Methods, devices, equipment, vehicles and systems for calculating the real-time price index of on-board computing power

Abstract

This application provides a method, apparatus, device, vehicle, and system for calculating the real-time price index of onboard computing power. The method includes: acquiring the hardware specifications of the onboard processors of target vehicle nodes distributed within a preset parking physical area; querying a computing power conversion mapping table to extract benchmark conversion coefficients; using the benchmark conversion coefficients to convert the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence; receiving a task processing request message and parsing it to obtain the benchmark identifier of the target computing power unit and the computing power acquisition threshold parameter; retrieving a matching set of expected computing power supply parameters and performing ascending sorting to generate a supply ladder distribution sequence; using the sequence node values ​​to sequentially perform bidirectional data constraint iterative operations on the computing power acquisition threshold parameter, extracting the converged value as the benchmark price value under convergence conditions; and broadcasting the benchmark price value according to an adaptive refresh cycle.

Description

Technical Field

[0001] This application relates to the fields of data processing and distributed control technology, and more specifically, to a method, apparatus, electronic device, vehicle, and system for calculating the real-time price index of vehicle computing power. Background Technology

[0002] With the continuous growth of onboard computing power in new energy vehicles, distributed networks with parked vehicles as supply nodes are gradually taking shape. Reasonable evaluation and coordinated scheduling of heterogeneous onboard processing resources under different parking scenarios have become an important application scenario requirement in the field of distributed edge computing protocols.

[0003] Existing computing resource evaluation and scheduling schemes typically employ a centralized cloud architecture or a fixed edge node static allocation model. This scheme first involves a central server uniformly collecting resource availability and network bandwidth parameters from each computing node within a specific region. Then, the central server generates a fixed resource price reference table based on the collected parameters, or updates unilateral data at fixed intervals and publishes it publicly. Finally, the computing power requesting node, based on the published fixed reference table and its own task requirements, negotiates the task distribution through a point-to-point data handshake with a single node.

[0004] However, this fixed and centralized processing solution has significant technical drawbacks. Because different vehicles use different underlying processing chips with varying specifications, and vehicle entry and exit are highly dynamic in terms of time and space, existing solutions lack a unified conversion and calibration mechanism for heterogeneous hardware specifications. Furthermore, the fixed reference table relying on unidirectional data collection and distribution cannot accurately reflect the real-time changes in bidirectional data interaction, and it lacks a dynamic data iteration and convergence mechanism based on bidirectional constraints. These shortcomings result in generated guidance data that fails to reflect the true node collaboration status and is ill-suited to the frequently changing online characteristics of parked vehicles. Consequently, the efficiency of issuing scheduling instructions to computing power requesting nodes is low, failing to meet the agile scheduling requirements for resource collaboration in highly dynamic vehicular computing networks. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, vehicle, and system for calculating the real-time price index of vehicle computing power, in order to at least alleviate the aforementioned technical problems.

[0006] A method for calculating the real-time price index of vehicle-mounted computing power includes the following steps:

[0007] The hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area are obtained. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients.

[0008] The system receives a task processing request message sent by a computing power request node, parses the task processing request message to obtain the target computing power unit baseline identifier, and extracts the computing power acquisition threshold parameter based on the task processing request message.

[0009] Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0010] Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value.

[0011] According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the network interface to instruct the computing power request node to generate a target scheduling instruction.

[0012] Optionally, the step of converting the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients includes:

[0013] Extract the integer precision computing power index of the target vehicle node, multiply the integer precision computing power index using the first coefficient in the benchmark conversion coefficient to obtain the integer precision standard computing power unit value, and establish a correspondence between the integer precision standard computing power unit value and the integer precision standard computing power unit benchmark identifier;

[0014] Extract the floating-point precision computing power index of the target vehicle node, multiply the floating-point precision computing power index using the second coefficient in the reference conversion coefficient to obtain the floating-point precision standard computing power unit value, and establish a correspondence between the floating-point precision standard computing power unit value and the floating-point precision standard computing power unit reference identifier.

[0015] The integer precision standard computing power unit value, the integer precision standard computing power unit reference identifier, the floating point precision standard computing power unit value, and the floating point precision standard computing power unit reference identifier are all written into the dual reference standard computing power unit sequence.

[0016] Optionally, the step of performing bidirectional data constraint iteration operations on the computing power acquisition threshold parameter sequentially using the sequence node values ​​includes:

[0017] Set the computing power acquisition threshold parameter as the first constraint condition, and set the current sequence node value in the sequence node values ​​as the second constraint condition;

[0018] In a single data loop processing process, the first constraint is applied to the supply ladder distribution sequence to filter out a subset of available nodes that meet the first constraint, and the corresponding first node response parameter is calculated based on the subset of available nodes.

[0019] Apply the second constraint to the first node response parameter to calculate the corresponding second node request parameter;

[0020] The response parameters of the first node are compared with the request parameters of the second node, and the deviation value of the comparison output is used as the iteration index of the current data loop processing process, and the iteration index is used as the basis for determining whether the deviation value falls into the preset convergence interval.

[0021] If the deviation value does not fall within the preset convergence interval, the next data loop processing process is triggered based on the value of the next sequence node in the supply ladder distribution sequence.

[0022] Optionally, before the step of broadcasting the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh period, the method further includes:

[0023] Collect the total number of available vehicle nodes in the preset parking physical area that are in a network connected state, and count the total number of pending tasks to which the task processing request message belongs;

[0024] When the total number of available vehicle nodes is greater than the first preset node number threshold, or when the total number of pending tasks in the queue is greater than the first preset queue number threshold, the adaptive refresh cycle is set to the first cycle duration parameter.

[0025] When the total number of available vehicle nodes is lower than the second preset node number threshold and the total number of pending tasks in the queue is lower than the second preset queue number threshold, the adaptive refresh cycle is set to the second cycle duration parameter.

[0026] The time interval for broadcasting the benchmark price value to the preset node cluster via the network interface is controlled according to the first cycle duration parameter or the second cycle duration parameter.

[0027] Optionally, before the step of obtaining the hardware specification characteristics of the on-board processors of the target vehicle nodes distributed within the preset parking physical area, the method further includes:

[0028] Obtain the spatial location attribute information of the preset parking physical area, match the corresponding computing power supply stability coefficient based on the spatial location attribute information, and obtain the credibility weight coefficient corresponding to the preset parking physical area.

[0029] The computing power supply stability coefficient is multiplied by the credibility weight coefficient to obtain the node collaboration comprehensive weight value of the preset stationed physical area;

[0030] After obtaining the benchmark price value, the node collaborative comprehensive weight value is multiplied by the benchmark price value to obtain the regional weighted price adjustment parameter;

[0031] The regional weighted price adjustment parameter is used as an alternative broadcast result after regional weighting of the benchmark price value, and the regional weighted price adjustment parameter is broadcast to the preset node cluster through the network interface.

[0032] Optionally, the step of obtaining the spatial location attribute information of the preset parking physical area includes:

[0033] Collect the geographic coordinate point dataset of the preset parking physical area, input the geographic coordinate point dataset into the preset point of interest location parsing program to output the location classification label, and use the location classification label as the spatial location attribute information;

[0034] When the location parsing program for points of interest does not output the missing location classification label, the statistical data on vehicle dwell time distribution of the preset parking physical area within a historical set period is retrieved. Based on the statistical data on vehicle dwell time distribution, an alternative location classification label is determined, and the alternative location classification label is used to replace the original location classification label. The alternative location classification label is then used as the spatial location attribute information.

[0035] Optionally, in the step of parsing the task processing request message to obtain the target computing unit reference identifier, if the parsed and extracted target computing unit reference identifier contains mixed precision indication information, the method further includes:

[0036] Extract the first execution stage data block and the second execution stage data block from the task processing request message;

[0037] The first computational load corresponding to the data block in the first execution stage and the second computational load corresponding to the data block in the second execution stage are calculated respectively.

[0038] A total computing load is generated based on the first computing load and the second computing load. A first percentage parameter of the first computing load in the total computing load is calculated based on the total computing load. A second percentage parameter of the second computing load in the total computing load is calculated based on the total computing load.

[0039] The reference identifier for the standard computing power unit with integer precision is assigned according to the first proportion parameter, and the reference identifier for the standard computing power unit with floating-point precision is assigned according to the second proportion parameter.

[0040] Based on the integer precision standard computing power unit reference identifier, the first proportion parameter, the floating point precision standard computing power unit reference identifier, and the second proportion parameter combination, a hybrid reference target computing power unit reference identifier is generated, and the hybrid reference target computing power unit reference identifier is used as the target computing power unit reference identifier.

[0041] Optionally, after the step of triggering the next data loop processing process based on the next sequence node value in the supply ladder distribution sequence, provided that the deviation value does not fall within the preset convergence interval, the method further includes:

[0042] Record the cumulative number of times the data loop processing process is executed;

[0043] Determine whether the cumulative execution count has reached the preset execution count limit parameter;

[0044] Under abnormal triggering conditions where the cumulative execution count reaches the preset execution count limit parameter, the bidirectional data constraint iteration operation is stopped;

[0045] Retrieve the benchmark price value broadcast in the previous period and extract the initial sequence node value in the supply ladder distribution sequence;

[0046] A weighted average calculation is performed on the benchmark price value broadcast in the previous period and the initial sequence node value to extract the weighted average calculation result as the compensation benchmark price value, and then the compensation benchmark price value is used to replace the benchmark price value in the current period.

[0047] Optionally, after the step of broadcasting the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh period, the method further includes:

[0048] The benchmark price value, along with the corresponding timestamp information, is written into the historical sequence database;

[0049] Extract the historical benchmark price value sequence within a preset backtracking time period from the historical sequence database, and perform a moving weighted average processing operation on the historical benchmark price value sequence to obtain the moving average parameter;

[0050] Calculate the data distribution slope parameter corresponding to the supply ladder distribution sequence, and perform a fusion processing operation on the data distribution slope parameter and the moving average parameter to generate a prediction interval value;

[0051] The predicted interval values ​​are sent to the computing power request node to instruct the computing power request node to plan the task access timing.

[0052] A vehicle-mounted computing power real-time price index calculation device, comprising:

[0053] The standard computing power conversion module is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area, query the preset computing power conversion mapping table based on the hardware specifications, extract the benchmark conversion coefficients corresponding to the hardware specifications, and then use the benchmark conversion coefficients to convert the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence.

[0054] The request message parsing module is used to receive the task processing request message sent by the computing power request node, parse the task processing request message to obtain the target computing power unit benchmark identifier, and extract the computing power acquisition threshold parameter based on the task processing request message.

[0055] The supply ladder generation module is used to retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit in the dual benchmark standard computing power unit sequence, and perform an ascending sorting operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0056] The bidirectional iterative pricing module is used to extract the sequence node values ​​in the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iterative operations on the computing power acquisition threshold parameter in sequence to obtain continuous iterative output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iterative operations falls within the preset convergence interval, the convergence value of the final iterative output is extracted as the benchmark price value.

[0057] The benchmark price broadcasting module is used to broadcast the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh cycle, so as to instruct the computing power requesting node to generate a target scheduling instruction.

[0058] An electronic device includes a processor and a memory, the memory storing a computer program, and the processor performing the following processing when running the computer program:

[0059] The hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area are obtained. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients.

[0060] The system receives a task processing request message sent by a computing power request node, parses the task processing request message to obtain the target computing power unit baseline identifier, and extracts the computing power acquisition threshold parameter based on the task processing request message.

[0061] Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0062] Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value.

[0063] According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the network interface to instruct the computing power request node to generate a target scheduling instruction.

