Task List Scheduling Method Based on Vehicle Cooperative Clusters

Abstract

The present invention provides a task list scheduling method based on a vehicle cooperation cluster. First, a task model of the vehicle and a vehicle networking communication model are designed, considering the link occupancy situation of V2V communication. Then, a task list scheduling algorithm based on communication occupancy and predicted task priority is designed. The algorithm predicts the computing priority of each task, selects the tasks and vehicle computing devices to execute according to the priority value, and then selects the optimal path to transmit the task data. Finally, through experimental comparison of the algorithm in this paper with other algorithm schemes, the simulation results show that the algorithm in this paper reduces the average task completion time by about 27% and improves the efficiency of vehicle computing devices by about 18%.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication networks, and particularly to a task list scheduling method based on a vehicle cooperation cluster. Background Art

[0002] With the rapid development of intelligent transportation systems, the Internet of Vehicles (IoV) is gradually maturing. Idle vehicles on the road can form a vehicle cooperation cluster, and many tasks use the computing resources of the vehicle cooperation cluster for task scheduling. Since the communication of vehicle-to-vehicle network (V2V) is direct communication, the scheduling of dependent tasks needs to consider the link occupancy of IoV communication.

[0003] However, many current studies in the IoV do not consider the communication occupancy situation. Therefore, this paper proposes a task list scheduling method based on a vehicle cooperation cluster. First, a task model of the vehicle and an IoV communication model are designed, considering the link occupancy of V2V communication. Then, the computing priority of each task is designed, and the tasks and vehicle computing devices to be executed are selected according to the priority value, and then the optimal path is selected to transmit the task data. Summary of the Invention

[0004] To solve the above problems, the present invention discloses a task list scheduling method based on a vehicle cooperation cluster, a scheduling algorithm for completing dependent tasks in the scenario of a vehicle cooperation cluster, which considers the situation of V2V communication link occupancy, improves the performance in heterogeneous computing systems, and is superior to the proposed algorithms in terms of completion time, acceleration, and efficiency.

[0005] To achieve the above object, the technical solution adopted by the present invention is: a task scheduling method based on a vehicle cooperation cluster, including the following specific steps:

[0006] Step 1, a user requests a task flow to be executed from a vehicle cooperation cluster;

[0007] Step 2, design a task flow model of the user, which is represented by a directed acyclic graph, the nodes of the graph are each task, and the edges of the graph are the dependency relationships between the executions of two tasks;

[0008] Step 3, design a communication environment model of the vehicle cooperation cluster, and give an execution cost model of the task in the computing device of the vehicle;

[0009] Step 4, in the case of link occupancy in the communication environment model, define the scheduling problem of the task flow, and give an objective function for minimizing time;

[0010] Step 5, design a task scheduling algorithm based on communication link occupancy and predicted priority, calculate the task priority, determine the execution order of the tasks, and schedule the tasks to the computing devices of the vehicle cooperation cluster for execution.

[0011] Furthermore, in the step 1, the task flow that the user requests the vehicle cooperation cluster to schedule is composed of multiple tasks with dependencies, and the tasks need to be executed on each vehicle computing device in the vehicle cooperation cluster according to the dependencies.

[0012] Furthermore, in the step 2, the task flow model of the user is composed of a set of dependent tasks, which is represented as a directed acyclic graph (DAG), denoted as G=(T, E, D), where T represents the set of |T| tasks with dependencies requested by the user; E is the set of edges of the predecessor tasks and successor tasks, representing the dependencies between tasks; D is the set of the sizes of the transmitted data between the predecessor tasks and successor tasks. If task i is a predecessor task of task j, it represents the size of the data that needs to be transmitted to task j after task i is executed.

[0013] Furthermore, the communication environment model of the vehicle cooperation cluster designed in the step 3 is as follows: Define an undirected graph to represent the heterogeneous vehicle communication environment, where is the set of vehicle computing devices, and k is the number of vehicles in the vehicle cooperation cluster; is the set of communication links, and q is the number of communication links between vehicles, is the set of transmission speeds of the communication links; where represents the transmission speed on the communication link After the predecessor task is completed on the source vehicle computing device, a communication link is selected to transmit the data required by the successor task to the target vehicle computing device for execution.

