k I;q k is the process noise factor, I is the unit matrix; k Slowly decrease from the first value to the second value according to the third formula; the third formula is:
[0021] q k =q max ·ω k-ks ,k∈[k s ,k e ];
[0022] k s and k e They represent the start and end time of adjustment respectively, and ω is q k The attenuation factor, q k is the process noise factor; q k In k s The value of the moment is the first value; q k In k e The value of the moment is the second value.
[0023] The method of determining a plurality of available edge servers around a target location based on connectivity between a cloud computer terminal and an edge server includes:
[0024] It is determined that the edge servers corresponding to the connectivity higher than the second threshold are multiple available edge servers around the target location.
[0025] The allocating the first task to at least two of the available edge servers according to the transmission delay between the cloud computer terminal and the available edge server and the connectivity between the cloud computer terminal and the available edge server comprises:
[0026] According to the fourth formula, the first task is decomposed into N tasks, and the N tasks are respectively assigned to the N available edge servers; wherein the fourth formula is:
[0027]
[0028] in, T all =max(t1+t d,1 ,t2+t d,2 ,...,t N +t d,N ); N is the number of edge services available for the assigned task; C(d i ) is the connectivity between the cloud computer terminal and the i-th edge server; p i is the computing power of the ith edge server; t i The time required for computing the assigned task for the i-th edge server; T all is the maximum time that the cloud computer terminal can collect all the calculation results at one time; t d,i is the transmission delay between the cloud computer terminal and the i-th edge server; M sum This is the first task.
[0029] Wherein, the method further comprises:
[0030] Determine the interruption probability P between the cloud computer terminal and the first edge server as a function of the distance d between the cloud computer terminal and the first edge server Outage (d);
[0031] According to the interruption probability P Outage (d), determining connectivity C(d) between the cloud computer terminal and the first edge server;
[0032] The first edge server is any edge server or any optional edge server.
[0033] The interruption probability P between the cloud computer terminal and the first edge server is determined according to the function of the distance d between the cloud computer terminal and the first edge server. Outage (d) including:
[0034] According to the transmission power P of the cloud computer terminal Tx , the path loss L between the cloud computer terminal and the first edge server and the shadow gain G caused by the log-normal shadow, determine the receiving power P of the cloud computer terminal Rx (d);
[0035] According to the receiving power P of the cloud computer terminalRx (d) and the noise power P N , determine the signal-to-noise ratio SNR(d);
[0036] According to the signal-to-noise ratio SNR(d), the interruption probability P is obtained. Outage (d); The interruption probability P Outage (d) is the probability that the signal-to-noise ratio SNR(d) is lower than the first threshold.
[0037] The present application also provides a task processing device, which is applied to a cloud computer terminal, and the device includes:
[0038] a determination module, configured to determine, when the cloud computer terminal has a first task that needs to be processed, a plurality of available edge servers around a target location according to connectivity between the cloud computer terminal and the edge server; wherein the target location is the location of the cloud computer terminal;
[0039] an allocation module, configured to allocate the first task to at least two of the available edge servers according to a transmission delay between the cloud computer terminal and the available edge servers and connectivity between the cloud computer terminal and the available edge servers;
[0040] The receiving module is used to receive the task processing results returned after the at least two edge servers can be used to process the tasks.
[0041] An embodiment of the present application also provides a cloud computer terminal, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor implements the task processing method described above when executing the program.
[0042] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the task processing method described above are implemented.
[0043] An embodiment of the present application also provides a computer program product, including computer instructions, which implement the steps of the task processing method described above when executed by a processor.
[0044] The above technical solution of the present application has at least the following beneficial effects:
[0045] In the task processing method, device and cloud computer terminal of the embodiment of the present application, multiple available edge servers around the target location are screened out according to connectivity, and then a distributed computing method is used to efficiently allocate tasks to multiple available edge servers according to transmission delay and connectivity; on the one hand, in view of the low latency requirement of the cloud computer terminal, the embodiment of the present application allocates a task to multiple available edge servers around for processing, and optimizes the transmission delay between each edge server and the cloud computer terminal in the allocation algorithm, thereby effectively reducing the overall delay; on the other hand, in view of the path loss and shadow effect existing in the wireless communication environment, the allocation algorithm can reduce the impact of path loss and shadow effect on task allocation taking into account the connectivity between each edge server and the cloud computer terminal; therefore, while efficiently allocating tasks to multiple available edge servers around, the embodiment of the present application can also effectively reduce the overall delay and reduce the impact of path loss and shadow effect, thereby effectively improving the success rate of collecting task processing results. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 A flowchart showing the steps of the task processing method provided in an embodiment of the present application;
[0047] Figure 2 A schematic diagram showing the main flow of the task processing method provided in an embodiment of the present application;
[0048] Figure 3 represents the process noise factor q in the task processing method provided in the embodiment of the present application k Schematic diagram of the transition plan;
[0049] Figure 4 A schematic diagram showing task allocation in a task processing method provided in an embodiment of the present application;
[0050] Figure 5 It shows one of the simulation example diagrams in the embodiment of the present application;
[0051] Figure 6 It shows the second simulation example diagram in the embodiment of the present application;
[0052] Figure 7 It represents the third simulation example diagram in the embodiment of the present application;
[0053] Figure 8 A schematic diagram showing the structure of a task processing device provided in an embodiment of the present application;
[0054] Fig. 9 A schematic diagram showing the structure of a cloud computer terminal provided in an embodiment of the present application. DETAILED DESCRIPTION
[0055] In order to make the technical problems, technical solutions and advantages to be solved by the present application clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.