[0064] A vehicle includes an onboard processor, an onboard communication interface, and an onboard memory. The onboard memory stores a real-time price index calculation program for onboard computing power. When the onboard processor runs the real-time price index calculation program for onboard computing power, it performs the following processing:

[0065] The vehicle communication interface is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients.

[0066] The system receives task processing request messages sent by computing power request nodes through the vehicle communication interface, parses the task processing request messages to obtain the target computing power unit baseline identifier, and extracts computing power acquisition threshold parameters based on the task processing request messages.

[0067] Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0068] Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value.

[0069] According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the vehicle communication interface to instruct the computing power request node to generate a target scheduling instruction.

[0070] A real-time price index calculation system for vehicle-mounted computing power includes a target vehicle node, a computing power request node, a price index calculation node, and a preset node cluster.

[0071] The target vehicle node is used to send the hardware specifications and available computing power of the on-board processor to the price index calculation node.

[0072] The computing power request node is used to send a task processing request message to the price index calculation node, and receive the benchmark price value broadcast by the price index calculation node or receive the benchmark price value forwarded by the preset node cluster, so as to generate a target scheduling instruction.

[0073] The price index calculation node is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area, query the preset computing power conversion mapping table based on the hardware specifications, extract the benchmark conversion coefficient corresponding to the hardware specifications, and then use the benchmark conversion coefficient to convert the available computing power of each target vehicle node into a dual benchmark standard computing power unit sequence.

[0074] The price index calculation node is also used to receive the task processing request message sent by the computing power request node, parse the task processing request message to obtain the target computing power unit benchmark identifier, and extract the computing power acquisition threshold parameter based on the task processing request message.

[0075] The price index calculation node is also used to retrieve the set of expected computing power supply parameters that match the target computing power unit benchmark identifier in the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0076] The price index calculation node is also used to extract the sequence node values ​​in the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, the convergence value of the final iteration output is extracted as the benchmark price value.

[0077] The price index calculation node is also used to broadcast the benchmark price value to the preset node cluster via a network interface according to a preset adaptive refresh cycle.

[0078] The preset node cluster is used to receive the benchmark price value broadcast by the price index calculation node, and send the benchmark price value to the computing power request node to instruct the computing power request node to generate the target scheduling instruction.

[0079] The technical advantages of the technical solution provided in this application are:

[0080] This application's method for calculating the real-time price index of onboard computing power addresses the shortcomings of existing technologies, such as the lack of a unified conversion and calibration mechanism for heterogeneous hardware specifications and the reliance on fixed reference tables that rely on one-way data collection and publication, which fail to accurately reflect changes in bidirectional data interaction and result in low efficiency in issuing scheduling commands. By acquiring the hardware specification characteristics of the onboard processors of target vehicle nodes distributed within a preset parking physical area, and querying a computing power conversion mapping table to extract benchmark conversion coefficients, the available computing power is converted into a dual-benchmark standard computing power unit sequence. This solves the problem of inconsistent measurement and comparison of computing power across different chip hardware specifications in traditional solutions. Compared to traditional solutions that directly use raw heterogeneous parameter data for discrete scheduling, this application utilizes a mapping table and benchmark conversion coefficients to convert the heterogeneous available computing power of each node into a unified dual-benchmark sequence format, achieving standardized conversion of heterogeneous resources and laying a data benchmark for the subsequent generation of objective and fair guidance data.

[0081] Based on the generated dual-benchmark standard computing power unit sequence, this application further receives task processing request messages and extracts computing power acquisition threshold parameters. It then retrieves the corresponding set of expected computing power supply parameters to generate a supply tiered distribution sequence. Using the sequence node values, it sequentially performs bidirectional data constraint iteration operations on the computing power acquisition threshold parameters, extracting the converged value as the benchmark price value. This solves the problem of traditional schemes relying on unilateral fixed static data, which prevents the formation of an effective node collaborative state. Compared to the unidirectional fixed acquisition of traditional schemes, this application introduces bidirectional constraint iteration verification based on supply tiered data and request threshold parameters. This ensures that the final output benchmark value is an objective result of multiple cross-validations and convergence of supply and demand data, significantly improving the rationality and consensus of the generated guidance data.

[0082] Finally, this application broadcasts the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh cycle, instructing computing power requesting nodes to generate target scheduling instructions. This solves the problem that traditional solutions struggle to adapt to the high-frequency dynamic entry and exit of parked vehicles. Compared to traditional solutions that rely on fixed, long-term information dissemination, this application combines adaptive refresh cycles with network broadcasting, enabling computing power requesting nodes to obtain more timely benchmark values. This allows subsequent scheduling instructions to closely match the real-time online status of parked vehicle nodes, significantly improving both the scheduling agility and instruction execution effectiveness of the overall computing network. Attached Figure Description

[0083] Figure 1 This application provides an embodiment of a real-time price index calculation scenario for vehicle-mounted computing power.

[0084] Figure 2 This application provides a method for calculating the real-time price index of vehicle-mounted computing power.

[0085] Figure 3 This application provides an embodiment of a vehicle-mounted computing power real-time price index calculation device.

[0086] Figure 4 This is an electronic device according to an embodiment of the present application.

[0087] Figure 5 This is a computer-readable storage medium according to an embodiment of the present application.

[0088] Figure 6 This application provides an embodiment of a vehicle-mounted computing power real-time price index calculation system. Detailed Implementation

[0089] like Figure 1 The image shows a scenario for calculating a real-time price index for vehicle-mounted computing power according to an embodiment of this application; as shown... Figure 2The image shows a method for calculating the real-time price index of vehicle-mounted computing power according to an embodiment of this application. The method includes the following steps:

[0090] The hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area are obtained. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients.

[0091] The system receives a task processing request message sent by a computing power request node, parses the task processing request message to obtain the target computing power unit baseline identifier, and extracts the computing power acquisition threshold parameter based on the task processing request message.

[0092] Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0093] Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value.

[0094] According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the network interface to instruct the computing power request node to generate a target scheduling instruction.

[0095] Optionally, the step of converting the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients includes:

[0096] Extract the integer precision computing power index of the target vehicle node, multiply the integer precision computing power index using the first coefficient in the benchmark conversion coefficient to obtain the integer precision standard computing power unit value, and establish a correspondence between the integer precision standard computing power unit value and the integer precision standard computing power unit benchmark identifier;

[0097] Extract the floating-point precision computing power index of the target vehicle node, multiply the floating-point precision computing power index using the second coefficient in the reference conversion coefficient to obtain the floating-point precision standard computing power unit value, and establish a correspondence between the floating-point precision standard computing power unit value and the floating-point precision standard computing power unit reference identifier.

[0098] The integer precision standard computing power unit value, the integer precision standard computing power unit reference identifier, the floating point precision standard computing power unit value, and the floating point precision standard computing power unit reference identifier are all written into the dual reference standard computing power unit sequence.

[0099] Preferably, the specific implementation process of converting the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficient is as follows: After the hardware specification features are obtained, the vehicle processor model identifier, vehicle processor precision type identifier, vehicle processor nominal computing power record, and vehicle processor availability status record are first read from the hardware specification features; wherein, the vehicle processor model identifier is used to locate the target vehicle node mapping entry corresponding to the target vehicle node in the computing power conversion mapping table, the target vehicle node mapping entry is used to carry the benchmark conversion coefficient corresponding to the target vehicle node, the vehicle processor precision type identifier is used to distinguish the integer precision computing capability and floating-point precision computing capability that the target vehicle node can provide, the vehicle processor nominal computing power record is used to provide the original computing power value of the target vehicle node under the corresponding precision type, and the vehicle processor availability status record is used to limit the range of available computing power that the target vehicle node can currently participate in the conversion process. Using the above reading method, the subsequent conversion process does not directly use the original computing power values ​​reported by different vehicles. Instead, it first breaks down the hardware specification features into structured fields that can be recognized by the computing power conversion mapping table, and locates the target vehicle node mapping entry based on the structured fields, so that the extraction process of the baseline conversion coefficient can correspond to the actual on-board processor specifications of the target vehicle node.

[0100] Preferably, during the pre-setting process of the computing power conversion mapping table, the computing power conversion mapping table is configured according to the vehicle processor model identifier, the vehicle processor precision type identifier, the integer precision reference field, the floating-point precision reference field, and the reference conversion coefficient field; wherein, the integer precision reference field is used to record the unit time computing power conversion relationship based on integer eight-bit precision, the floating-point precision reference field is used to record the unit time computing power conversion relationship based on half-precision floating-point calculation precision, and the reference conversion coefficient field is used to record the first coefficient corresponding to the integer precision reference field and the second coefficient corresponding to the floating-point precision reference field. After the computing power conversion mapping table is configured with the above fields, the vehicle processor model identifier can locate the corresponding target vehicle node mapping entry. The reference conversion coefficient field in the target vehicle node mapping entry can provide the reference conversion coefficient. The vehicle processor precision type identifier can distinguish the first coefficient and the second coefficient from the reference conversion coefficient field, thereby enabling the integer precision computing power index to enter the integer precision standardized conversion channel and the floating-point precision computing power index to enter the floating-point precision standardized conversion channel, avoiding mixing the original computing power values ​​of different dimensions into the same conversion caliber.

[0101] Preferably, when extracting the integer precision computing power index of the target vehicle node, the nominal integer precision computing power value is first selected from the nominal computing power record of the vehicle processor according to the precision type identifier of the vehicle processor. Then, the available range of the nominal integer precision computing power value is limited according to the available status record of the vehicle processor to obtain the integer precision computing power index. The integer precision computing power index represents the available integer precision computing power that the target vehicle node can provide in the quantized inference task, rather than the full nominal computing power of the vehicle processor. For example, quantized inference tasks such as object detection, image classification, and natural language inference usually correspond to integer precision calculation. After the available range of the integer precision computing power index is limited, the computing power portion that has been occupied by the vehicle basic task can be excluded, so that when the multiplication operation is performed using the first coefficient, the currently schedulable available integer precision computing power is used in the operation, and the available integer precision computing power can continue to be converted into the integer precision standard computing power unit value along the integer precision standardization conversion channel.

[0102] Preferably, when multiplying the integer precision computing power index using the first coefficient, the first coefficient originates from the benchmark conversion coefficient field in the computing power conversion mapping table, which corresponds to both the vehicle processor model identifier and the integer precision benchmark field. The first coefficient is used to convert the original computing power values ​​of different vehicle processors at integer precision to a unified integer precision standard computing power unit caliber. During multiplication, the integer precision computing power index and the first coefficient are calculated using the same time benchmark to obtain the integer precision standard computing power unit value. The integer precision standard computing power unit value is then associated with the integer precision standard computing power unit benchmark identifier. This association indicates that the integer precision standard computing power unit value belongs to the integer precision standard computing power unit caliber, enabling the identification of the suitable integer precision standard computing power unit value from the dual benchmark standard computing power unit sequence when matching based on the target computing power unit benchmark identifier.

[0103] Preferably, when the hardware specification features also include available specifications for low-order quantization precision, the computing power conversion mapping table incorporates the precision correction relationship of low-order quantization precision into the configuration source of the first coefficient, instead of directly writing the original computing power value of low-order quantization precision into the dual-benchmark standard computing power unit sequence. Specifically, the available specifications for low-order quantization precision are first matched with the vehicle processor model identifier to determine whether the target vehicle node meets the conditions for allowable correction; when the target vehicle node meets the conditions for allowable correction, the precision correction relationship is used to adjust the first coefficient, and the adjusted first coefficient then participates in the multiplication operation of the integer precision computing power index, thereby obtaining the integer precision standard computing power unit value corresponding to the integer precision standard computing power unit benchmark identifier. By converging the difference in low-level quantization precision into the first coefficient, the dual-benchmark standard computing power unit sequence still maintains two main calibers: integer precision standard computing power unit values ​​and floating-point precision standard computing power unit values. When subsequently retrieving the expected parameter set of computing power supply, there is no need to introduce a third conversion caliber, and the third conversion caliber will not disrupt the isolation relationship between the two main calibers in the dual-benchmark standard computing power unit sequence.