[0014] Furthermore, in the step 3, it is assumed that the data needs to be transmitted from task to task . is the source vehicle computing device for processing task , is the target vehicle computing device for processing task , represents the set of communication routes between the source vehicle computing device and the target vehicle computing device . Each route passes through multiple vehicle computing devices and consists of multiple communication links with different transmission rates and is searched by the depth-first search algorithm or the breadth-first search algorithm; Define as the average communication speed between the source vehicle computing device and the target vehicle computing device .

[0015] , where represents the communication speed of the path ; due to the message forwarding of the vehicle using the direct communication scheme, from the source vehicle computing device and the target vehicle computing device the communication speed is determined by the minimum speed of the link, so the communication speed of the path is:

[0016] ; where represents the transmission speed on the communication link , and the average communication speed between the source vehicle computing device and other vehicle computing devices is defined as:

[0017] , where represents the number of available vehicle computing devices in the vehicle cooperation cluster.

[0018] Furthermore, the execution cost model of the task in the vehicle computing device in step 3 includes four parts: The first part is the data size of the task, and the data of task i is defined as ; The second part is the computing device equipped in each vehicle. Assuming that each vehicle is only equipped with one computing device, the set of vehicle computing devices is defined as , and the computing power of the device is represented by the set , represents the computing power of vehicle computing device k; The third part is to define as the computing time of task on the vehicle computing device , which is obtained by the following formula: ; The average execution time of task is defined as follows: , The fourth part is the estimated communication time of task at the vehicle computing device , that is, the ratio of the transmission data size from the predecessor task to the successor task to the average transmission rate of the vehicle computing device , which is calculated by the following formula: , where represents the average transmission rate of the vehicle computing device , represents the predecessor task to the successor task transmission data volume.

[0019] Furthermore, in step 4, since the earliest start time of a task depends on the task start transmission time, and the task transmission time is restricted by the actual link finish time LFT, it is necessary to obtain the link start time LST in advance. Through the above analysis, it is necessary to obtain the value of the actual link finish time LFT. Define the link start time as LST, and the definition of LST is shown in the following formula:

[0020] ; where represents the amount of transmitted data from the source vehicle computing device to the target vehicle computing device the start time of link x on communication path z, represents the available time of link x on path z, represents the task at the source vehicle computing device the actual end time; the start time of the first link on path z should take the maximum value between the link available time and the completion time of the previous task, and the start time of the subsequent link is the start time of the previous path; similarly, the value of LFT is obtained as follows:

[0021] , where represents the amount of transmitted data from the source vehicle computing device to the target vehicle computing device the end time of link x on communication path z; each task has different communication times on different vehicle computing devices, and the start time of the task is determined by the available time of the vehicle computing device and the available time of the transmission link; the start time of the starting task

[0022] is 0, and the start time of other tasks is obtained from the following formula, ; where

[0023] ; where is the set of previous tasks of task , is the vehicle computing device that executes the previous task , is the target vehicle computing device the earliest available time, represents the amount of transmitted data from the source vehicle computing device to the target vehicle computing device The end time of the last link on communication path z. The above equation shows that EST and LFT are closely related. When a vehicle device wants to send task data, it may not transmit immediately. Since the communication link to be transmitted may be occupied by other messages until it becomes idle. Thus, the end time of task i can be obtained ; where represents task at the target vehicle computing device 's estimated start time, represents task at the target vehicle computing device 's execution time; finally, makespan is defined as the completion time of the task flow. Makespan is determined by the completion time of the ending task. The objective function of the problem is to minimize the completion time, which is defined by the following equation:

[0024] , where represents the completion time of the ending task at the target vehicle computing device .