[0056] like Figure 1 As shown, an embodiment of the present application provides a task processing method, which is applied to a cloud computer terminal, and the method includes:
[0057] Step 101, when the cloud computer terminal has a first task that needs to be processed, multiple available edge servers around a target location are determined according to the connectivity between the cloud computer terminal and the edge server; wherein the target location is the location of the cloud computer terminal;
[0058] Step 102, allocating the first task to at least two of the available edge servers according to the transmission delay between the cloud computer terminal and the available edge servers and the connectivity between the cloud computer terminal and the available edge servers;
[0059] Step 103: Receive the task processing results returned by the at least two edge servers after processing the tasks.
[0060] Optionally, step 103 specifically includes that when all the edge servers that are assigned tasks have processed the assigned tasks, the task processing structure will be sent to the cloud computer terminal for aggregation in turn; if the cloud computer terminal has new tasks that need to be requested for processing, the next round of loop will be started, that is, steps 101 to 103 will be executed in a loop.
[0061] On the one hand, the embodiment of the present application targets the low latency demand of cloud computer terminals. The embodiment of the present application allocates a task to multiple available edge servers in the surrounding area for processing, and optimizes the transmission delay between each edge server and the cloud computer terminal in the allocation algorithm, thereby effectively reducing the overall latency. On the other hand, the embodiment of the present application targets the path loss and shadow effect existing in the wireless communication environment. Taking into account the connectivity between each available edge server and the cloud computer terminal, the allocation algorithm can reduce the impact of path loss and shadow effect on task allocation. Therefore, while efficiently allocating tasks to multiple available edge servers in the surrounding area, the embodiment of the present application can also effectively reduce the overall latency and reduce the impact of path loss and shadow effect, thereby effectively improving the success rate of collecting task processing results.
[0062] In at least one embodiment of the present application, before step 101, the method further includes:
[0063] Detecting motion state information of the cloud computer terminal, wherein the motion state information includes at least one of motion speed, acceleration, and jerk;
[0064] When it is determined according to the motion state information that the cloud computer terminal is in a high dynamic scene, an improved standard linear Kalman filter (KF) is used to predict the position of the cloud computer terminal to obtain a predicted position, and the predicted position is used as the target position;
[0065] Alternatively, when it is determined according to the motion state information that the cloud computer terminal is in a static scene or a low-dynamic scene, the current position of the cloud computer terminal is used as the target position.
[0066] That is to say, the task processing method provided in the embodiment of the present application is applicable to both cloud computer terminals in static scenes or low-dynamic scenes and cloud computer terminals in high-dynamic scenes. For cloud computer terminals in static scenes or low-dynamic scenes, the target position in step 101 is the current position of the cloud computer terminal, which can also be called the position where the cloud computer terminal confirms that there is a first task that needs to be requested for processing; for high-dynamic scenes, the target position in step 101 is the predicted position of the cloud computer terminal.
[0067] like Figure 2 As shown, the main process of the task processing method provided in the embodiment of the present application includes: when the cloud computer terminal needs to request a task to be processed by the edge server, the cloud computer terminal or the edge server detects the motion state of the current cloud computer terminal, that is, the current speed v, acceleration α and jerk j. th , acceleration threshold α th , jerk threshold j th To determine whether the cloud computer terminal is in a high-dynamic scene. th , α≥α th , j ≥ j th It is determined to be a high-dynamic scene. The mobile position of the cloud computer terminal is predicted by using an improved standard linear Kalman filter; the available edge servers around the predicted position are selected based on connectivity; the tasks are efficiently assigned to the available edge servers based on transmission delay and connectivity, and finally the edge servers can be used to process the tasks and return the results. On the contrary, if v = 0km / h, α = 0km / h 2 , j = 0 km / h 3 , then the cloud computer terminal is judged to be in a static scene; if v <v th , α<α th , j <j th, then the cloud computer terminal is determined to be in a low-dynamic scene. When the current scene is static or low-dynamic, the step of using the improved standard linear Kalman filter to predict the location of the cloud computer terminal is skipped, and the available edge servers around the predicted location are directly selected based on connectivity, and the tasks are efficiently assigned to the available edge servers based on transmission delay and connectivity. Finally, the edge servers can be used to process the tasks and return the results. After the current task processing is completed, if the cloud computer terminal has new tasks to request processing, the next round of loop is started.
[0068] In summary, in order to solve the problem of low success rate of collecting task processing results in high-dynamic scenarios in traditional MEC task management methods, the task processing method provided by the embodiment of the present application first predicts the mobile position of the cloud computer terminal by using the improved standard linear KF in high-dynamic scenarios, then selects the available edge servers around the predicted position according to connectivity and efficiently allocates tasks to the available edge servers according to transmission delay and connectivity, and finally uses the edge servers to process tasks and return results. Simulation results show that the improved method proposed in the embodiment of the present application can effectively improve the success rate of collecting task processing results in high-dynamic scenarios such as medium and high speeds.