[0104] Preferably, when extracting the floating-point precision computing power index of the target vehicle node, firstly, half-precision floating-point nominal computing power values ​​are selected from the nominal computing power records of the vehicle processor according to the precision type identifier of the vehicle processor. Then, the usable range of the half-precision floating-point nominal computing power values ​​is limited according to the available status records of the vehicle processor to obtain the floating-point precision computing power index. The floating-point precision computing power index characterizes the available floating-point precision computing power that the target vehicle node can provide in small model fine-tuning training, complex multimodal inference, and high-precision numerical calculation tasks. The floating-point precision computing power index is not the same as the integer precision computing power index in terms of technical meaning. Therefore, the floating-point precision computing power index is not calculated with the first coefficient, but enters the floating-point precision normalization conversion channel corresponding to the second coefficient, so that the available floating-point precision computing power can be further converted into the floating-point precision standard computing power unit value along the floating-point precision normalization conversion channel.

[0105] Preferably, when multiplying the floating-point precision computing power index using the second coefficient, the second coefficient originates from the reference conversion coefficient field in the computing power conversion mapping table, which corresponds to both the vehicle processor model identifier and the floating-point precision reference field. The second coefficient is used to convert the original computing power values ​​of different vehicle processors at floating-point precision to a unified floating-point precision standard computing power unit (DPU) caliber. During multiplication, the floating-point precision computing power index and the second coefficient are calculated using the same time reference to obtain the floating-point precision standard computing power unit (DPU) value. The DPU value is then further associated with the DPU reference identifier to establish a floating-point precision reference identifier correspondence. This correspondence indicates that the DPU value belongs to the DPU caliber, enabling the retrieval of the expected computing power supply parameter set matching the floating-point precision task from the dual-reference standard computing power unit sequence when the target computing power unit reference identifier points to a floating-point precision task.

[0106] Preferably, the integer precision standard computing unit reference identifier and the floating-point precision standard computing unit reference identifier are not simple text labels for the integer precision standard computing unit value and the floating-point precision standard computing unit value, but rather index fields that participate in subsequent matching processing together with the target computing unit reference identifier. Specifically, after the integer precision standard computing unit value is generated, the integer precision standard computing unit value, the integer precision standard computing unit reference identifier, the correspondence between the integer precision reference identifiers, and the corresponding target vehicle node identifier are written into the same integer precision standard computing record; after the floating-point precision standard computing unit value is generated, the floating-point precision standard computing unit value, the floating-point precision standard computing unit reference identifier, the correspondence between the floating-point precision reference identifiers, and the corresponding target vehicle node identifier are written into the same floating-point precision standard computing record. Both the integer precision standard computing power record and the floating-point precision standard computing power record retain the corresponding target vehicle node identifier, so that when the expected computing power supply parameter set is subsequently formed, the standardized integer precision standard computing power unit value or the standardized floating-point precision standard computing power unit value can be re-associated with the corresponding target vehicle node, avoiding the loss of node source information after standardization conversion.

[0107] Preferably, when writing the integer precision standard computing power unit value, the integer precision standard computing power unit reference identifier, the floating-point precision standard computing power unit value, and the floating-point precision standard computing power unit reference identifier into the dual-reference standard computing power unit sequence, a node-level sequence position is first established based on the target vehicle node identifier. Then, the integer precision standard computing power record and the floating-point precision standard computing power record are written into the integer precision reference branch and the floating-point precision reference branch under the same node-level sequence position. The node-level sequence position is used to maintain a one-to-one correspondence between the target vehicle node and its corresponding dual-reference conversion result. The integer precision reference branch is used to maintain the computational caliber association between the integer precision standard computing power unit value and the integer precision standard computing power unit reference identifier. The floating-point precision reference branch is used to maintain the computational caliber association between the floating-point precision standard computing power unit value and the floating-point precision standard computing power unit reference identifier. Through the joint organization of the node-level sequence position, the integer precision benchmark branch, and the floating-point precision benchmark branch, the dual-benchmark standard computing power unit sequence can express both the available integer precision computing power of each target vehicle node and the available floating-point precision computing power of the same target vehicle node, thereby providing a direct data source for subsequently retrieving the expected set of computing power supply parameters based on the target computing power unit benchmark identifier.

[0108] Preferably, in one scenario, the preset parking physical area is a parking lot area, and the target vehicle nodes include target vehicle nodes equipped with neural network acceleration units and target vehicle nodes equipped with graphics processing units. For target vehicle nodes equipped with neural network acceleration units that are mainly for quantization inference tasks, the nominal integer computing power value in the hardware specification features is converted into the integer computing power index, and the integer computing power index is multiplied by the first coefficient to form the integer standard computing power unit value; for target vehicle nodes equipped with graphics processing units that can perform half-precision floating-point calculations, the nominal half-precision floating-point computing power value in the hardware specification features is converted into the floating-point computing power index, and the floating-point computing power index is multiplied by the second coefficient to form the floating-point standard computing power unit value. Therefore, different vehicle-mounted processors in the same parking lot area no longer participate in subsequent sorting with their original computing power values, but instead participate in subsequent retrieval and ascending order sorting with the unified caliber in the dual-benchmark standard computing power unit sequence. This enables the supply ladder distribution sequence to reflect a comparable standardized computing power supply status, which continues to serve as the data basis for the subsequent bidirectional data constraint iteration operation.

[0109] Preferably, after the dual-benchmark standard computing power unit sequence is generated, a benchmark integrity check can be performed on the dual-benchmark standard computing power unit sequence to ensure the stability of the data source when subsequently retrieving the expected parameter set of computing power supply. The benchmark integrity check first checks whether the integer precision benchmark branch and the floating-point precision benchmark branch exist at each node-level sequence position, then checks whether the integer precision standard computing power unit benchmark identifier and the integer precision standard computing power unit value exist in the integer precision benchmark branch, and checks whether the floating-point precision standard computing power unit benchmark identifier and the floating-point precision standard computing power unit value exist in the floating-point precision benchmark branch; when a target vehicle node only has one type of precision computing capability, the dual-benchmark standard computing power unit sequence retains the integer precision benchmark branch or the floating-point precision benchmark branch already generated for the target vehicle node, and marks the ungenerated integer precision benchmark branch or the ungenerated floating-point precision benchmark branch as unretrievalable. The unretrievable state continues to participate in the baseline integrity verification of the dual-benchmark standard computing unit sequence, so that when matching is performed according to the target computing unit baseline identifier, target vehicle nodes that do not have the corresponding precision computing capabilities will not be written into the computing power supply expected parameter set.

[0110] Preferably, after the integer precision standard computing unit value and the floating-point precision standard computing unit value are written into the dual-reference standard computing unit sequence, the coefficient source relationship with the reference conversion coefficient is also retained respectively. The coefficient source relationship includes the integer precision coefficient source relationship between the first coefficient and the integer precision standard computing unit value, and the floating-point precision coefficient source relationship between the second coefficient and the floating-point precision standard computing unit value. The integer precision coefficient source relationship is used to retain the integer precision standard computing unit caliber used in the current period when the reference price value is subsequently written into the historical sequence database, and the floating-point precision coefficient source relationship is used to retain the floating-point precision standard computing unit caliber used in the current period when the reference price value is subsequently written into the historical sequence database. The integer precision coefficient source relationship is also used to distinguish the integer precision standard computing unit reference identifier required for different execution stages when mixed precision indication information appears in the future, and the floating-point precision coefficient source relationship is also used to distinguish the floating-point precision standard computing unit reference identifier required for different execution stages when mixed precision indication information appears in the future. By preserving the source relationships of the integer precision coefficients and the floating-point precision coefficients in the dual-benchmark standard computing unit sequence, the target computing unit benchmark identifier, the set of expected computing power supply parameters, and the supply ladder distribution sequence can be continuously transmitted along the same benchmark, reducing subsequent iteration deviations caused by differences in the original computing power values ​​of heterogeneous vehicle processors.

[0111] Optionally, the step of performing bidirectional data constraint iteration operations on the computing power acquisition threshold parameter sequentially using the sequence node values ​​includes:

[0112] Set the computing power acquisition threshold parameter as the first constraint condition, and set the current sequence node value in the sequence node values ​​as the second constraint condition;

[0113] In a single data loop processing process, the first constraint is applied to the supply ladder distribution sequence to filter out a subset of available nodes that meet the first constraint, and the corresponding first node response parameter is calculated based on the subset of available nodes.

[0114] Apply the second constraint to the first node response parameter to calculate the corresponding second node request parameter;

[0115] The response parameters of the first node are compared with the request parameters of the second node, and the deviation value of the comparison output is used as the iteration index of the current data loop processing process, and the iteration index is used as the basis for determining whether the deviation value falls into the preset convergence interval.

[0116] If the deviation value does not fall within the preset convergence interval, the next data loop processing process is triggered based on the value of the next sequence node in the supply ladder distribution sequence.

[0117] Preferably, the specific implementation process of setting the computing power acquisition threshold parameter as the first constraint condition and setting the current sequence node value in the sequence node values ​​as the second constraint condition is as follows: Before entering the bidirectional data constraint iteration operation, the target computing power unit reference identifier, the computing power acquisition threshold parameter, and the supply ladder distribution sequence are read first; wherein, the target computing power unit reference identifier is used to limit the integer precision standard computing power unit caliber or floating-point precision standard computing power unit caliber used in this bidirectional data constraint iteration operation, the computing power acquisition threshold parameter is used to limit the request constraint boundary that the computing power request node can accept under the same computing power unit caliber, and the supply ladder distribution sequence is used to provide supply-side sequence nodes that change step by step according to the ascending order of the computing power supply expected parameter set. By first reading the target computing power unit reference identifier, the computing power acquisition threshold parameter and the supply ladder distribution sequence can enter the bidirectional data constraint iteration operation under the same computing power unit caliber. This avoids directly mixing the sequence node values ​​under the integer precision standard computing power unit caliber with the computing power acquisition threshold parameter under the floating point precision standard computing power unit caliber, and allows the target computing power unit reference identifier to continue to participate in the caliber limitation of the subsequent first constraint condition, second constraint condition, first node response parameter and second node request parameter.

[0118] Preferably, the first constraint is not a simple numerical comparison threshold, but a request-side constraint record formed by the computing power acquisition threshold parameter, the target computing power unit benchmark identifier, and the task requirement field in the task processing request message. The task requirement field expresses the computing power requirement, task access timing requirements, and computing power unit benchmark requirements corresponding to the current task processing request message of the computing power request node. The request-side constraint record uses the target computing power unit benchmark identifier as a caliber marker, the computing power acquisition threshold parameter as a filtering boundary, and the task requirement field as the source of demand, thus forming a first constraint that can act on the supply tiered distribution sequence. By organizing the computing power acquisition threshold parameter into the request-side constraint record, the first constraint can directly limit which target vehicle nodes can enter the available node subset in subsequent single data loop processing processes, rather than making isolated judgments on the values ​​of individual sequence nodes. Simultaneously, the request-side constraint record continues to provide the task requirement field and the computing power acquisition threshold parameter when subsequently calculating the second node request parameters, ensuring consistency of the request-side data source between the first constraint and the second node request parameters.

