Vehicle-road cloud integrated cooperative computing method

Abstract

The invention discloses a vehicle-road cloud integrated cooperative computing method, and relates to the technical field of surface mine unmanned driving, and the method comprises the steps: defining a computing dependency relationship among a vehicle end, a road end and a cloud end, and constructing a computing relevancy index based on the computing dependency relationship for quantifying the dependency influence degree between the vehicle end and the road end or the cloud end; wherein the calculation dependency relationship comprises strong dependency, weak dependency and no dependency; respectively establishing calculation models for the vehicle end, the road end and the cloud end according to task real-time requirements and resource constraints; the task allocation of the vehicle end, the road end and the cloud end is globally optimized through a genetic algorithm with the goal of minimizing the maximum task completion time and the total energy consumption, and the constraint condition is that each task is only executed by a single node of the vehicle end, the road end or the cloud end. Three layers of resources of vehicles, roads and cloud are cooperated, so that the defect of insufficient vehicle-mounted computing power is made up, and the overall efficiency of the system is improved; the energy utilization efficiency is improved through refined task allocation and balanced calculation and transmission energy consumption.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of open-pit mine unmanned driving, and in particular to a vehicle-road-cloud integrated collaborative computing method. BACKGROUND

[0002] The open-pit mine unmanned driving system involves integrated collaborative computing of vehicle end, road end and cloud end, needs to process high-complexity computing tasks such as environment perception, path planning and real-time control, and simultaneously faces technical challenges of large data transmission volume and strict real-time requirements.

[0003] At present, on the one hand, the vehicle-mounted terminal is limited by physical space and energy consumption, and it is difficult to integrate high-performance chips, resulting in high delay in processing complex tasks, and even unable to complete in real time, directly affecting the safety of unmanned driving; on the other hand, if all tasks are uploaded to the cloud for processing, a large amount of vehicle-mounted sensor data needs to be transmitted through a wireless communication link, which is easily affected by mine terrain (such as valleys and obstacles) to cause unstable transmission delay, which may cause decision lag or system out of control; in addition, in the open-pit mine scene, the road end (such as fixed base stations and edge servers) usually has local computing and storage capabilities, but the existing technology does not establish an effective vehicle-road-cloud collaborative mechanism, resulting in idle road end resources that cannot make up for the defects of vehicle-mounted and cloud.

[0004] Therefore, there is an urgent need for a vehicle-road-cloud collaborative computing strategy to utilize the local computing capabilities of the road end, integrate vehicle end, road end and cloud end resources, optimize task allocation and scheduling mechanism, and improve resource utilization efficiency and system robustness while ensuring real-time performance. SUMMARY

[0005] To solve the above technical problems, the present application provides a vehicle-road-cloud integrated collaborative computing method, which comprises: Defining the computing dependency relationship between the vehicle end, the road end and the cloud end, based on the computing dependency relationship, constructing a computing correlation index for quantifying the dependency influence degree between the vehicle end and the road end or the cloud end; wherein the computing dependency relationship includes strong dependency, weak dependency and no dependency; According to the real-time requirement of the task and the resource constraint, a computing model is established for the vehicle end, the road end and the cloud end respectively; Minimizing the maximum task completion time and total energy consumption as the target, the global optimization of vehicle end, road end and cloud end task allocation is carried out through genetic algorithm, and the constraint condition is that each task is executed by a single node of the vehicle end, the road end or the cloud end.

[0006] Further, the strong dependency means that part of the vehicle end computing task must be completed by the road end or the cloud end; the weak dependency means that the vehicle end can independently complete the task but the participation of the road end or the cloud end can improve the efficiency; and the no dependency means that the vehicle end completely independently completes the task.

[0007] Further, the calculation dependency relationship is refined as: The vehicle end and the road end / cloud end are mutually strong dependent, marked as ; The vehicle end is strongly dependent on the road end / cloud end, and the road end / cloud end is weakly dependent on the vehicle end, marked as ; The vehicle end is strongly dependent on the road end / cloud end, and the road end / cloud end is not dependent on the vehicle end, marked as ; The vehicle end and the road end / cloud end are mutually weak dependent, marked as ; The vehicle end is weakly dependent on the road end / cloud end, and the road end / cloud end is not dependent on the vehicle end, marked as ; The vehicle end and the road end / cloud end are mutually independent, marked as .

[0008] Further, the road end server resource set is , and the parameters of each road end server are a three-tuple , wherein is the upload bandwidth, is the road end computing capability, is the download bandwidth.