[0069] In the embodiment of the present application, the movement prediction of the cloud computer terminal adopts an improved standard linear Kalman filter KF. Compared with other types of KF, the advantage of the standard linear KF is that the calculation complexity is lower, which is very suitable for edge computing scenarios that have certain requirements for low latency. At the same time, in order to predict the position more accurately, the embodiment of the present application improves the state space model and Kalman gain of the standard linear KF to obtain an improved standard linear KF. Specifically, the process equation of the state space model of the improved standard linear Kalman filter is expressed by the first formula, and the linear observation equation of the state space model of the improved standard linear Kalman filter is expressed by the second formula; wherein,
[0070] The first formula is:
[0071] Where F represents the transition matrix; v k Represented as additive process noise, it follows a normal distribution Q k is represented as the covariance matrix of the process noise; x k represents the state vector at the current moment; x k-1 Represents the state vector at the previous moment; T s Indicates the sampling period;
[0072] The second formula is: k =Hx k +n k ;
[0073] Among them, y k represents the observed value; H = [1 0 0 0] is the observation transformation matrix; x k Represents the state vector at the current moment; n k represents observation noise, which follows a normal distribution R k Represents the covariance matrix of the observation noise.
[0074] In order to obtain the state space model of the standard linear KF, the state vector must be defined and its time evolution must be appropriately modeled. The embodiment of the present application adopts a uniform acceleration linear motion model because the movement of the cloud computer terminal can be regarded as a linear model in a short period of time. At the same time, compared with the commonly used uniform linear motion and uniform acceleration linear motion models, it is closer to the actual motion law of the cloud computer terminal. In the standard linear KF algorithm, considering the speed, acceleration and jerk, the moving coordinates of the cloud computer terminal can be modeled according to the following formula (1):
[0075]
[0076] Where k represents the discrete time instant, T s is the sampling period, L x,k L y,k are the x-coordinate and y-coordinate of the cloud computer terminal, L x,0 , L y,0 are random constants of the x- and y-coordinates of the cloud computer terminal, and v x 、v y are the x and y components of the cloud computer terminal velocity vector, α x , α y are the x and y components of the cloud computer terminal acceleration vector, j x 、j y are the x and y components of the cloud computer terminal's acceleration vector. In this case, the state vector x k It is expressed by the following formula (2):
[0077]
[0078] Thus, the process equation in the state space model can be obtained as shown in formula (3), which specifically represents how the state vector evolves over time:
[0079]
[0080] Where F is the transition matrix, the additive process noise v k Normal distribution Q k is the covariance matrix of process noise, Qk It can represent the error of any model and is usually designed a priori based on the actual problem and the system working conditions. In addition, the linear observation equation of the state space model is expressed as formula (4):
[0081] y k =Hx k +n k (4)
[0082] Where H = [1 0 0 0] is the observation transformation matrix, and the observation noise n k Normal distribution R k is the observation noise covariance matrix, which includes thermal noise and other propagation interference. Formula (3) and Formula (4) respectively define the process equation and the linear observation equation in the state space model of the improved standard linear KF provided in the embodiment of the present application.
[0083] It should be noted that the specific steps of the improved standard linear KF algorithm are usually divided into two stages: prediction stage and update stage. In the prediction stage, the process noise covariance matrix Q k , Last updated status and update error covariance P x,k-1∣k-1 Based on the estimated forecast state And the prediction error covariance P x,k∣k-1 The update phase consists of News and the Kalman gain K k To calculate the updated state in is the observed value y k Based on The predicted observed value of The difference K k represents the reliability of the observation, which depends on P x,k∣k-1 and the observation noise covariance matrix R k The improved standard linear KF algorithm can be used in real time according to Get the predicted location of the cloud computer terminal.
[0084] It should be noted that the predicted position (prediction value) of the cloud computer terminal is based on the Kalman gain K k Recursively adjust to minimize the root mean square error between the observed and predicted values. K k After a period of recursive adjustment, it will tend to be stable, as shown in formula (5):
[0085] lim k→∞ K k →K ss (5)
[0086] This period of time can be divided into two regions: transient and steady state. Kalman gain K k It can be transformed into formula (6), and it can be known that in the transient state, it mainly depends on P x,0∣0 / R k The relative weight, in steady state, depends mainly on Q k / R k The relative weight of:
[0087] K k =(FP x,k-1∣k-1 F T +Q k )H T (HP x,k∣k-1 H T +R k ) -1 (6)
[0088] In order to make the predicted position more accurate and further reduce the root mean square error between the observed value and the predicted value, the embodiment of the present application proposes a new improved method, the main idea of which is to reasonably adjust Q k This controls K in the improved standard linear KF k From the steady-state value K ss,1 Effectively reduced to the steady-state value K ss,2 Specifically, the improved standard linear Kalman filter adjusts the covariance matrix Q of the process noise to avoid large jitters that cause KF failure. k To control the Kalman gain K k slowly decreasing from a first steady-state value to a second steady-state value;
[0089] Among them, Q k =q k I;q k is the process noise factor, I is the unit matrix; k Slowly decrease from the first value to the second value according to the third formula; the third formula is:
[0090]
[0091] k s and k e They represent the start and end time of adjustment respectively, and ω is q k The attenuation factor, q k is the process noise factor; q k In k s The value of the moment is the first value; q k In k e The value of the moment is the second value
[0092] like Figure 3As shown, set R k is a fixed value, Q k =q k I, q k is the process noise factor, and I is the unit matrix. The whole tracking process is divided into three stages.