[0119] Preferably, the second constraint is formed by the value of the current sequence node in the supply ladder distribution sequence, and the value of the current sequence node originates from the supply-side sequence node currently traversed in the supply ladder distribution sequence. The supply-side sequence node corresponds to the target vehicle node that has completed the standardization transformation, and the supply-side sequence node includes the target vehicle node identifier, the benchmark matching standard computing unit value that matches the target computing unit benchmark identifier, and the computing power supply expectation parameter corresponding to the benchmark matching standard computing unit value; the value of the current sequence node is extracted from the supply-side sequence node and used as the second constraint in this single data loop processing process. By using the value of the current sequence node as the second constraint, the bidirectional data constraint iteration operation can progressively expand or adjust the supply range that can participate in the calculation according to the ascending order of the supply ladder distribution sequence, so that each single data loop processing process corresponds to a supply-side constraint state, and the supply-side constraint state continues to participate in the calculation of the subsequent first node response parameters and second node request parameters.

[0120] Preferably, in a single data loop processing step, when applying the first constraint to the supply ladder distribution sequence, firstly, supply-side sequence nodes that do not match the target computing power unit benchmark identifier in the request-side constraint record are filtered out. Then, the remaining supply-side sequence nodes are boundary-filtered according to the computing power acquisition threshold parameter in the request-side constraint record to obtain a subset of available nodes that satisfy the first constraint. Each target vehicle node in the subset of available nodes retains a corresponding target vehicle node identifier, a benchmark matching standard computing power unit value, and a computing power supply expectation parameter. The target vehicle node identifier is used to trace the node source when calculating the first node response parameter in the subsequent calculation. The benchmark matching standard computing power unit value is used to provide a computing power quantity basis under the same caliber when calculating the first node response parameter in the subsequent calculation. The computing power supply expectation parameter is used to express the ladder position relationship between the current sequence node value and the subset of available nodes when the second constraint is applied in the subsequent calculation.

[0121] Preferably, in the specific technical implementation of calculating the first node response parameter based on the available node subset, the baseline matching standard computing power unit value corresponding to each target vehicle node in the available node subset is first read, then the network connectivity status, vehicle basic task occupancy status, and current availability status corresponding to each target vehicle node in the available node subset are read, and then the baseline matching standard computing power unit value is effectively limited according to the network connectivity status, the vehicle basic task occupancy status, and the current availability status to obtain the node effective standard computing power unit value. The node effective standard computing power unit value is further summarized according to the arrangement order in the supply ladder distribution sequence to form the first node response parameter. The first node response parameter is used to express the effective supply response under the same computing power unit caliber that the current available node subset can provide under the first constraint condition, rather than directly adding the original computing power values ​​of the target vehicle nodes; the first node response parameter continues to serve as the basis for supply-side calculation when the second constraint condition takes effect, in order to participate in the calculation of the second node request parameter.

[0122] Preferably, when applying the second constraint to the first node response parameter, the current sequence node value is first associated with the available node subset corresponding to the first node response parameter to determine the current tier position of the current sequence node value in the supply tier distribution sequence. Then, based on the current tier position, the effective standard computing power unit value of the node not higher than the current tier position is extracted from the first node response parameter to form the current tier response parameter. The current tier response parameter is further adapted and calculated with the task requirement field in the task processing request message to obtain the second node request parameter. The second node request parameter is used to express the request-side calculation result that the computing power request node can form a corresponding relationship with the first node response parameter under the second constraint represented by the current sequence node value, so that the second node request parameter and the first node response parameter are both under the same target computing power unit benchmark identifier, and the current tier response parameter becomes the intermediate calculation result connecting the first node response parameter and the second node request parameter.

[0123] Preferably, the calculation process of the second node request parameter does not directly replace the task requirement with the computing power acquisition threshold parameter. Instead, it uses the computing power acquisition threshold parameter as the request-side constraint boundary, the task requirement field as the request-side quantity source, and the current tiered response parameter as the supply-side quantity source for collaborative calculation. Specifically, the required computing power unit quantity is first determined based on the task requirement field, then the effective supply computing power unit quantity is determined based on the current tiered response parameter, and the required computing power unit quantity is compared with the effective supply computing power unit quantity. Subsequently, based on the comparison result, the number of requested computing power units that can be handled by the available node subset in this single data loop processing process is determined, and this number of requested computing power units is used as the second node request parameter. Through this processing method, the second node request parameter is limited by both the computing power acquisition threshold parameter and the current tiered position corresponding to the current sequence node value, thus enabling the second node request parameter to reflect the data state after the combined effect of request-side and supply-side constraints.

[0124] Preferably, when comparing the first node response parameter and the second node request parameter, a caliber verification is first performed on the first node response parameter and the second node request parameter. This caliber verification checks whether the target computing power unit benchmark identifier corresponding to the first node response parameter is consistent with the target computing power unit benchmark identifier corresponding to the second node request parameter. After the caliber verification passes, the difference between the number of effective supply computing power units in the first node response parameter and the number of request computing power units in the second node request parameter is calculated to obtain the deviation value. The deviation value expresses the degree of difference between the first node response parameter after filtering by the first constraint condition on the supply side and the second node request parameter after being constrained by the second constraint condition on the request side during this single data loop processing process. The deviation value is not the result of directly comparing parameters with different dimensions, but rather iterative judgment data calculated based on the same target computing power unit benchmark identifier after the caliber verification passes.

[0125] Preferably, when using the deviation value from the comparison output as the iteration index for the current data loop processing process, the deviation value, the current sequence node value, the current ladder position, the available node subset, the target computing unit benchmark identifier, the first node response parameter, and the second node request parameter are jointly written into the iteration state record corresponding to the current single data loop processing process. The iteration state record describes the technical correlation between the first constraint, the second constraint, the first node response parameter, the second node request parameter, and the deviation value in the current single data loop processing process. After reading the deviation value from the iteration state record, the iteration index continues to serve as the basis for determining whether the deviation value falls within the preset convergence interval. By setting the iteration state record, subsequent single data loop processing processes can inherit the data state corresponding to the current sequence node value and the deviation value from the previous single data loop processing process, instead of re-traversing the supply ladder distribution sequence without any correlation.

[0126] Preferably, the preset convergence interval is configured before the bidirectional data constraint iteration operation starts, based on the target computing power unit benchmark identifier, task type, and adaptive refresh cycle. The preset convergence interval is used to limit the acceptable deviation range of the deviation value. For quantization inference tasks, the preset convergence interval can be configured according to the integer precision standard computing power unit caliber; for half-precision floating-point calculation tasks, the preset convergence interval can be configured according to the floating-point precision standard computing power unit caliber; for scenarios with short adaptive refresh cycles, the preset convergence interval can adopt a smaller allowable fluctuation range to reduce numerical jitter during rapid broadcasting. The preset convergence interval and the target computing power unit benchmark identifier jointly participate in the determination of the iteration index, ensuring that deviation values ​​under different precision calibers are not mixed into the same convergence judgment rule; the preset convergence interval, together with the iteration status record, is used to determine whether the current single data loop processing process can form continuous iterative output results.

[0127] Preferably, when the deviation value falls within the preset convergence interval, the bidirectional data constraint iteration operation uses the current sequence node value, current step position, first node response parameter, and second node request parameter in this single data loop processing process as the convergence data source for the continuous iteration output result. The continuous iteration output result retains the current step position corresponding to the current sequence node value, which indicates which level of supply-side sequence node in the supply ladder distribution sequence the convergence value comes from; the continuous iteration output result also retains the target computing power unit benchmark identifier, which indicates the computing power unit caliber corresponding to the convergence value. Thus, when the convergence value of the final iteration output is subsequently extracted as the benchmark price value, the benchmark price value can be consistent with the target computing power unit benchmark identifier used in this bidirectional data constraint iteration operation, and can be traced back to the corresponding supply-side sequence node in the supply ladder distribution sequence through the current step position.

[0128] Preferably, under the condition that the deviation value does not fall within the preset convergence interval, the next data cycle processing process is triggered based on the value of the next sequence node in the supply ladder distribution sequence. The value of the next sequence node comes from the supply-side sequence node adjacent to the value of the current sequence node, and the value of the next sequence node is set as the second constraint condition in the next single data cycle processing process; at the same time, the available node subset, the first node response parameter, the second node request parameter, and the deviation value in the iteration state record corresponding to the previous single data cycle processing process are used as reference data for the next single data cycle processing process. The next single data cycle processing process reapplies the first node response parameter based on the value of the next sequence node to update the second node request parameter, and continues to compare the updated first node response parameter with the updated second node request parameter, so that the supply ladder distribution sequence can approach the preset convergence interval step by step according to the ascending order.

[0129] Preferably, when performing the bidirectional data constraint iteration operation within the parking area, the supply tiered distribution sequence is formed by ascending order of the expected computing power supply parameters of multiple target vehicle nodes, and the computing power acquisition threshold parameter originates from the task processing request message sent by the computing power request node. For the quantization inference task, the target computing power unit benchmark identifier points to the integer precision standard computing power unit benchmark identifier. The first constraint condition first filters out target vehicle nodes that meet the computing power acquisition threshold parameter and have integer precision standard computing power unit values. The available node subset is then used to calculate the first node response parameter under the integer precision standard computing power unit caliber. Subsequently, the current sequence node value acts as a second constraint condition on the first node response parameter to calculate the second node request parameter under the integer precision standard computing power unit caliber. As the next sequence node values ​​are selected sequentially, the first node response parameter and the second node request parameter gradually approach each other until the deviation value falls into the preset convergence interval. The continuous iterative output results formed after the deviation value falls into the preset convergence interval continue to serve as the convergence data source for the benchmark price value.

[0130] Preferably, the bidirectional data constraint iterative operation differs from the processing method that only reads a single value from a fixed reference table. Its technical processing simultaneously uses both the first and second constraints, creating a mutual constraint relationship between them in each single data loop processing step. The first constraint prevents supply-side sequence nodes in the supply tiered distribution sequence that do not meet the computing power acquisition threshold parameter from entering the available node subset. The second constraint prevents the first node response parameter from deviating from the current tiered position represented by the current sequence node value and directly forming the second node request parameter. Through these mutual constraints, the deviation value can reflect the dynamic matching state between the computing power request node and the target vehicle node under the same computing power unit caliber. The subsequently generated benchmark price value can also originate from the convergence result of continuous data loop processing, rather than from a single static read result.

[0131] Optionally, before the step of broadcasting the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh period, the method further includes:

[0132] Collect the total number of available vehicle nodes in the preset parking physical area that are in a network connected state, and count the total number of pending tasks to which the task processing request message belongs;

[0133] When the total number of available vehicle nodes is greater than the first preset node number threshold, or when the total number of pending tasks in the queue is greater than the first preset queue number threshold, the adaptive refresh cycle is set to the first cycle duration parameter.

[0134] When the total number of available vehicle nodes is lower than the second preset node number threshold and the total number of pending tasks in the queue is lower than the second preset queue number threshold, the adaptive refresh cycle is set to the second cycle duration parameter.

[0135] The time interval for broadcasting the benchmark price value to the preset node cluster via the network interface is controlled according to the first cycle duration parameter or the second cycle duration parameter.