[0009] Further, the cloud end server parameter is a three-tuple , wherein is the upload bandwidth, is the cloud end computing capability, is the download bandwidth.

[0010] Further, the computing model of the vehicle end, the road end, and the cloud end includes: Vehicle end computing model: the vehicle end task completion time is , wherein is the number of resources required for the computing task; and the vehicle end computing energy consumption is ; Road end computing model: the road end task completion time is , wherein is the input data volume of the computing task; is the output data volume of the computing task; and the road end computing energy consumption is ; Cloud end computing model: the cloud end task completion time is , wherein is the upload bandwidth of the task; is the cloud server computing capability; is the task download bandwidth; and the cloud end computing energy consumption is .

[0011] Further, the optimization process of the genetic algorithm comprises: According to the calculation of the correlation degree, the resource conflict task is solved; An optimal task allocation scheme is generated through crossover, mutation and selection operations; The task allocation proportion of the vehicle end, the road end and the cloud end is dynamically adjusted to balance the real-time performance and the energy consumption.

[0012] Further, the method is applied to the unmanned driving scene of the open-pit mine, and the road end server is deployed at the edge of the mine operation area.

[0013] Compared with the prior art, the embodiments of the present application have the following beneficial effects: On the one hand, the road end is calculated nearby to reduce the transmission delay of the cloud and meet the high real-time requirement of the unmanned driving task; on the other hand, the need for directly uploading a large amount of data to the cloud is reduced, and the network load is reduced; in addition, the resources of the vehicle, the road and the cloud are coordinated to make up for the deficiency of insufficient on-board computing power, and the overall efficiency of the system is improved; and through the refined task allocation model, the computing and transmission energy consumption is balanced, and the energy utilization efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0015] Figure 1 The overall flowchart disclosed in the embodiments of the present application. DETAILED DESCRIPTION

[0016] In order to enable those skilled in the art to better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0017] The application aims to provide a vehicle-road-cloud integrated collaborative computing method, which is applied to an open-pit mine unmanned driving scene, and a road end server is deployed at the edge of a mine operation area. According to the real-time requirement of a task, a computing task is dynamically allocated to a vehicle end, a road end or a cloud end. High real-time tasks are preferentially processed by the road end server, and the dependence on the cloud end is reduced. The dynamic collaborative scheduling of vehicle end, road end and cloud end resources is used to solve the problems of insufficient vehicle computing power, high cloud delay and high network bandwidth pressure, so as to improve the real-time performance, energy efficiency and robustness of the unmanned driving system.

[0018] Please refer to Figure 1 The method mainly includes the following steps: S1, define the computing dependency relationship between the vehicle end, the road end and the cloud end, and construct a computing correlation index based on the computing dependency relationship, which is used to quantify the dependency influence degree between the vehicle end and the road end or the cloud end; wherein the computing dependency relationship includes strong dependency, weak dependency and no dependency.

[0019] The person skilled in the art further explains that, on the special network of vehicle-road-cloud integration in the open-pit mine, although each computing member of the vehicle end, the road end and the cloud end is an independent entity, due to the flow of computing tasks between the preceding and subsequent links, the computing dependency relationship between the preceding and subsequent links is generated. The closeness of this computing dependency relationship is called computing dependency.

[0020] Among them, the strong dependency means that the vehicle end computing task part must be completed by the road end or the cloud end; the weak dependency means that the vehicle end can independently complete the task, but the participation of the road end or the cloud end can improve the efficiency; and the no dependency means that the vehicle end completely independently completes the task.

[0021] In the further scheme of the embodiment, it is assumed that two directly connected computing members are x and y, the computing processes are a and b respectively, and the outputs are , Then, there are three kinds of computing dependencies between x and y: strong dependency, weak dependency and no dependency, specifically: (1) Expression is called x strongly dependent on y, which means that the output of x cannot be independently completed, and part of the work in the computing process a must be completed by relying on y.

[0022] (2) Expression is called x weakly dependent on y, which means that x can independently output , but if the work in the computing process a is replaced by y, the computing efficiency will be higher.

[0023] (3) Expression is called x not dependent on y, which means that x can independently calculate , and all the work in the computing process a is irrelevant to y.