[0093] The first stage is responsible for quickly and roughly predicting the location of the cloud computer terminal. Set q k Initial value, i.e. q0 = q max , where q max Set to a larger value, K k From the initial transient value K tr,1 After a certain period of adjustment, it tends to the steady-state value K ss,1 , so that the standard linear KF can be recursively adjusted relatively quickly and roughly predict the location of the cloud computer terminal.
[0094] The second stage is the transition stage between the first stage and the third stage. k Transition plan, q k From q max Slowly decrease to q min During this period, K k Also from K ss,1 Start to slowly adjust to K tr,2 It should be noted that q k The choice of transition plan must avoid K k Excessive jitter occurs during the recursive adjustment to ensure that the standard linear KF can normally predict the cloud computer terminal position in the next stage. The present application embodiment provides a q as shown in formula (7) k Transition plan:
[0095] q k =q max ·ω k-ks ,k∈[k s ,k e ] (7)
[0096] where k s and k e are the start and end time of the transition, ω is q k The attenuation factor, q k In k e The moment will become q min . It should be noted that setting q max / q min Too large and transition time k e -k s When it is too short, K k During the recursive adjustment process, there will be too much jitter, which makes the standard linear KF unable to predict the position. In addition, the transition time ke -k s Too long, K k The recursive adjustment time will also be prolonged, which is meaningless in practical applications.
[0097] The third stage is responsible for accurately predicting the location of the cloud computer terminal. k Fixed to q min , K k The second stage K tr,2 As the starting point, recursively adjust again, and finally reach K ss,2 , the improved standard linear KF can further reduce the root mean square error between the observed value and the predicted value, making the predicted position more accurate.
[0098] In summary, the advantage of the standard linear KF in the embodiment of the present application is that the calculation complexity is lower, which is very suitable for edge computing scenarios that have certain requirements for low latency. The embodiment of the present application has improved the standard linear KF, and the improvements are in two points: 1. State space model. The state space model currently used by various KF algorithms is too simple and does not take into account the influence of acceleration and jerk. The embodiment of the present application improves these two factors in the state space model and optimizes the transition and observation conversion matrix in the standard linear KF, which is closer to the actual motion law of the cloud computer terminal; 2. Kalman gain. The predicted position of the cloud computer terminal is recursively adjusted according to the Kalman gain to minimize the root mean square error between the observed value and the predicted value. The Kalman gain will tend to stabilize after a period of recursive adjustment. In order to make the predicted position more accurate, the embodiment of the present application proposes an improved scheme to reasonably adjust Q k , further reducing the root mean square error between the observed and predicted values, and this method involves a q k The transition scheme can avoid the Kalman gain from generating large jitters and causing KF failure.
[0099] As an optional embodiment, step 101 includes:
[0100] It is determined that the edge servers corresponding to the connectivity higher than the second threshold are multiple available edge servers around the target location.
[0101] Wherein, the method further comprises:
[0102] Determine the interruption probability P between the cloud computer terminal and the first edge server as a function of the distance d between the cloud computer terminal and the first edge server Outage (d);
[0103] According to the interruption probability P Outage (d), determining connectivity C(d) between the cloud computer terminal and the first edge server;
[0104] The first edge server is any edge server or any optional edge server.
[0105] In an optional implementation, the interruption probability P between the cloud computer terminal and the first edge server is determined based on a function of the distance d between the cloud computer terminal and the first edge server. Outage (d) including:
[0106] According to the transmission power P of the cloud computer terminal Tx , the path loss L between the cloud computer terminal and the first edge server and the shadow gain G caused by the log-normal shadow, determine the receiving power P of the cloud computer terminal Rx (d);
[0107] According to the receiving power P of the cloud computer terminal Rx (d) and the noise power P N , determine the signal-to-noise ratio SNR(d);
[0108] According to the signal-to-noise ratio SNR(d), the interruption probability P is obtained. Outage (d); The interruption probability P Outage (d) is the probability that the signal-to-noise ratio SNR(d) is lower than the first threshold.
[0109] After obtaining the coordinates of the predicted position, it is possible to select the surrounding connectable edge servers with the predicted position as the center. However, due to the existence of multipath fading and shadow effects, the received signal power in the wireless channel between the cloud computer terminal and the edge server fluctuates violently, so the cloud computer terminal may lose each calculation result. In order to derive connectivity, the embodiment of the present application refers to the radio propagation model. Radio propagation is mainly affected by three effects: path loss, shadowing, and multipath fading. Multipath fading can be easily mitigated using multi-antenna technology, so multipath fading can be ignored. In addition, it is assumed that the connectivity between the cloud computer terminal and each edge server depends only on the signal-to-noise power ratio (SNR). The embodiment of the present application describes the connectivity model based on these assumptions.