[0136] Preferably, the specific implementation process for collecting the total number of available vehicle nodes in a network-connected state within the preset parking physical area is as follows: After the benchmark price value is extracted and before it is broadcast through the network interface, a node connectivity sampling window is generated based on the preset parking physical area. The node connectivity sampling window is used to limit the time and area range used for this statistical analysis of the total number of available vehicle nodes. Within the node connectivity sampling window, the network connectivity status, available computing power reporting timestamp, and vehicle basic task occupancy status of each target vehicle node are read. Target vehicle nodes whose network connectivity status is connected, whose available computing power reporting timestamp falls within the node connectivity sampling window, and whose vehicle basic task occupancy status corresponds to an occupancy level lower than the available resources of the vehicle processor are written into the available vehicle node statistical record. The available vehicle node statistical record retains the corresponding target vehicle node identifier and target computing power unit benchmark identifier, ensuring that the available vehicle node statistical record maintains the same node source relationship with the aforementioned dual-benchmark standard computing power unit sequence. Subsequently, the available vehicle node statistical record is deduplicated and counted based on the target vehicle node identifier to obtain the total number of available vehicle nodes. The total number of available vehicle nodes continues to participate in the construction of subsequent activity determination data groups, so that the statistical record of available vehicle nodes is not an isolated statistical result, but serves as the node-side data source set by the adaptive refresh cycle.

[0137] Preferably, the total number of pending task queues is not directly calculated using the historical cumulative request count. Instead, within the task queue statistics window to which the task processing request message belongs, status filtering is performed on task processing request messages that are still in the waiting access, waiting allocation, or waiting broadcast confirmation state to obtain pending task queue statistics records. The task queue statistics window and the node connectivity sampling window use the same time base. The pending task queue statistics records include a task processing request message identifier, a target computing power unit benchmark identifier, a computing power acquisition threshold parameter, and a queue status identifier. The target computing power unit benchmark identifier is used to distinguish between pending task queue statistics records under the integer precision standard computing power unit caliber and those under the floating-point precision standard computing power unit caliber. The queue status identifier is used to exclude task processing request messages that have completed access or have expired. By counting the pending task queue statistics records, the total number of pending task queues is obtained, enabling the total number of pending task queues to reflect the scale of request-side data that still needs to be scheduled based on the benchmark price value before the current broadcast. The total number of pending tasks continues to participate in the construction of subsequent activity determination data groups, so that there is a direct data transmission relationship between the pending task queue statistics and the adaptive refresh cycle setting.

[0138] Preferably, the total number of available vehicle nodes and the total number of pending task queues form an activity determination data group. This activity determination data group is then compared with the first preset node number threshold, the first preset queue number threshold, the second preset node number threshold, and the second preset queue number threshold to obtain a refresh cycle determination result. The first preset node number threshold characterizes the node-side high activity boundary when the total number of available vehicle nodes in the preset parking physical area is large; the first preset queue number threshold characterizes the request-side high activity boundary when the total number of pending task queues is large; the second preset node number threshold characterizes the node-side low activity boundary when the total number of available vehicle nodes is small; and the second preset queue number threshold characterizes the request-side low activity boundary when the total number of pending task queues is small. The first preset node number threshold, the first preset queue number threshold, the second preset node threshold, and the second preset queue number threshold are pre-configured in an initialization configuration table. The initialization configuration table stores these thresholds according to the parking scenario type, the node connectivity sampling window length, and the task queue statistics window length, thereby ensuring that the refresh cycle determination result corresponds to the network connectivity status change characteristics and queue status change characteristics in different parking scenarios such as parking areas, residential parking areas, or charging parking areas. The refresh cycle determination result is then used to generate high-activity state records, low-activity state records, or intermediate-activity state records.

[0139] Preferably, when the total number of available vehicle nodes is greater than a first preset node number threshold, or when the total number of pending task queues is greater than a first preset queue number threshold, a high-activity status record is generated based on the refresh cycle determination result, and the adaptive refresh cycle is set as a first cycle duration parameter based on the high-activity status record. The high-activity status record includes the total number of available vehicle nodes, the total number of pending task queues, a trigger field that triggers the first cycle duration parameter, and a target computing unit benchmark identifier corresponding to the benchmark price value; wherein, the trigger field is used to record whether the current trigger is due to the total number of available vehicle nodes, the total number of pending task queues, or both. The first cycle duration parameter is used to form a shorter broadcast time interval (for example, if the total number of available vehicle nodes exceeds the first preset node number threshold or the total number of pending task queues exceeds the first preset queue number threshold, it can be configured as a second-level broadcast time interval), so that the benchmark price value can be written into the broadcast scheduling record with a shorter broadcast time interval when network connectivity and queue status change rapidly. The broadcast scheduling record continues to read the first cycle duration parameter to control the subsequent broadcast time interval of the network interface.

[0140] Preferably, when the total number of available vehicle nodes is lower than a second preset node threshold and the total number of pending task queues is lower than a second preset queue threshold, a low-activity status record is generated based on the refresh cycle determination result, and the adaptive refresh cycle is set as the second cycle duration parameter based on the low-activity status record. The low-activity status record includes the total number of available vehicle nodes, the total number of pending task queues, the second preset node threshold, the second preset queue threshold, and the target computing unit benchmark identifier corresponding to the benchmark price value; wherein, the total number of available vehicle nodes and the total number of pending task queues must simultaneously meet the low-activity determination condition before the adaptive refresh cycle is set as the second cycle duration parameter. The second cycle duration parameter is used to form a longer broadcast time interval (for example, if the total number of available vehicle nodes is lower than the second preset node threshold and the total number of pending task queues is lower than the second preset queue threshold, it can be configured as a minute-level broadcast time interval), so that the benchmark price value reduces the number of invalid broadcasts when the network connectivity and queue status changes slowly, and enables the network interface to use network transmission resources for subsequent task processing request messages and target scheduling instructions. The broadcast scheduling record continues to read the second period duration parameter to control the subsequent broadcast time interval of the network interface.

[0141] Preferably, when the total number of available vehicle nodes is not greater than the first preset node number threshold and the total number of pending task queues is not greater than the first preset queue number threshold, and simultaneously the total number of available vehicle nodes is not lower than the second preset node number threshold or the total number of pending task queues is not lower than the second preset queue number threshold, an intermediate active state record is generated based on the refresh cycle determination result, and the adaptive refresh cycle is transitionally set based on the intermediate active state record. The intermediate active state record continues to read the adaptive refresh cycle of the previous cycle, the total number of available vehicle nodes of the previous cycle, and the total number of pending task queues of the previous cycle, and compares the difference between the total number of available vehicle nodes of the previous cycle and the total number of pending task queues of the previous cycle and the total number of pending task queues of the current cycle to generate an activity change record. The activity change record is used to determine whether the network connectivity status and queue status within the preset parking physical area are changing towards high activity or low activity, and then selects a transition cycle duration parameter between the first cycle duration parameter and the second cycle duration parameter. The transition period duration parameter continues to be written into the broadcast scheduling record, so that the adaptive refresh period does not frequently change due to single count fluctuations.

[0142] Preferably, when the total number of available vehicle nodes is zero or the total number of pending task queues is zero, a broadcast hold-up record is generated. This record indicates that the current period will no longer trigger the broadcast time interval corresponding to the first period duration parameter based on the new bidirectional data constraint iteration operation result. Instead, it reads the benchmark price value already broadcast in the previous period and the timestamp information corresponding to the previous period to form a last valid benchmark price record. This last valid benchmark price record includes the benchmark price value already broadcast in the previous period, the target computing unit benchmark identifier of the previous period, and the adaptive refresh period of the previous period. The last valid benchmark price record is continued to be written into the broadcast scheduling record, so that when the total number of available vehicle nodes and the total number of pending task queues recover to non-zero, the last valid benchmark price record can be used as broadcast recovery reference data. Through the cooperation of the broadcast hold-up record and the last valid benchmark price record, the network interface will not repeatedly broadcast benchmark price values ​​lacking a data basis when there are no available vehicle nodes or no pending task queues. The last valid benchmark price record can also serve as a time and caliber reference for subsequent broadcast recovery.

[0143] Preferably, when controlling the time interval for broadcasting the benchmark price value to the preset node cluster via the network interface according to the first cycle duration parameter or the second cycle duration parameter, the benchmark price value, the target computing power unit benchmark identifier, the adaptive refresh cycle, and the refresh cycle determination result of this cycle are first written into the broadcast scheduling record. The broadcast scheduling record is used to carry the data content and time control content to be broadcast this week. The benchmark price value is used to provide the main data for broadcasting this cycle. The target computing power unit benchmark identifier is used to indicate the integer precision standard computing power unit caliber or floating-point precision standard computing power unit caliber corresponding to the benchmark price value. The adaptive refresh cycle is used to control the broadcast time interval between two adjacent broadcasts. The refresh cycle determination result is used to indicate that the adaptive refresh cycle originates from a high-activity state record, a low-activity state record, or an intermediate-activity state record. After the broadcast scheduling record is generated, a timing trigger is initiated according to the first cycle duration parameter, the second cycle duration parameter, or the transition cycle duration parameter. Only after the timing trigger is reached is the benchmark price value sent to the preset node cluster via the network interface.

[0144] Preferably, before sending the benchmark price value, the network interface performs a field integrity check on the broadcast scheduling record. This field integrity check includes checking if the benchmark price value exists, whether the target computing unit benchmark identifier corresponds to the benchmark price value, whether the adaptive refresh period has been determined by the first period duration parameter, the second period duration parameter, or the transition period duration parameter, and whether the node address record of the preset node cluster is in a transmittable state. The node address record is used to store the network address and most recent reception status of each node in the preset node cluster. After the field integrity check, the node address record continues to participate in broadcast target selection, ensuring that the benchmark price value is only sent to nodes in the preset node cluster that are in a transmittable state. Through the field integrity check, a continuous data association is formed between the benchmark price value, the adaptive refresh period, and the preset node cluster, avoiding technical breakpoints where the benchmark price value has been calculated but broadcasting has not been performed according to the corresponding adaptive refresh period. The result of the field integrity check is further written into the broadcast scheduling record so that the broadcast scheduling record can record whether the conditions for transmission through the network interface are met in the current period.

[0145] Preferably, when the above adaptive refresh cycle setting is executed within the parking area, the total number of available vehicle nodes reflects the number of target vehicle nodes in the parking area that are in a network-connected state and can still provide available computing power, and the total number of pending task queues reflects the number of task processing request messages currently waiting to be accessed. If vehicles in the parking area enter a parking state in a concentrated manner, causing the total number of available vehicle nodes to exceed the first preset node number threshold, or if multiple computing power request nodes send task processing request messages within a similar time period, causing the total number of pending task queues to exceed the first preset queue number threshold, then the high-activity state record triggers the first cycle duration parameter; if many vehicles leave the parking area and there are few task processing request messages, causing the total number of available vehicle nodes to be lower than the second preset node number threshold and the total number of pending task queues to be lower than the second preset queue number threshold, then the low-activity state record triggers the second cycle duration parameter. Thus, the broadcast time interval of the benchmark price value can be adjusted according to changes in the network connectivity and queue status within the preset parking physical area, avoiding a large delay between the benchmark price value and the current node status caused by using a fixed time interval. The broadcast time interval continues to act on the broadcast scheduling record, causing the network interface to perform broadcasts according to the broadcast scheduling record.

[0146] Preferably, the adaptive refresh cycle is also coordinated with the iteration state record of the bidirectional data constraint iteration operation. The iteration state record records the current sequence node value, the first node response parameter, the second node request parameter, and the deviation value corresponding to the benchmark price value. When writing the benchmark price value, the broadcast scheduling record simultaneously reads the deviation value and the target computing power unit benchmark identifier from the iteration state record, and writes the deviation value and the target computing power unit benchmark identifier as broadcast auxiliary verification fields into the broadcast scheduling record. The broadcast auxiliary verification field is used to verify whether the benchmark price value of this cycle comes from the converged continuous iteration output result when the preset node cluster receives the benchmark price value. When the broadcast auxiliary verification field matches the benchmark price value, the target computing power unit benchmark identifier, and the adaptive refresh cycle in the broadcast scheduling record, the preset node cluster receives the benchmark price value and continues to send the benchmark price value to the computing power request node to instruct the computing power request node to generate a target scheduling instruction. The target scheduling instruction continues to use the target computing unit benchmark identifier and the benchmark price value, so that the bidirectional data constraint iteration operation, the adaptive refresh cycle, the network interface broadcast, and the target scheduling instruction maintain the same data link.