[0024] According to different situations of the task allocation of the vehicle-road-cloud cooperative computing, the computing dependency relationship is refined into six types, as shown in the following table: Table 1. Computing dependency relationship

[0025] According to the computing dependency relationship shown in Table 1, the following can be summarized: The vehicle end and the road end / cloud end are mutually strongly dependent, marked as . The vehicle end is strongly dependent on the road end / cloud end, and the road end / cloud end is weakly dependent on the vehicle end, marked as . The vehicle end is strongly dependent on the road end / cloud end, and the road end / cloud end is not dependent on the vehicle end, marked as . The vehicle end and the road end / cloud end are mutually weakly dependent, marked as . The vehicle end is weakly dependent on the road end / cloud end, and the road end / cloud end is not dependent on the vehicle end, marked as . The vehicle end and the road end / cloud end are mutually independent, marked as .

[0026] In the further scheme of the embodiment, the calculation of y is stopped, which may affect the normal production of x. The degree of calculation correlation between y and x is referred to as a calculation correlation degree D[y x], which takes values of {0, 1, 2, 3, 4, 5, 6}; specifically, D[y x]=0, D[y x]=1, D[y x]=2, D[y x]=3, D[y x]=4, D[y x]=5, D[x x]=6.

[0027] (1) When the computing dependency relationship between x and y belongs to type 1, that is, x and y are mutually strongly dependent, if one of the members stops computing, the other member must stop computing.

[0028] (2) When the computing dependency relationship between x and y belongs to types 2 and 3, that is, x is strongly dependent on y, and y is weakly (not) dependent on x, if y stops computing, x must stop computing; if x stops computing, y will not stop computing.

[0029] (3) When the computing dependency relationship between x and y belongs to types 4, 5 and 6, that is, x and y are mutually weakly (not) dependent, if one of the computing members stops computing, the other computing member will not stop computing.

[0030] S2, according to the task real-time requirement and resource constraint, respectively, the vehicle end, road end, cloud end to establish a computational model.

[0031] Specifically: The vehicle end uses to represent, wherein, is the computing power, is the computing power, is the transmission power, is the download power, is the idle power.

[0032] The road end is composed of a plurality of road end servers, assuming that there are road end servers, the road end server set is , and each road end server The parameters of the road end server are a three-tuple , wherein, is the upload bandwidth, is the road end computing power, is the download bandwidth.

[0033] The cloud end is composed of a plurality of cloud end servers , wherein, is the upload bandwidth, is the cloud computing power, is the download bandwidth.

[0034] In the further scheme of the embodiment, the computing model of the vehicle end, the road end and the cloud end includes: Vehicle end computing model: the vehicle end task completion time is , wherein, is the number of resources required for computing tasks; the vehicle end computing energy consumption is ; the task only considers the task computing time when computing in the vehicle end.

[0035] Road end computing model: the road end task completion time is , wherein, is the input data volume of the computing task; is the data volume output by the computing task; the road end computing energy consumption is .

[0036] Cloud end computing model: the cloud end task completion time is , wherein, is the upload bandwidth of the task; is the cloud server computing power; is the task download bandwidth; the cloud end computing energy consumption is .

[0037] Suppose there are tasks in the vehicle, tasks in the road, tasks in the cloud, then the vehicle-road-cloud integrated computing time is:

[0038] The total energy consumption of the vehicle-road-cloud integrated computing task is:

[0039] where , , {0,1}, indicating that the task can only be calculated in one of the vehicle, road, and cloud.

[0040] S3, to minimize the maximum task completion time and total energy consumption, the genetic algorithm is used to globally optimize the task allocation of the vehicle, road, and cloud, and the constraint condition is that each task is executed by a single node of the vehicle, road, or cloud.

[0041] In the further scheme of the embodiment, the optimization process of the genetic algorithm includes: S31, according to the calculation correlation, solve the resource conflict task.

[0042] S32, generate the optimal task allocation scheme through crossover, mutation, and selection operations.

[0043] S33, dynamically adjust the task allocation proportion of the vehicle, road, and cloud to balance the real-time performance and energy consumption.

[0044] Specifically, integer coding is used, and each individual (chromosome) represents a task allocation scheme. The length of the chromosome is the total number of tasks , and the gene bit value is {0, 1, 2}, which represents the allocation of the task to the vehicle, road, or cloud. An initial population is randomly generated to ensure a diverse allocation scheme. In the coding process, the constraint condition of "each task is executed only in one place" is forcibly met, that is, each gene bit has a unique value.

[0045] In the further scheme, the fitness value of the objective function is calculated by weighting the maximum task completion time and the total energy consumption :

[0046] where , by default =0.7, =0.3, and The historical maximum reference value is used for normalization.