[0110] When the cloud computer terminal moves, the connectivity between the cloud computer terminal and the edge server will change. The embodiment of the present application uses the interruption probability to determine the connectivity between the cloud computer terminal and the edge server. The interruption probability can be estimated as a function of the distance d between the cloud computer terminal and the edge server. Therefore, the transmission power is defined as P Tx [dBm], the path loss is L[dB], the shadow gain caused by log-normal shadowing is G[dB], and its standard deviation is σ[dB]. Rx [dBm] is expressed by formula (8):
[0111] PRx (d) = P Tx -L + G(8)
[0112] The signal-to-noise ratio (SNR) SNR(d) can be expressed by Equation (9):
[0113] SNR(d) = P Rx (d) - P N (9)
[0114] where the noise power is given by P N and belongs to additive white Gaussian noise (AWGN). The probability density function of SNR(d) follows a log-normal distribution, as expressed by Equation (10):
[0115]
[0116] The outage probability P Outage , that is, the probability that the received signal is lower than the SNR threshold Z (SNR(d) < Z), is expressed by Equation (11):
[0117]
[0118] where erf(·) is the error function. The connectivity between the cloud computer terminal and the edge server is defined by Equation (12):
[0119] C(d) = 1 - P Outage (d) (12)
[0120] After calculating the connectivity of the surrounding connectable edge servers, since edge servers with too low connectivity will affect the overall efficiency of task allocation, edge servers with connectivity higher than C th are filtered out according to the connectivity threshold C th . These filtered edge servers are called available edge servers.
[0121] In at least one embodiment of the present application, step 102 includes:
[0122] Decompose the first task into N tasks according to the fourth formula and allocate the N tasks to the N available edge servers respectively; where the fourth formula is:
[0123]
[0124] where, T all = max(t1 + t d,1 , t2 + t d,2,...,t N +t d,N ); N is the number of edge services available for the assigned task; C(d i ) is the connectivity between the cloud computer terminal and the i-th edge server; p i is the computing power of the ith edge server; t i The time required for computing the assigned task for the i-th edge server; T all is the maximum time that the cloud computer terminal can collect all the calculation results at one time; t d,i is the transmission delay between the cloud computer terminal and the i-th edge server; M sum This is the first task.
[0125] like Figure 4 As shown, the embodiment of the present application adopts distributed computing to decompose the first task into many small parts, which are then assigned to multiple edge servers for processing, so as to reduce the overall task processing time.
[0126] Assume that the total number of dimensionless tasks (first tasks) on the cloud computer terminal is defined as M sum The computing capacity of each edge server is defined as the amount of computation per second p i (i=1,2,...,N), where N is the number of edge servers. In practice, p i It will be determined based on factors such as CPU and terminal specifications. In addition, the embodiment of the present application also takes into account the transmission delay t between the cloud computer terminal and the edge server. d,i The essence of cloud computing is to transmit virtual machine resources in the cloud to local terminals in real time. Excessive transmission delay between the two will significantly affect user usage. To simplify the calculation, assume that p i It follows a uniform distribution, with a minimum value of 1.0 and a maximum value of 10.0. d,i (i=1,2,...,N) is uniformly distributed, with a minimum of 10ms and a maximum of 500ms. The edge servers are randomly distributed in the area and periodically send their locations and computing power p i Then the cloud computer terminal sends the information of each edge server, the connectivity between the cloud computer terminal and each edge server, and the transmission delay t d,i These three key information are combined to efficiently complete the task M i (i=1,2,...,N) are assigned to the corresponding edge servers. It is worth noting that M i and M sum The relationship between needs to satisfy formula (13):
[0127]
[0128] From formula (12), it can be seen that the cloud computer terminal may lose each calculation result in an unstable wireless communication environment. sum The cloud computer terminal iteratively requests the calculation until all calculations are completed, as shown in the following process.
[0129] 1) Based on the beacons sent by the edge servers, the cloud computer terminal checks the connectivity and transmission delay of each edge server.
[0130] 2) Calculate M i And assigned to the already filtered available edge servers.
[0131] 3) After the calculation time is over, the cloud computer terminal collects the calculation results. The single time from sending to receiving is equal to t i +t d,i [s], where t i =M i / p i .
[0132] 4) Iterate the above process until all M are collected sum .
[0133] Since the time interval between step 1) and step 2) is very short, it is assumed that the task allocation in step 2) can be successfully carried out. In addition, without considering the cooperation between edge servers, M i The calculation results of can only be collected from the i-th ES. It is worth noting that when assigned to the i-th ES, the time required for calculation in the ES can be written as, t i =M i / p i [s]. The overall collection time is equal to the maximum time that the cloud computer terminal can collect all the calculation results at one time. Therefore, the overall collection time can be written as formula (14):
[0134] T all =max(t1+t d,1 ,t2+t d,2 ,...,t N +t d,N )(14)
[0135] Finally, by solving the nonlinear optimization problem shown in formula (15), the cloud computer terminal can effectively improve the efficiency of task allocation:
[0136]
[0137] In order to evaluate the task processing method provided by the embodiment of the present application, the embodiment of the present application performed a computer performance simulation. The success rate of collecting task processing results of the task processing method provided by the present application was compared with the success rate of collecting task processing results of the traditional MEC task management method. In the simulation, the number of edge servers ES n, the moving speed v and acceleration α of the cloud computer terminal were changed, and the acceleration j remained unchanged. The simulation parameters are shown in Table 1. The initial value of the state vector is as follows:
[0138]
[0139] ES is evenly distributed in a fixed area. When the received SNR exceeds SNRZ, the cloud computer terminal connects to ES, and the cloud computer terminal can collect the task processing results. When predicting the location of the cloud computer terminal, the task processing time is set to T s , which is the sampling interval of KF. Let the task volume be M sum , the processing capacity of ES is p i (i=1,2,...,n), p i It follows a uniform distribution, with a minimum value of 1.0 and a maximum value of 10.0. The transmission delay t between the cloud computer terminal and the ES d,i (i=1,2,...,n) obeys uniform distribution, with a minimum value of 10ms and a maximum value of 500ms. This application does not consider the calculation time required for the prediction itself.