[0147] Optionally, before the step of obtaining the hardware specification characteristics of the on-board processors of the target vehicle nodes distributed within the preset parking physical area, the method further includes:

[0148] Obtain the spatial location attribute information of the preset parking physical area, match the corresponding computing power supply stability coefficient based on the spatial location attribute information, and obtain the credibility weight coefficient corresponding to the preset parking physical area.

[0149] The computing power supply stability coefficient is multiplied by the credibility weight coefficient to obtain the node collaboration comprehensive weight value of the preset stationed physical area;

[0150] After obtaining the benchmark price value, the node collaborative comprehensive weight value is multiplied by the benchmark price value to obtain the regional weighted price adjustment parameter;

[0151] The regional weighted price adjustment parameter is used as an alternative broadcast result after regional weighting of the benchmark price value, and the regional weighted price adjustment parameter is broadcast to the preset node cluster through the network interface.

[0152] Preferably, before the step of obtaining the hardware specifications of the on-board processors of target vehicle nodes distributed within the preset parking physical area, the preset parking physical area is first subjected to regional attribute reading to obtain the spatial location attribute information. The regional attribute reading uses the geographic coordinate point data, regional boundary data, and vehicle dwelling records of the preset parking physical area as data sources. The geographic coordinate point data is used to define the location of the preset parking physical area on the electronic map, the regional boundary data is used to define whether subsequent target vehicle nodes belong to the preset parking physical area, and the vehicle dwelling records are used to express the time distribution of vehicle entry, parking, and departure within the preset parking physical area. Location attribution analysis is performed on the geographic coordinate point data and the regional boundary data to generate location attribution analysis results. Based on the location attribution analysis results, location classification labels are generated and written into the spatial location attribute information. When the location classification labels need to be supplemented by the vehicle dwelling records, a vehicle dwelling duration distribution record is formed based on the vehicle dwelling records, and the spatial location attribute information is supplemented based on the vehicle dwelling duration distribution record. Therefore, the spatial location attribute information is not an isolated location description, but rather carries the location classification label, the location attribution resolution result, and the vehicle dwell time distribution record to continue participating in the subsequent matching process of computing power supply stability coefficient and credibility weight coefficient.

[0153] Preferably, after the spatial location attribute information is generated, a preset scene weight configuration table is read. This table configures different computing power supply stability coefficients and reliability weight coefficients for parking areas, residential parking areas, charging parking areas, commercial parking areas, roadside parking areas, and park parking areas. The computing power supply stability coefficient in the scene weight configuration table characterizes the stability of the target vehicle node's network connectivity and continuous reporting of available computing power under the corresponding parking scenario. The reliability weight coefficient characterizes the data reliability between the vehicle's parking record and the subsequent broadcast of the benchmark price value under the corresponding parking scenario. After matching the location classification label in the spatial location attribute information with the scene weight configuration table, the scene weight configuration entry corresponding to the location classification label is located. The computing power supply stability coefficient and reliability weight coefficient corresponding to the preset parking physical area are extracted from the scene weight configuration entry, ensuring that the subsequent node collaborative comprehensive weight value originates from the actual parking scenario of the preset parking physical area, rather than from a fixed, uniform weight.

[0154] Preferably, the matching process for the computing power supply stability coefficient is as follows: First, the scene weight configuration entry in the scene weight configuration table is located based on the location classification label in the spatial location attribute information. Then, the vehicle dwell time reference field, network connectivity continuity field, and available computing power reporting continuity field in the scene weight configuration entry are read. The vehicle dwell time reference field expresses the common dwell time range of the target vehicle node in the corresponding dwelling scene; the network connectivity continuity field expresses the continuity of the target vehicle node maintaining network connectivity within the preset dwelling physical area; and the available computing power reporting continuity field expresses the continuity of the target vehicle node continuously reporting available computing power during dwelling. The vehicle dwell time reference field, the network connectivity continuity field, and the available computing power reporting continuity field are jointly mapped to the computing power supply stability coefficient. These three fields are retained as the matching source for the computing power supply stability coefficient, ensuring that the computing power supply stability coefficient reflects the stable data foundation for the target vehicle node to participate in subsequent benchmark price calculations in the corresponding dwelling scene.

[0155] Preferably, the matching process of the credibility weight coefficient is as follows: after the spatial location attribute information matches the scene weight configuration entry, the scene boundary identification field, the complete vehicle dwelling record field, and the historical node online fluctuation field in the scene weight configuration entry are read. The scene boundary identification field expresses whether the regional boundary data of the preset parking physical area facilitates stable identification of the target vehicle node; the complete vehicle dwelling record field expresses whether the vehicle dwelling record within the preset parking physical area can continuously cover the state changes of vehicle entry and exit; and the historical node online fluctuation field expresses the degree of change in the network connectivity status of the target vehicle node within the preset parking physical area over a historical period. The credibility weight coefficient is generated based on the scene boundary identification field, the complete vehicle dwelling record field, and the historical node online fluctuation field. These three fields are retained as the matching source for the credibility weight coefficient, allowing it to continue participating in the calculation of the node collaborative comprehensive weight value and forming a parking scene reliability correction result for subsequent regional weighted price adjustment parameters.

[0156] Preferably, the specific implementation process of multiplying the computing power supply stability coefficient by the credibility weight coefficient to obtain the node collaboration comprehensive weight value of the preset parking physical area is as follows: First, the computing power supply stability coefficient and the credibility weight coefficient are checked within the same interval. The same interval check is used to confirm that both the computing power supply stability coefficient and the credibility weight coefficient belong to normalized weight data that can be used for multiplication processing. The normalized weight data is used to ensure that the computing power supply stability coefficient and the credibility weight coefficient participate in multiplication processing within the same numerical interval. After the same interval check is passed, the computing power supply stability coefficient and the credibility weight coefficient are multiplied to obtain the node collaboration comprehensive weight value. The node collaboration comprehensive weight value retains the location classification label corresponding to the spatial location attribute information, the matching source of the computing power supply stability coefficient, and the matching source of the credibility weight coefficient, so that the node collaboration comprehensive weight value can continue to express the parking scene stability and data credibility of the preset parking physical area when it is subsequently multiplied with the benchmark price value.

[0157] Preferably, after the node collaborative comprehensive weight value is generated, it is further correlated and verified with the distribution status of target vehicle nodes within the preset parking physical area. The distribution status of target vehicle nodes is derived from the total number of available vehicle nodes, the target vehicle node identifier, and the regional affiliation result of the target vehicle node within the preset parking physical area; wherein, the regional affiliation result is generated based on the regional boundary data and the location reporting data corresponding to the target vehicle node identifier, and is used to confirm whether the target vehicle node is located within the preset parking physical area described by the spatial location attribute information. After correlating and verifying the node collaborative comprehensive weight value with the distribution status of target vehicle nodes, a node collaborative weight verification record is generated. The node collaborative weight verification record is used to indicate that the node collaborative comprehensive weight value corresponds to the target vehicle node participating in the benchmark price value calculation in this period, rather than corresponding to the weight data of other parking areas. The node collaborative weight verification record continues to participate in the generation of subsequent regional weighted price adjustment parameters, so that the node collaborative comprehensive weight value and the benchmark price value have the same regional source relationship.

[0158] Preferably, after obtaining the benchmark price value, the node collaborative comprehensive weight value and the node collaborative weight verification record are read, and the node collaborative comprehensive weight value and the benchmark price value are confirmed to correspond to the same preset parking physical region based on the node collaborative weight verification record. After confirming the same region source relationship, the node collaborative comprehensive weight value and the benchmark price value are multiplied to obtain the regional weighted price adjustment parameter. The regional weighted price adjustment parameter retains the benchmark price value, the node collaborative comprehensive weight value, the spatial location attribute information, and the target computing power unit benchmark identifier, so that the regional weighted price adjustment parameter can express both the benchmark price value obtained by the bidirectional data constraint iterative operation and the parking scene weight correction result corresponding to the preset parking physical region. The parking scene weight correction result continues to serve as the field source that distinguishes the regional weighted price adjustment parameter from the benchmark price value.

[0159] Preferably, the regional weighted price adjustment parameter and the benchmark price value are not two unrelated broadcast data, but rather have a substitute broadcast relationship. This substitute broadcast relationship is recorded by a regional weighted broadcast identifier, which indicates that the data broadcast through the network interface in the current period is the regional weighted price adjustment parameter, not the benchmark price value without correction by the stationary scenario weight correction result. When generating the regional weighted broadcast identifier, the node collaborative comprehensive weight value, spatial location attribute information, stationary scenario weight correction result, and target computing power unit benchmark identifier in the regional weighted price adjustment parameter are read, and these elements are jointly written into the regional weighted broadcast record. The regional weighted broadcast record continues to serve as the field source before network interface broadcasting, enabling the preset node cluster to recognize that the currently received data has undergone regional weighting processing.

[0160] Preferably, before broadcasting the regional weighted price adjustment parameter to the preset node cluster via the network interface, a field consistency check is performed on the regional weighted broadcast record. The field consistency check includes checking whether the regional weighted price adjustment parameter corresponds to the benchmark price value, whether the regional weighted price adjustment parameter corresponds to the node collaborative comprehensive weight value, whether the node collaborative comprehensive weight value corresponds to the spatial location attribute information, whether the regional weighted price adjustment parameter corresponds to the stationary scenario weight correction result, and whether the target computing power unit benchmark identifier is consistent with the benchmark price value. After the field consistency check passes, the regional weighted price adjustment parameter, the regional weighted broadcast identifier, the spatial location attribute information, the stationary scenario weight correction result, and the target computing power unit benchmark identifier are jointly written into the broadcast scheduling record, so that the broadcast scheduling record can use the regional weighted price adjustment parameter instead of the benchmark price value as the data subject for broadcast in this period.

[0161] Preferably, when the network interface sends the regional weighted price adjustment parameter to the preset node cluster according to the broadcast scheduling record, the preset node cluster receives the regional weighted price adjustment parameter, the regional weighted broadcast identifier, and the target computing power unit benchmark identifier, and identifies the regional weighted price adjustment parameter as a replacement broadcast result after regional weighting of the benchmark price value according to the regional weighted broadcast identifier. The preset node cluster continues to send the regional weighted price adjustment parameter to the computing power request node, and the computing power request node generates a target scheduling instruction based on the regional weighted price adjustment parameter and the target computing power unit benchmark identifier, so that the target scheduling instruction can simultaneously use the convergence data source of the benchmark price value and the dwell scenario weight correction result corresponding to the node collaborative comprehensive weight value. The target scheduling instruction continues to retain the regional weighted broadcast identifier, so that the computing power request node can distinguish that the current one is using the regional weighted price adjustment parameter, rather than the benchmark price value that has not undergone regional weighting.