[0047] The fitness value of the individual violating the resource conflict (such as the overload of the road end server) is increased by a penalty factor =10. Tournament selection is used to randomly select =5 individuals from the population, and the individual with the optimal fitness is selected to enter the next generation. The top 5% elite individuals are retained in each generation to avoid loss of high-quality genes. The crossover probability =0.8; a gene site is randomly selected, and its value is changed according to the probability =0.05, such as changing "vehicle end" to "road end". If the fitness is not improved for 10 generations, the mutation probability is increased to =0.1.

[0048] The mechanism of dynamic task allocation ratio adjustment is as follows: the task allocation ratios of the vehicle end, the road end and the cloud end are counted for 10 generations, and if the utilization rate of a layer of resources is lower than a threshold value (such as the vehicle end <20%), the mutation operation is used to guide the task migration to the layer.

[0049] The application can be widely applied to unmanned driving scenes such as open-pit mines, ports and logistics parks, and is especially suitable for environments with complex network conditions and high real-time requirements, and has significant economic benefits and technical popularization value.

[0050] Although the embodiments of the application have been shown and described, it can be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the application, and the scope of the application is defined by the appended claims and their equivalents.

Claims

1. A vehicle-road-cloud integrated collaborative computing method, characterized in that, The method includes: Define the computational dependencies between the vehicle, road, and cloud. Based on these dependencies, construct a computational relevance index to quantify the degree of dependency between the vehicle and the road or cloud. The computational dependencies include strong dependencies, weak dependencies, and no dependencies. Based on the real-time requirements of the task and resource constraints, computing models are established for the vehicle end, road end, and cloud end respectively; With the goal of minimizing the maximum task completion time and total energy consumption, a genetic algorithm is used to globally optimize the task allocation on the vehicle, road, and cloud sides, with the constraint that each task is executed by only a single node on the vehicle, road, or cloud side.

2. The vehicle-road-cloud integrated collaborative computing method according to claim 1, characterized in that, The term "strong dependency" indicates that the vehicle-side computing task must be partially completed by the roadside or cloud; the term "weak dependency" indicates that the vehicle-side can complete the task independently, but the participation of the roadside or cloud can improve efficiency; and the term "no dependency" indicates that the vehicle-side completes the task completely independently.

3. The vehicle-road-cloud integrated collaborative computing method according to claim 1 or 2, characterized in that, The computational dependencies are refined as follows: The vehicle-side and the roadside / cloud-side are strongly interdependent, marked as ; The vehicle side is strongly dependent on the roadside / cloud side, and the roadside / cloud side is weakly dependent on the vehicle side. ; Vehicle-side devices are strongly dependent on roadside / cloud-side devices, while roadside / cloud-side devices are independent of vehicle-side devices; these are marked as such. ; The vehicle and road / cloud terminals have a weak dependency on each other, marked as ; Vehicle-side weakly dependent on roadside / cloud-side, roadside / cloud-side not dependent on vehicle-side, marked as ; The vehicle-side and the roadside / cloud-side are independent of each other, marked as .

4. The vehicle-road-cloud integrated collaborative computing method according to claim 1, characterized in that, The roadside server resource set is as follows: Each roadside server The parameter is a triplet ,in, For upload bandwidth, For roadside computing capabilities, This refers to the download bandwidth.

5. The vehicle-road-cloud integrated collaborative computing method according to claim 1, characterized in that, The cloud server parameters are triples. Composition, among which, For upload bandwidth, For cloud computing capabilities, This refers to the download bandwidth.

6. The vehicle-road-cloud integrated collaborative computing method according to claim 1 or 5, characterized in that, The computing models for the vehicle, road, and cloud terminals include: Vehicle-side computing model: Vehicle-side task completion time is In the formula, The number of resources required to calculate the task; the energy consumption of vehicle-side computing is... ; Roadside computation model: Roadside task completion time is In the formula, The amount of input data for the computation task; The amount of data output by the calculation task; the power consumption of the roadside calculation is ; Cloud computing model: Cloud task completion time In the formula, For the upload bandwidth of the task; For cloud server computing power; The download bandwidth for the task; cloud computing energy consumption is... .

7. The vehicle-road-cloud integrated collaborative computing method according to claim 1, characterized in that, The optimization process of the genetic algorithm includes: Based on the calculated relevance, resolve resource conflict tasks; The optimal task allocation scheme is generated through crossover, mutation, and selection operations; Dynamically adjust the task allocation ratio between vehicle, road, and cloud terminals to balance real-time performance and energy consumption.

8. The vehicle-road-cloud integrated collaborative computing method according to any one of claims 1-7, characterized in that, The method is applied to unmanned driving scenarios in open-pit mines, with the roadside server deployed at the edge of the mine operation area.

Images