[0140] Table 1 Simulation parameters
[0141]
[0142]
[0143] Figure 5 The success rate of collecting task processing results when the cloud computer terminal moves at a speed of 5km / h is shown. Since the current speed is less than the speed threshold, it is judged as a low-dynamic scene, and the ES closest to the cloud computer terminal is selected to process the task and return the task processing results to the cloud computer terminal. The results show that there is almost no difference in the collection success rate between the improved method proposed in this paper and the traditional method. This is because when the cloud computer terminal moves slowly, the cloud computer terminal moves a shorter distance during the task processing of the selected ES, and the ES can still maintain connectivity from the initiation of the task request to the end of the task processing.
[0144] Figure 6 The cloud computer terminal was demonstrated to be able to run at a medium speed of 50km / h and 2km / h 2 Acceleration and 0.1km / h 3The success rate of collecting task processing results when moving with acceleration. Since the current speed and acceleration are both greater than the threshold, it is determined to be a high dynamic scene. The results show that when the cloud computer terminal moves at medium speed, the improved method proposed in the embodiment of the present application has an overall slight improvement in the collection success rate compared to the traditional method. In addition, within a fixed area, an increase in the number of ES n will increase the success rate of collecting task processing results. This is because as the density of ES increases, the distance between the cloud computer terminal and the ES is shortened, and the connectivity will also increase, thereby increasing the collection success rate.
[0145] Figure 7 The cloud computer terminal was demonstrated to be able to reach a high speed of 100km / h and 4km / h 2 Acceleration and 0.1km / h 3 The success rate of collecting task processing results when moving with accelerated acceleration. Since the current speed and acceleration are both greater than the threshold, it is also determined to be a high dynamic scene. The results show that when the cloud computer terminal moves at high speed, the longer the processing time, the lower the success rate of collecting task processing results. This is because the longer the processing time, the longer the moving distance of the cloud computer terminal, while the ES that originally processed the task is still connected at the requested location, the connection between the two will often be interrupted, and the collection success rate will also decrease. If the collection success rate is too low, the cloud computer terminal will not be able to work normally. Compared with traditional methods, the improved method proposed in the embodiment of the present application has an overall significant improvement in the collection success rate, which effectively solves the problem of the low collection success rate of traditional methods when moving at high speeds.
[0146] In summary, the embodiment of the present application screens out multiple available edge servers around the target location based on connectivity, and then uses a distributed computing method to efficiently allocate tasks to multiple available edge servers based on transmission delay and connectivity; on the one hand, in response to the low latency requirement of the cloud computer terminal, the embodiment of the present application allocates a task to multiple available edge servers around for processing, and optimizes the transmission delay between each edge server and the cloud computer terminal in the allocation algorithm, effectively reducing the overall delay; on the other hand, in response to the path loss and shadow effect existing in the wireless communication environment, taking into account the connectivity between each edge server and the cloud computer terminal, the allocation algorithm can reduce the impact of path loss and shadow effect on task allocation; therefore, while efficiently allocating tasks to multiple available edge servers around, the embodiment of the present application can also effectively reduce the overall delay and reduce the impact of path loss and shadow effect, thereby effectively improving the success rate of collecting task processing results.
[0147] like Figure 8 As shown, the embodiment of the present application also provides a task processing device, which is applied to a cloud computer terminal, and the device includes:
[0148] Determination module 801 is used to determine a plurality of available edge servers around a target location according to connectivity between the cloud computer terminal and the edge server when there is a first task that needs to be processed by the cloud computer terminal; wherein the target location is the location of the cloud computer terminal;
[0149] an allocation module 802, configured to allocate the first task to at least two of the available edge servers according to a transmission delay between the cloud computer terminal and the available edge servers and connectivity between the cloud computer terminal and the available edge servers;
[0150] The receiving module 803 is configured to receive the task processing results returned after the at least two edge servers can be used to process the tasks.
[0151] As an optional embodiment, the device further includes:
[0152] A detection module, used to detect the motion state information of the cloud computer terminal, wherein the motion state information includes: at least one of motion speed, acceleration, and jerk;
[0153] A prediction module, for predicting the position of the cloud computer terminal using an improved standard linear Kalman filter when it is determined that the cloud computer terminal is in a high dynamic scene according to the motion state information, to obtain a predicted position, and using the predicted position as the target position;
[0154] Alternatively, the position determination module is used to use the current position of the cloud computer terminal as the target position when it is determined that the cloud computer terminal is in a static scene or a low-dynamic scene based on the motion state information.
[0155] As an optional embodiment, the process equation of the state space model of the improved standard linear Kalman filter is represented by a first formula, and the linear observation equation of the state space model of the improved standard linear Kalman filter is represented by a second formula; wherein,
[0156] The first formula is:
[0157] Where F represents the transition matrix; v k Represented as additive process noise, it follows a normal distribution Q k is represented as the covariance matrix of the process noise; x k represents the state vector at the current moment; x k-1 Represents the state vector at the previous moment; T s Indicates the sampling period;
[0158] The second formula is: k =Hxk +n k ;
[0159] Among them, y k represents the observed value; H = [1 0 0 0] is the observation transformation matrix; x k Represents the state vector at the current moment; n k represents observation noise, which follows a normal distribution R k Represents the covariance matrix of the observation noise.