[0162] Preferably, when different parking areas, residential parking areas, charging parking areas, commercial parking areas, roadside parking areas, and park parking areas participate in the calculation of the real-time price index for vehicle computing power, different preset parking physical areas can be matched with different computing power supply stability coefficients and reliability weight coefficients through different spatial location attribute information. Parking areas and residential parking areas typically correspond to longer vehicle dwell times and more stable network connectivity, therefore, the scenario weight configuration table configures relatively higher computing power supply stability coefficients and reliability weight coefficients for parking areas and residential parking areas; roadside parking areas typically correspond to shorter vehicle dwell times and higher network connectivity fluctuations, therefore, the scenario weight configuration table configures relatively lower computing power supply stability coefficients and reliability weight coefficients for roadside parking areas. After different preset parking physical areas undergo the same node collaboration comprehensive weight value calculation and regional weighted price adjustment parameter generation process, they can obtain regional weighted price adjustment parameters that match their respective spatial location attribute information, enabling subsequent alternative broadcast results to reflect the differences in node collaboration between different parking scenarios. The alternative broadcast results continue to maintain an alternative broadcast relationship with the benchmark price value through the regional weighted broadcast identifier.

[0163] Preferably, the regional weighted price adjustment parameter also maintains the same field caliber as the historical sequence database records in the historical sequence database. Specifically, after the regional weighted price adjustment parameter is broadcast, the regional weighted price adjustment parameter, the benchmark price value, the node collaborative comprehensive weight value, the spatial location attribute information, the docking scenario weight correction result, and the target computing power unit benchmark identifier are written into the historical sequence database to form a historical sequence database record. When the historical sequence database is subsequently used to extract historical benchmark price value sequences or perform moving weighted average processing operations, it can distinguish the historical sequence database records corresponding to different preset docking physical regions based on the spatial location attribute information, and distinguish between integer precision standard computing power unit caliber and floating-point precision standard computing power unit caliber based on the target computing power unit benchmark identifier. Therefore, the regional weighted price adjustment parameter not only participates in the network interface broadcast of the current period but also continues to participate in the generation of subsequent prediction interval values, maintaining data continuity between regional weighted processing and subsequent task access timing planning.

[0164] Optionally, the step of obtaining the spatial location attribute information of the preset parking physical area includes:

[0165] Collect the geographic coordinate point dataset of the preset parking physical area, input the geographic coordinate point dataset into the preset point of interest location parsing program to output the location classification label, and use the location classification label as the spatial location attribute information;

[0166] When the location parsing program for points of interest does not output the missing location classification label, the statistical data on vehicle dwell time distribution of the preset parking physical area within a historical set period is retrieved. Based on the statistical data on vehicle dwell time distribution, an alternative location classification label is determined, and the alternative location classification label is used to replace the original location classification label. The alternative location classification label is then used as the spatial location attribute information.

[0167] Preferably, the specific implementation process of collecting the geographic coordinate point dataset of the preset parking physical area is as follows: Before obtaining the spatial location attribute information, the regional boundary data, vehicle parking location reporting data, and map coordinate reference information corresponding to the preset parking physical area are read first; wherein, the regional boundary data is used to limit the spatial range of the preset parking physical area, the vehicle parking location reporting data is used to provide the location points reported by multiple target vehicle nodes when parking within the preset parking physical area, and the map coordinate reference information is used to unify the coordinate reference of the regional boundary data and the vehicle parking location reporting data. After unifying the coordinate reference of the regional boundary data and the vehicle parking location reporting data based on the map coordinate reference information, the location points in the vehicle parking location reporting data are mapped to a coordinate reference consistent with the regional boundary data, and the location points falling within the regional boundary data are selected from the mapped location points to form the geographic coordinate point dataset. The geographic coordinate point dataset continues to serve as the input data for the point of interest location parsing program, ensuring that the generation process of the location classification label maintains a correspondence with the actual spatial range of the preset parking physical area.

[0168] Preferably, after the geographic coordinate point dataset is formed, point stability processing is performed on the geographic coordinate point dataset to reduce the impact of the location drift of a single target vehicle node on the location classification label. Specifically, the geographic coordinate point dataset is first grouped according to the reporting time and target vehicle node identifier of each location point in the geographic coordinate point dataset to obtain node point group records; then, based on the distance change and reporting time interval between adjacent location points under the same target vehicle node identifier, abnormal drift points are identified from the node point group records and removed from the node point group records to obtain stable point group records; subsequently, the regional center point and regional edge point of the preset parking physical area are calculated based on the stable point group records, and the regional center point, the regional edge point, and the regional boundary data are jointly written into the point of interest parsing input record. The point of interest (POI) parsing input record retains the source relationship between the geographic coordinate point dataset, the node point grouping record, and the stable point grouping record. When the POI location parsing program outputs the location classification label based on the POI parsing input record, it can trace back to the geographic coordinate point dataset along the POI parsing input record.

[0169] Preferably, before being invoked, the preset point-of-interest (POI) location parsing program pre-configures an POI index table and a location classification mapping table. The POI index table stores the POI location ranges corresponding to parking areas, residential parking areas, charging parking areas, commercial parking areas, roadside parking areas, and park parking areas. The location classification mapping table maps the POI location ranges to location classification labels. When the geographic coordinate point dataset is input into the POI location parsing program, the program first reads the region center point, region edge point, and region boundary data from the POI parsing input record. Then, it performs spatial overlap matching between the region center point, region edge point, and region boundary data and the POI location ranges in the POI index table to generate POI matching records. The point of interest matching record continues to enter the location classification mapping table to output the location classification label, so that the location classification label is not directly determined by a single location point, but is determined by the regionalized spatial location result formed by the region center point, the region edge point, and the region boundary data; the regionalized spatial location result continues to be retained in the point of interest matching record to participate in the subsequent source tracing of location classification labels.

[0170] Preferably, after the location classification label is output, the location classification label is written into the spatial location attribute information, and simultaneously, the point of interest matching record is written into the spatial location attribute information. The location classification label is used to represent the parking scene type to which the preset parking physical area belongs, and the point of interest matching record is used to record the source process of the location classification label being jointly generated by the geographic coordinate point dataset, the point of interest index table, and the location classification mapping table. After the spatial location attribute information is written into the location classification label and the point of interest matching record, it continues to participate in the matching of the subsequent computing power supply stability coefficient and the credibility weight coefficient, so that the computing power supply stability coefficient and the credibility weight coefficient can correspond to the parking scene type of the preset parking physical area, rather than using the same matching criterion for all parking areas.

[0171] Preferably, when the point of interest location parsing program does not output the location classification label, the point of interest location parsing program generates a missing status record. This missing status record indicates that the geographic coordinate point dataset and the point of interest index table do not form a matching record that meets the output conditions, or that the matching record cannot be mapped to a location classification label through the location classification mapping table. After the missing status record is written into the supplementary field of the spatial location attribute information, it triggers the retrieval of vehicle dwell time distribution statistics within a historically set period for the preset parking physical area. The vehicle dwell time distribution statistics are derived from the vehicle entry time, vehicle dwell duration, and vehicle departure time of the target vehicle node within the preset parking physical area. These three times together express the historical dwell status of the target vehicle node within the preset parking physical area. The vehicle dwell time distribution statistics provide alternative classification criteria when the location classification label is missing, thus allowing the spatial location attribute information to still enter the subsequent matching process of the computing power supply stability coefficient and credibility weight coefficient.

[0172] Preferably, after the vehicle dwell time distribution statistics are retrieved, they are first segmented according to the historical set period to obtain segmented dwell time statistics records. These segmented dwell time statistics records include vehicle dwell time distribution, vehicle entry frequency distribution, and vehicle departure frequency distribution. The vehicle dwell time distribution expresses the continuous dwelling characteristics of the target vehicle node within the preset dwelling physical area, while the vehicle entry frequency distribution and vehicle departure frequency distribution express the entry and departure variation characteristics of the target vehicle node within the preset dwelling physical area. Dwelling stability description data is extracted based on the segmented dwell time statistics records, and this data is further matched with preset dwelling scenario classification rules to determine the alternative location classification label. The dwelling stability description data originates from the segmented dwell time statistics records and continues to serve as the classification basis for the alternative location classification label.

[0173] Preferably, the parking scenario classification rules are pre-configured according to the vehicle parking duration distribution statistics corresponding to different parking scenarios. The parking scenario classification rules include long-term parking rules, short-term parking rules, periodic parking rules, and charging parking rules. Specifically, the long-term parking rules are used for residential parking areas or park parking areas, the short-term parking rules are used for roadside parking areas or commercial parking areas, the periodic parking rules are used for parking lot areas, and the charging parking rules are used for charging parking areas. After matching the parking stability description data with the long-term parking rules, the short-term parking rules, the periodic parking rules, and the charging parking rules respectively, parking scenario matching records are obtained. The alternative location classification label is determined based on the parking scenario matching records. The alternative location classification label retains its source relationship with the vehicle parking duration distribution statistics, so that the alternative location classification label can continue to explain its data source when the location classification label is subsequently replaced.

[0174] Preferably, after determining the alternative location classification label, the alternative location classification label is first associated with the missing status record to generate an alternative classification association record. The alternative classification association record includes the missing status record, the vehicle dwell time distribution statistics, the segmented dwell time statistics, the dwell stability description data, and the dwell scene matching record; wherein, the missing status record explains the reason for activating the alternative location classification label, the vehicle dwell time distribution statistics and the segmented dwell time statistics explain the statistical source of the alternative location classification label, and the dwell stability description data and the dwell scene matching record explain the classification process of the alternative location classification label. The alternative classification association record continues to write the spatial location attribute information, so that the alternative location classification label is not an isolated replacement result, but a spatial location attribute component with a complete source and subsequent usage path.

[0175] Preferably, when replacing the location category label with the alternative location category label, the missing status record is not deleted. Instead, the missing status record, the alternative location category label, and the alternative category association record are all written into the spatial location attribute information. The currently valid category field in the spatial location attribute information is updated to the alternative location category label. The currently valid category field is used for subsequent matching of the computing power supply stability coefficient and the credibility weight coefficient. The missing status record and the alternative category association record are used to indicate that the currently valid category field comes from the vehicle dwell time distribution statistics, rather than from the location category label directly output by the point of interest location parsing program. Through the above writing method, the alternative location category label can still maintain a technical correlation with the vehicle dwell time distribution statistics after replacing the location category label.

[0176] Preferably, after using the location classification label or the alternative location classification label as the current valid classification field, the spatial location attribute information is further written with a scene classification source identifier. The scene classification source identifier distinguishes whether the current valid classification field originates from the point of interest location parsing program or from the vehicle dwell time distribution statistics. When the current valid classification field originates from the point of interest location parsing program, the scene classification source identifier corresponds to the point of interest matching record; when the current valid classification field originates from the vehicle dwell time distribution statistics, the scene classification source identifier corresponds to the alternative classification association record. The scene classification source identifier subsequently participates in the credibility weight coefficient matching process, enabling the credibility weight coefficient to be configured differently based on the source type of the spatial location attribute information.

[0177] Preferably, within a parking area, the geographic coordinate point dataset can spatially overlap with the location range of points of interest corresponding to the parking area in the point of interest index table. The point of interest location parsing program outputs the location classification label corresponding to the parking area based on the point of interest matching record, and uses the location classification label corresponding to the parking area as the currently valid classification field in the spatial location attribute information. In roadside parking areas or commercial parking areas, if the geographic coordinate point dataset cannot form a point of interest matching record that satisfies the output conditions with the point of interest index table, the missing state record triggers the retrieval of the vehicle dwell time distribution statistics, and determines the alternative location classification label corresponding to the roadside parking area or commercial parking area through the dwell time scenario classification rules. Thus, the spatial location attribute information can be compatible with both the processing scenario where the point of interest location parsing program directly outputs the location classification label, and the processing scenario where the point of interest location parsing program determines the alternative location classification label based on the vehicle dwell time distribution statistics after failing to output the location classification label, enabling subsequent node collaborative comprehensive weight value calculations to continuously obtain scenario classification input.