[0025] Furthermore, in step 5, the task scheduling algorithm based on communication link occupancy and prediction priority is the PPTS_CC algorithm. It uses the concept of predictability to determine the task priority stage and the vehicle computing device selection stage. First, based on the prediction cost matrix calculate the priority of each task, and then, select a vehicle with an unoccupied communication link to execute the task, and select the vehicle computing device with the minimum completion time for the task among all available vehicles; define PCM as a matrix, where each element represents the maximum value of the priority of task on each vehicle computing device ; Each task is recursively calculated by traversing from the ending task to the starting task through the given vehicle task flow model; is recursively determined by the following equation: , where represents 's successor task, represents the average transmission time of task and task on the vehicle computing device .

[0026] Furthermore, the first stage of task scheduling is to prioritize tasks. First, calculate the priority value of each task , represented by , which represents the average PCM value of task , obtained by the following equation: 。

[0027] Further, in the vehicle computing device selection phase of step 5, first, calculate the forward prediction value of each task on the arriving vehicle computing device ; then, select the vehicle computing device with the smallest value to execute the task, and mark a path between the source vehicle computing device and the target vehicle computing device as occupied until the task data transmission is completed; the goal of this phase is to ensure that the successor tasks of the current task can be completed in advance without increasing the algorithm complexity; the calculation method of the forward prediction value is as follows: 。

[0028] Advantages of the present invention:

[0029] Compared with the prior art, in an environment where the V2V communication link in the vehicle cooperation cluster is occupied, the present invention adopts a scheduling algorithm (PPTS_CC) that considers communication link occupancy and predicted task priorities to solve the problem of scheduling dependent task flows, improves the utilization rate of computing resources in the vehicle cooperation cluster, reduces the task completion time, and increases the acceleration of task completion. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a task scheduling model diagram of the vehicle cooperation cluster of the present invention.

[0031] Figure 2 is a task flow model diagram of the user of the present invention.

[0032] Figure 3 is a flowchart of the PPTS_CC algorithm of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0033] The present invention will be further clarified below in conjunction with the drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. It should be noted that the terms "front", "rear", "left", "right", "up" and "down" used in the following description refer to the directions in the drawings, and the terms "inside" and "outside" refer to the directions towards or away from the geometric center of a specific component respectively.

[0034] As Figures 1-3 shown, this embodiment provides a task list scheduling method based on a vehicle cooperation cluster, including the following steps:

[0035] In step 1, the task flow requested by the user consists of multiple tasks with dependencies, and the tasks need to be executed on each vehicle computing device in the vehicle cooperation cluster according to the dependencies.

[0036] In step 2, the user's task flow model consists of a set of dependent tasks, represented as a directed acyclic graph (DAG) G=(T, E, D), where T represents the set of |T| dependent tasks requested by the user; E is the set of edges between the predecessor tasks and the successor tasks, representing the dependency relationships between tasks; D is the set of the sizes of the transmitted data between the predecessor tasks and the successor tasks. Task i is the predecessor task of task j. represents the size of the data that needs to be transmitted to task j after task i is executed.

[0037] In step 3, the communication environment model of the vehicle cooperation cluster is given as follows: Define an undirected graph represents the heterogeneous vehicle communication environment. is the set of vehicle computing devices, where k is the number of vehicles in the vehicle cooperation cluster; is the set of communication links, where q is the number of communication links between vehicles. is the set of transmission speeds of the communication links; where represents the transmission speed on the communication link ; Define as the average communication speed between the source vehicle computing device and the target vehicle computing device .

[0038] ; According to Figure 2 the case of the vehicle cooperation cluster communication model diagram of, the path transmission rate values between the vehicle computing devices of UG can be obtained, {V 1 -V 2 : 1, V 2 -V 3 :1, V 2 -V 4 : 2}, V 3 -V 4 : 4, V 4 - V 5 : 4}. For example There are two paths in total, where ; where represents the communication speed of the path ; Since the vehicle uses message forwarding in the direct communication scheme, the communication speed from the source vehicle computing device and the target vehicle computing device is determined by the minimum speed of the link. Therefore, the communication speed of the path is: ; where Denote the communication speed of link q, and define the source vehicle computing device The average communication speed with other vehicle computing devices is:

[0039] , where Denote the number of available vehicle computing devices in the vehicle cooperation cluster.