[0160] As an optional embodiment, the improved standard linear Kalman filter is implemented by adjusting the covariance matrix Q of the process noise k To control the Kalman gain K k slowly decreasing from a first steady-state value to a second steady-state value;
[0161] Among them, Q k =q k I;q k is the process noise factor, I is the unit matrix; k Slowly decrease from the first value to the second value according to the third formula; the third formula is:
[0162]
[0163] k s and k e They represent the start and end time of adjustment respectively, and ω is q k The attenuation factor, q k is the process noise factor; q k In k s The value of the moment is the first value; q k In k e The value of the moment is the second value.
[0164] As an optional embodiment, the determining module includes:
[0165] The determination submodule is used to determine that the edge servers corresponding to the connectivity higher than the second threshold are multiple available edge servers around the target location.
[0166] As an optional embodiment, the allocation module includes:
[0167] The allocation submodule is used to decompose the first task into N tasks according to a fourth formula, and allocate the N tasks to the N available edge servers respectively; wherein the fourth formula is:
[0168]
[0169] in, T all =max(t1+t d,1 ,t2+t d,2 ,...,t N +t d,N ); N is the number of edge services available for the assigned task; C(d i ) is the connectivity between the cloud computer terminal and the i-th edge server; p i is the computing power of the ith edge server; t i The time required for computing the assigned task for the i-th edge server; T all is the maximum time that the cloud computer terminal can collect all the calculation results at one time; t d,i is the transmission delay between the cloud computer terminal and the i-th edge server; M sum This is the first task.
[0170] As an optional embodiment, the device further includes:
[0171] A probability determination module is used to determine the interruption probability P between the cloud computer terminal and the first edge server according to a function of the distance d between the cloud computer terminal and the first edge server. Outage (d);
[0172] A connectivity determination module is used to determine the connectivity of the network according to the interruption probability P Outage (d), determining connectivity C(d) between the cloud computer terminal and the first edge server;
[0173] The first edge server is any edge server or any optional edge server.
[0174] As an optional embodiment, the probability determination module includes:
[0175] The first submodule is used to determine the transmission power P of the cloud computer terminal according to the transmission power P of the cloud computer terminal. Tx , the path loss L between the cloud computer terminal and the first edge server and the shadow gain G caused by the log-normal shadow, determine the receiving power P of the cloud computer terminal Rx (d);
[0176] The second submodule is used to receive the power P of the cloud computer terminal according to the received power P Rx (d) and the noise power P N , determine the signal-to-noise ratio SNR(d);
[0177] The third submodule is used to obtain the interruption probability P according to the signal-to-noise ratio SNR(d). Outage (d); The interruption probability P Outage(d) is the probability that the signal-to-noise ratio SNR(d) is lower than the first threshold.
[0178] The embodiment of the present application selects multiple available edge servers around the target location according to connectivity, and then uses a distributed computing method to efficiently allocate tasks to multiple available edge servers according to transmission delay and connectivity; on the one hand, in response to the low latency requirement of cloud computer terminals, the embodiment of the present application allocates a task to multiple available edge servers around for processing, and optimizes the transmission delay between each edge server and the cloud computer terminal in the allocation algorithm, effectively reducing the overall delay; on the other hand, in response to the path loss and shadow effect existing in the wireless communication environment, taking into account the connectivity between each edge server and the cloud computer terminal, the allocation algorithm can reduce the impact of path loss and shadow effect on task allocation; therefore, the embodiment of the present application can effectively reduce the overall delay and reduce the impact of path loss and shadow effect while efficiently allocating tasks to multiple available edge servers around, thereby effectively improving the success rate of collecting task processing results.
[0179] It should be noted that the task processing device provided in the embodiment of the present application is a device capable of executing the above-mentioned task processing method. All embodiments of the above-mentioned task processing method are applicable to the device and can achieve the same or similar beneficial effects, which will not be repeated here.
[0180] like Fig. 9 As shown, an embodiment of the present application also provides a cloud computer terminal, including a memory 910, a processor 900, and a computer program stored in the memory 910 and executable on the processor 900. When the processor 900 executes the program, each process in the task processing method embodiment as described above is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.
[0181] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, each process in the task processing method embodiment described above is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0182] The embodiment of the present application also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the various processes of the task processing method embodiment described above are implemented, and the same technical effect can be achieved. To avoid repetition, they are not described here.
[0183] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.
[0184] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A device that specifies functions in one or more processes and / or one or more blocks.
[0185] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable storage medium produce a paper product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0186] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that the computer or other programmable device executes a series of operating steps to produce a computer-implemented process, thereby providing instructions executed on the computer or other programmable device for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0187] The above is a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.
Claims
1. A task processing method, characterized in that: Applied to a cloud computer terminal, the method includes: In the case that the cloud computer terminal has a first task that needs to be processed, determining a plurality of available edge servers around a target location according to connectivity between the cloud computer terminal and the edge server; wherein the target location is the location of the cloud computer terminal; Allocating the first task to at least two of the available edge servers according to a transmission delay between a cloud computer terminal and an available edge server and connectivity between the cloud computer terminal and the available edge server; Receive task processing results returned by the at least two edge servers after processing the tasks.
2. The method according to claim 1, characterized in that Before determining a plurality of available edge servers around the target location based on connectivity between the cloud computer terminal and the edge server, the method further includes: Detecting motion state information of the cloud computer terminal, wherein the motion state information includes at least one of motion speed, acceleration, and jerk; When it is determined according to the motion state information that the cloud computer terminal is in a high dynamic scene, using an improved standard linear Kalman filter to predict the position of the cloud computer terminal to obtain a predicted position, and using the predicted position as the target position; Alternatively, when it is determined according to the motion state information that the cloud computer terminal is in a static scene or a low-dynamic scene, the current position of the cloud computer terminal is used as the target position.
3. The method according to claim 2, characterized in that The process equation of the state space model of the improved standard linear Kalman filter is represented by a first formula, and the linear observation equation of the state space model of the improved standard linear Kalman filter is represented by a second formula; wherein, The first formula is: Where F represents the transition matrix; v k Represented as additive process noise, which follows a normal distribution Q k is represented as the covariance matrix of the process noise; x k represents the state vector at the current moment; x k-1 Represents the state vector at the previous moment; T s Indicates the sampling period; The second formula is: k =Hx k +n k ; Among them, y k represents the observed value; H = [1 0 0 0] is the observation transformation matrix; x k Represents the state vector at the current moment; n k represents observation noise, which follows a normal distribution R k Represents the covariance matrix of the observation noise.
4. The method according to claim 2, characterized in that: The modified standard linear Kalman filter is constructed by adjusting the covariance matrix Q of the process noise k To control the Kalman gain K k slowly decreasing from a first steady-state value to a second steady-state value; Among them, Q k =q k I;q k is the process noise factor, I is the unit matrix; k Slowly decrease from the first value to the second value according to the third formula; the third formula is: k s and k e They represent the start and end time of adjustment respectively, and ω is q k The attenuation factor, q k is the process noise factor; q k In k s The value of the moment is the first value; q k In k e The value of the moment is the second value.
5. The method according to claim 1, characterized in that Determining multiple available edge servers around the target location based on the connectivity between the cloud computer terminal and the edge server includes: It is determined that the edge servers corresponding to the connectivity higher than the second threshold are multiple available edge servers around the target location.
6. The method according to claim 1, characterized in that The allocating the first task to at least two of the available edge servers according to the transmission delay between the cloud computer terminal and the available edge server and the connectivity between the cloud computer terminal and the available edge server comprises: According to the fourth formula, the first task is decomposed into N tasks, and the N tasks are respectively assigned to the N available edge servers; wherein the fourth formula is: in, T all =max(t1+t d,1 ,t2+t d,2 ,...,t N +t d,N ); N is the number of edge services available for the assigned task; C(d i ) is the connectivity between the cloud computer terminal and the i-th edge server; p i is the computing power of the ith edge server; t i The time required for computing the assigned task for the i-th edge server; T all is the maximum time that the cloud computer terminal can collect all the calculation results at one time; t d,i is the transmission delay between the cloud computer terminal and the i-th edge server; M sum This is the first task.
7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Determine the interruption probability P between the cloud computer terminal and the first edge server as a function of the distance d between the cloud computer terminal and the first edge server Outage (d); According to the interruption probability P Outage (d), determining connectivity C(d) between the cloud computer terminal and the first edge server; The first edge server is any edge server or any optional edge server.
8. The method according to claim 7, characterized in that Determine the interruption probability P between the cloud computer terminal and the first edge server as a function of the distance d between the cloud computer terminal and the first edge server Outage (d) including: According to the transmission power P of the cloud computer terminal Tx , the path loss L between the cloud computer terminal and the first edge server and the shadow gain G caused by the log-normal shadow, determine the receiving power P of the cloud computer terminal Rx (d); According to the receiving power P of the cloud computer terminal Rx (d) and the noise power P N , determine the signal-to-noise ratio SNR(d); According to the signal-to-noise ratio SNR(d), the interruption probability P is obtained. Outage (d); The interruption probability P Outage (d) is the probability that the signal-to-noise ratio SNR(d) is lower than the first threshold.
9. A task processing device, characterized in that: Applied to a cloud computer terminal, the device comprises: a determination module, configured to determine, when the cloud computer terminal has a first task that needs to be processed, a plurality of available edge servers around a target location according to connectivity between the cloud computer terminal and the edge server; wherein the target location is the location of the cloud computer terminal; an allocation module, configured to allocate the first task to at least two of the available edge servers according to a transmission delay between the cloud computer terminal and the available edge servers and connectivity between the cloud computer terminal and the available edge servers; The receiving module is used to receive the task processing results returned after the at least two edge servers can be used to process the tasks.
10. A cloud computer terminal, comprising a memory, a processor, and a program stored in the memory and executable on the processor; characterized in that: When the processor executes the program, the task processing method according to any one of claims 1 to 9 is implemented.
11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the task processing method according to any one of claims 1 to 9 are implemented.
12. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the task processing method according to any one of claims 1 to 9.
Citation Information
Patent Citations
MEC-based V2X mobility prediction and content cache unloading scheme
CN109391681A
Dynamic task allocation method and system for mobile edge computing server
CN112559187A
Resource and task allocation optimization system of unmanned aerial vehicle edge server
CN113395679A
Edge calculation unloading optimization method based on vehicle position prediction
CN113727308A
Mobile edge calculation method and device, equipment and medium
CN116827952A