[0040] For example, the above path , , because the transmission rate of link is the minimum.

[0041] According to Figure 2 The vehicle cooperation cluster communication model diagram case, the average transmission rate values of each vehicle computing device in the cooperation cluster can be obtained through the above steps as shown in Table 1:

[0042] Table 1: Average transmission rate values of each vehicle computing device

[0043]

[0044] The execution cost model of a task in a vehicle's computing device includes four parts: The first part is the data size of the task, and the data of task i is defined as ; The second part is the computing device equipped in each vehicle. Assume that each vehicle is equipped with only one computing device. Define the set of vehicle computing devices , and the computing power of the device is represented by the set , Denote the computing power of vehicle computing device k; The third part is to define as the computing time of task on vehicle computing device , which is obtained from the following formula: ; Each task is associated with the estimated computing time on vehicle computing device , denoted by , and the average execution time of task is defined as follows: , The fourth part is the estimated communication time of task at vehicle computing device , which is calculated from the following formula:

[0045] That is, the ratio of the transmission data size from the predecessor task to the successor task to the average transmission rate of computing device , which is calculated from the following formula: , where Represents a vehicle computing device The average transmission rate of Represents the predecessor task To the successor task The amount of transmitted data.

[0046] In step 4, since the earliest start time of a task depends on the task start transmission time; and the task transmission time is restricted by the actual link finish time LFT; therefore, it is necessary to obtain the link start time LST in advance; through the above analysis, it is necessary to obtain the value of the actual link finish time LFT, define the link start time as LST, and the definition of LST is shown in the following formula:

[0047] In step 4, since the earliest start time of a task depends on the task start transmission time; and the task transmission time is restricted by the actual link finish time LFT; therefore, it is necessary to obtain the link start time LST in advance; through the above analysis, it is necessary to obtain the value of the actual link finish time LFT, define the link start time as LST, and the definition of LST is shown in the following formula:

[0048] ; where Represents the amount of transmitted data From the source vehicle computing device To the target vehicle computing device The start time of link x on communication path z, Represents the available time of link x on path z, Represents the task At the source vehicle computing device The actual end time; the start time of the first link on path z should take the maximum value between the link available time and the completion time of the predecessor task, and the start time of the subsequent link is the start time of the previous path; similarly, the value of LFT is obtained as follows:

[0049] , where Represents the amount of transmitted data From the source vehicle computing device To the target vehicle computing device The end time of link x on communication path z; each task has different communication times on different vehicle computing devices, and the start time of the task is determined by the available time of the vehicle computing device and the available time of the transmission link; the start time of the starting task

[0050] Is 0, and the start time of other tasks Is obtained from the following formula ,

[0051] ; where is the task set of predecessor tasks of is the vehicle computing device that executes the predecessor task ; is the target vehicle computing device earliest available time of indicates the amount of transmitted data from the source vehicle computing device to the target vehicle computing device end time of the last link on communication path z. The above formula shows that EST and LFT are closely related; thus, the end time of task i is obtained as follows: ; where represents the task estimated start time of at the target vehicle computing device represents the task execution time of at the target vehicle computing device; finally, makespan is defined as the completion time of the task flow, and makespan is determined by the completion time of the end task. The objective function of the problem is to minimize the completion time, which is defined by the following formula:

[0052] , where represents the completion time of the end task at the target vehicle computing device .

[0053] In step 5, the communication link occupancy and prediction priority task scheduling algorithm is the PPTS_CC algorithm. It uses the concept of predictability to determine the task priority stage and vehicle computing device selection stage. First, based on the prediction cost matrix calculate the priority of each task. Then, select the vehicle with the unoccupied communication link to execute the task, and select the vehicle computing device with the minimum completion time for the task among all available vehicles; define PCM as a matrix, where each element represents the maximum value of the priority of task on each vehicle computing device ; Each task is recursively calculated by traversing from the end task to the start task through the given vehicle task flow model; is recursively determined by the following equation: , where represents successor task of represents the average transmission time between task and task on the vehicle computing device .

[0054] The first stage of task scheduling is to sort tasks by priority. First, calculate the priority value of each task . Use to represent the average PCM value of task , which is obtained by the following formula: .

[0055] In the vehicle computing device selection stage of step 5, first, calculate the forward prediction value of each task on the arriving vehicle computing device ; then, select the vehicle computing device with the smallest value to execute the task, and mark a path between the source vehicle computing device and the target vehicle computing device as occupied until the task data transmission is completed; the goal of this stage is to ensure that the successor tasks of the current task can be completed in advance without increasing the algorithm complexity; the calculation method of the forward prediction value is as follows: .

[0056] According to Figure 2 the task flow model diagram case of the user, the predicted cost PCM values of the DAG can be obtained from the above steps as shown in Table 2:

[0057] Table 2: Computational cost values of each task on different vehicle computing devices

[0058]

[0059] The priority of each task is shown in Table 3:

[0060] Table 3: Final priority values of each task

[0061]

[0062] The technical means disclosed in the solution of the present invention are not limited to the technical means disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features.

Claims

1. A task list scheduling method based on vehicle cooperative clusters, characterized in that: The specific steps include: Step 1: The user requests the vehicle cooperation cluster to schedule a task flow to be executed; in step 1, the task flow that the user requests the vehicle cooperation cluster to schedule is composed of multiple tasks with dependencies, and the tasks need to be executed on each vehicle computing device in the vehicle cooperation cluster according to the dependencies; Step 2, design the user's task flow model, represented by a directed acyclic graph, where the nodes of the graph are tasks, and the edges of the graph are the dependency relationships between the execution of two tasks; the user's task flow model in step 2 is composed of a set of tasks with dependencies, defined as a directed acyclic graph G=(T,E,D), where T represents a set of |T| tasks with dependencies requested by the user; E is a set of edges between the predecessor task and the successor task, representing the dependency relationship between the tasks; D is a set of transmission data sizes between the predecessor task and the successor task, , task i is the predecessor task of task j, Indicates the size of data that needs to be transferred to task j after task i is executed; Step 3: Design a communication environment model for the vehicle cooperation cluster and provide a cost model for executing tasks in the vehicle's computing device. The communication environment model for the vehicle cooperation cluster designed in step 3 is as follows: Define an undirected graph represents the heterogeneous vehicle communication environment, is the set of vehicle computing devices, where k is the number of vehicles in the vehicle cooperation cluster; is the set of communication links, where q is the number of communication links between vehicles, is the set of transmission speeds of the communication link; Indicates the communication link After the predecessor task is completed by the source vehicle computing device, a communication link is selected to transmit the data required for the successor task to the target vehicle computing device for execution; In step 3, it is assumed that the data needs to be From the task Transfer tasks , Is processing task The source vehicle computing device, Is processing task The target vehicle computing device, Representative source vehicle computing device and target vehicle computing device between A collection of communication routes, each route passing through multiple vehicle computing devices and composed of multiple communication links with different transmission rates Composition, searched by depth-first search algorithm or breadth-first search algorithm; definition Source vehicle computing device and target vehicle computing device Average communication speed; ,in Representative path communication speed; Since the vehicle adopts the direct communication scheme for message forwarding, the source vehicle computing device and target vehicle computing device The communication speed is determined by the minimum speed of the link, so the path The communication speed is: ;in Indicates the communication link The transmission speed on the source vehicle computing device is defined The average communication speed with other vehicle computing devices is: ,in represents the number of available vehicle computing devices in the vehicle cooperation cluster; The execution cost model of the task in the computing device of the vehicle in step 3 includes four parts: the first part is the data size of the task, and the data of task i is defined as The second part is the computing device equipped in each vehicle. Assuming that each vehicle is equipped with only one computing device, define the vehicle computing device set , the computing power of the device is composed of the collection express, represents the computing power of the vehicle computing device k; the third part is the definition It's a task In-vehicle computing equipment The calculation time on is given by: ;Task The average execution time is defined as follows: , the fourth part is the task In-vehicle computing equipment The estimated communication time at the location, that is, the size of the data transmitted from the predecessor task to the successor task and the vehicle computing device The ratio of the average transmission rate is calculated by the following formula: ,in Represents vehicle computing device The average transmission rate, Represents the predecessor task To the successor task The amount of data transmitted; Step 4: In the case of link occupancy in the communication environment model, define the scheduling problem of the task flow and give the objective function of minimizing the time. In step 4, since the earliest start time of the task depends on the task start transmission time, and the task transmission time is limited by the actual link completion time LFT, it is necessary to obtain the link start time LST in advance. After obtaining the value of the actual link completion time LFT, define the link start time as LST. The definition of LST is as follows: ;in Indicates the amount of data transferred From the source vehicle computing device To the target vehicle computing device The start time of link x on communication path z, represents the available time of link x on path z, Indicates the task In-source vehicle computing equipment The actual end time at the location; the start time of the first link on path z should take the maximum value between the link available time and the completion time of the predecessor task, and the start time of the subsequent links is the start time of the previous path; similarly, the value of LFT is as follows: ,in Indicates the amount of data transferred From the source vehicle computing device To the target vehicle computing device The end time of link x on communication path z; each task is implemented on a different vehicle computing device There are different communication times. The start time of the task is determined by the availability of the vehicle computing equipment and the availability of the transmission link; the start task The start time of other tasks is 0. The start time is given by the following formula: ;in It's a task The set of predecessor tasks, Is to perform the predecessor task Vehicle computing devices, Is the target vehicle computing device The earliest available time, Indicates the amount of data transferred From the source vehicle computing device To the target vehicle computing device The end time of the last link on the communication path z, the above formula shows that EST and LFT are closely related; thus, the end time of task i is obtained; ;in Indicates the task In the target vehicle computing device The estimated start time of Indicates the task In the target vehicle computing device Finally, define makespan as the completion time of the task flow. Makespan is determined by the completion time of the end task. The objective function of the problem is to minimize the completion time, which is defined by the following formula: ,in Indicates the end of the task In the target vehicle computing device The completion time of the place; Step 5: Design a task scheduling algorithm based on communication link occupancy and predicted priority, calculate task priority, determine the execution order of tasks, and schedule tasks to the computing devices of the vehicle cooperation cluster for execution; In step 5, the task scheduling algorithm based on communication link occupancy and predicted priority is the PPTS_CC algorithm, which uses the concept of predictability to determine the task priority stage and the vehicle computing device selection stage. First, based on the prediction cost matrix Calculate the priority of each task, then select a vehicle with an unoccupied communication link to perform the task, and select the vehicle computing device with the shortest completion time among all available vehicles for the task; define PCM as a A matrix where each element Indicates the task In each vehicle computing device The maximum value of the upper priority; Each task is recursively calculated by traversing the given vehicle task flow model from the end task to the start task; It is determined recursively by the following equation: ,in express The subsequent task, Indicates the task With the task In-vehicle computing equipment Average transfer time on ; The first stage of task scheduling is to prioritize tasks. First, calculate each task The priority value is Indicates, indicates tasks The average PCM value is given by: ; In step 5, during the vehicle computing device selection phase, first, the forward prediction value of each task on the arriving vehicle computing device is calculated. ; Then, select The smallest vehicle computing device performs the task and establishes a path between the source vehicle computing device and the target vehicle computing device. Mark as occupied until the task data transmission is completed; the goal of this stage is to ensure that the successor tasks of the current task can be completed in advance without increasing the complexity of the algorithm; forward prediction value The calculation method is as follows: .

Images