[0178] Preferably, after the spatial location attribute information is generated, it is also checked for consistency with the regional attribution result of the target vehicle node. The regional attribution result is generated based on the regional boundary data and the vehicle parking location reporting data, and is used to determine whether the target vehicle node is located within the preset parking physical area. After checking the consistency between the regional attribution result and the currently valid classification field in the spatial location attribute information, a spatial location attribute verification record is generated. The spatial location attribute verification record is used to indicate that the currently valid classification field and the target vehicle node participating in the subsequent available computing power conversion in this cycle belong to the same preset parking physical area. The spatial location attribute verification record continues to participate in the matching of the subsequent computing power supply stability coefficient and credibility weight coefficient, so that the spatial location attribute information and the available computing power conversion processing of the target vehicle node maintain consistency in regional source.

[0179] Optionally, in the step of parsing the task processing request message to obtain the target computing unit reference identifier, if the parsed and extracted target computing unit reference identifier contains mixed precision indication information, the method further includes:

[0180] Extract the first execution stage data block and the second execution stage data block from the task processing request message;

[0181] The first computational load corresponding to the data block in the first execution stage and the second computational load corresponding to the data block in the second execution stage are calculated respectively.

[0182] A total computing load is generated based on the first computing load and the second computing load. A first percentage parameter of the first computing load in the total computing load is calculated based on the total computing load. A second percentage parameter of the second computing load in the total computing load is calculated based on the total computing load.

[0183] The reference identifier for the standard computing power unit with integer precision is assigned according to the first proportion parameter, and the reference identifier for the standard computing power unit with floating-point precision is assigned according to the second proportion parameter.

[0184] Based on the integer precision standard computing power unit reference identifier, the first proportion parameter, the floating point precision standard computing power unit reference identifier, and the second proportion parameter combination, a hybrid reference target computing power unit reference identifier is generated, and the hybrid reference target computing power unit reference identifier is used as the target computing power unit reference identifier.

[0185] Preferably, during the process of parsing the task processing request message to obtain the target computing unit benchmark identifier, if the precision requirement field in the task processing request message contains both integer precision requirements and floating-point precision requirements, a precision requirement judgment result is generated, and the mixed precision indication information is generated based on the precision requirement judgment result. The mixed precision indication information is then written into the parsing result of the target computing unit benchmark identifier. The precision requirement field originates from the task description field in the task processing request message. The task description field describes the computational precision requirements, stage input data range, stage output data range, and stage sequence corresponding to different execution stages in the task to be processed. By reading the precision requirement field, the precision requirement judgment result can be generated. The precision requirement judgment result is used to determine whether the task to be processed only requires an integer precision standard computing unit benchmark identifier, whether it only requires a floating-point precision standard computing unit benchmark identifier, or whether it requires the simultaneous use of both the integer precision standard computing unit benchmark identifier and the floating-point precision standard computing unit benchmark identifier. If the accuracy requirement judgment result is that both the integer precision standard computing unit reference identifier and the floating-point precision standard computing unit reference identifier are used simultaneously, then the mixed precision indication information will be used as the triggering basis for the subsequent extraction of the first execution stage data block and the second execution stage data block, so that the generation process of the target computing unit reference identifier can be consistent with the calculation accuracy requirement in the task processing request message.

[0186] Preferably, the first execution stage data block and the second execution stage data block are not arbitrary data fragments extracted from the task processing request message, but rather structured data blocks obtained by parsing the stage boundary field, stage input field, stage output field, and stage precision field in the task processing request message. The stage boundary field defines the start and end positions of different execution stages in the task processing request message; the stage input field defines the data objects to be processed in the corresponding execution stage; the stage output field defines the intermediate or final processing result formed by the corresponding execution stage; and the stage precision field defines whether the corresponding execution stage uses integer precision or floating-point precision calculation. After jointly parsing the stage boundary field, the stage input field, the stage output field, and the stage precision field, the execution stage content calculated using integer precision is extracted as the first execution stage data block, and the execution stage content calculated using floating-point precision is extracted as the second execution stage data block. This ensures that the first and second execution stage data blocks each have a clear data source, precision caliber, and subsequent calculation purpose. The data source, precision caliber, and subsequent calculation purpose are then written into the corresponding first and second execution stage data blocks for subsequent statistical analysis of the first and second computational loads.

[0187] Preferably, after the first execution stage data block is extracted, the first stage input data volume, first stage operation type identifier, first stage repeated execution count, and first stage data precision identifier are read from the first execution stage data block. Based on these parameters, the first computational load is calculated. The first stage input data volume expresses the scale of data in the first execution stage data block that needs to be processed by integer precision calculation. The first stage operation type identifier expresses the type of integer precision calculation operation corresponding to the first execution stage data block. The first stage repeated execution count expresses the number of times the integer precision calculation operation corresponding to the first execution stage data block needs to be executed. The first stage data precision identifier confirms that the first execution stage data block belongs to the integer precision calculation scope. The first computational load obtained after statistically processing the first stage input data volume, first stage operation type identifier, first stage repeated execution count, and first stage data precision identifier continues to serve as the data source for calculating the overall computational load and the first proportion parameter, rather than merely as supplementary descriptive information for the first execution stage data block.

[0188] Preferably, after the second execution stage data block is extracted, the second stage input data volume, second stage operation type identifier, second stage repeated execution count, and second stage data precision identifier are first read from the second execution stage data block. Based on these parameters, the second computational load is calculated. The second stage input data volume expresses the scale of data in the second execution stage data block that needs to be processed by floating-point precision calculation. The second stage operation type identifier expresses the floating-point precision calculation operation type corresponding to the second execution stage data block. The second stage repeated execution count expresses the number of times the floating-point precision calculation operation corresponding to the second execution stage data block needs to be executed. The second stage data precision identifier confirms that the second execution stage data block belongs to the floating-point precision calculation scope. The second computational load obtained after statistical processing of the second stage input data volume, second stage operation type identifier, second stage repeated execution count, and second stage data precision identifier continues to serve as the data source for calculating the overall computational load and the second proportion parameter, thus establishing a direct technical correlation between the second execution stage data block and the floating-point precision standard computing unit benchmark identifier.

[0189] Preferably, after obtaining the first computational load and the second computational load, the first and second computational loads are first processed to unify their load scope, so that both the first and second computational loads are expressed as the stage computational load scale within the same task processing request message. This load scope unification process does not involve directly adding the integer precision standard computing unit value and the floating-point precision standard computing unit value. Instead, it performs stage load statistics under the same statistical scope on the data volume, operation type, and number of repeated executions corresponding to the first and second execution stage data blocks. After the load scope unification process is completed, the first and second computational loads are merged and statistically analyzed to generate the overall computational load. The overall computational load is used to express the overall stage load scale of the mixed-precision task in the task processing request message and continues to participate in the calculation of the first and second proportion parameters, ensuring that both the first and second proportion parameters originate from the overall computational load under the same statistical scope.

[0190] Preferably, the calculation process of the first proportion parameter is as follows: After the total computing load is generated, the first computing load and the total computing load are calculated as a proportion under the same statistical caliber to obtain the first proportion parameter. The first proportion parameter is used to express the proportion of the integer precision calculation stage corresponding to the first execution stage data block in the total computing load. The first proportion parameter continues to serve as the weight basis for allocating the integer precision standard computing power unit benchmark identifier. Through the first proportion parameter, the integer precision calculation requirement in the task processing request message can be retained in the target computing power unit benchmark identifier, so that when the expected set of computing power supply is retrieved from the dual benchmark standard computing power unit sequence, the integer precision standard computing power unit requirement corresponding to the first execution stage data block can be identified. The integer precision standard computing power unit requirement continues to establish a correspondence with the integer precision standard computing power unit benchmark identifier to participate in the subsequent generation of the hybrid benchmark target computing power unit benchmark identifier.

[0191] Preferably, the calculation process of the second proportion parameter is as follows: After the overall computing load is generated, the second computing load and the overall computing load are calculated as a proportion under the same statistical caliber to obtain the second proportion parameter. The second proportion parameter is used to express the proportion of the floating-point precision computing stage corresponding to the second execution stage data block in the overall computing load. The second proportion parameter continues to serve as the weight basis for allocating the floating-point precision standard computing unit benchmark identifier. Through the second proportion parameter, the floating-point precision computing requirement in the task processing request message can be retained in the target computing unit benchmark identifier, so that when the expected set of computing power supply is retrieved from the dual benchmark standard computing unit sequence, the floating-point precision standard computing unit requirement corresponding to the second execution stage data block can be identified. The floating-point precision standard computing unit requirement continues to establish a correspondence with the floating-point precision standard computing unit benchmark identifier to participate in the subsequent generation of the hybrid benchmark target computing unit benchmark identifier.

[0192] Preferably, when allocating the integer precision standard computing unit benchmark identifier according to the first proportion parameter, the first stage data precision identifier in the first execution stage data block is first read, and it is confirmed that the first stage data precision identifier corresponds to the integer precision standard computing unit benchmark identifier. After confirming the correspondence, the first proportion parameter, the first computing load, the first execution stage data block, the integer precision standard computing unit requirement, and the integer precision standard computing unit benchmark identifier are jointly written into the integer precision stage benchmark record. The integer precision stage benchmark record is used to indicate that the first execution stage data block should be subsequently matched through the integer precision standard computing unit benchmark identifier, while retaining the influence of the first proportion parameter on the subsequent retrieval of the expected parameter set for computing power supply. The integer precision stage benchmark record continues to participate in the generation of the hybrid benchmark target computing unit benchmark identifier, so that the integer precision standard computing unit benchmark identifier is not a standalone marker, but forms a continuous data link with the first execution stage data block, the first computing load, the integer precision standard computing unit requirement, and the first proportion parameter.

[0193] Preferably, when allocating the floating-point precision standard computing unit benchmark identifier according to the second proportion parameter, the second stage data precision identifier in the second execution stage data block is first read, and it is confirmed that the second stage data precision identifier corresponds to the floating-point precision standard computing unit benchmark identifier. After confirming the correspondence, the second proportion parameter, the second computing load, the second execution stage data block, the floating-point precision standard computing unit requirement, and the floating-point precision standard computing unit benchmark identifier are jointly written into the floating-point precision stage benchmark record. The floating-point precision stage benchmark record is used to indicate that the second execution stage data block should be subsequently matched through the floating-point precision standard computing unit benchmark identifier, while retaining the influence of the second proportion parameter on the subsequent retrieval of the expected parameter set for computing power supply. The floating-point precision stage benchmark record continues to participate in the generation of the hybrid benchmark target computing unit benchmark identifier, so that the floating-point precision standard computing unit benchmark identifier forms a continuous data link with the second execution stage data block, the second computing load, the floating-point precision standard computing unit requirement, and the second proportion parameter.

[0194] Preferably, the generation process of the hybrid benchmark target computing power unit benchmark identifier specifically involves: reading the integer precision stage benchmark record and the floating-point precision stage benchmark record; combining the integer precision standard computing power unit benchmark identifier, the first proportion parameter, the floating-point precision standard computing power unit benchmark identifier, and the second proportion parameter in the order of the stages to generate the hybrid benchmark target computing power unit benchmark identifier. The hybrid benchmark target computing power unit benchmark identifier includes an integer precision stage benchmark record reference field, a floating-point precision stage benchmark record reference field, a first proportion parameter field, and a second proportion parameter field. The integer precision stage benchmark record reference field is used to locate the integer precision standard computing power unit value when subsequently retrieving the dual-benchmark standard computing power unit sequence; the floating-point precision stage benchmark record reference field is used to locate the floating-point precision standard computing power unit value when subsequently retrieving the dual-benchmark standard computing power unit sequence; and the first proportion parameter field and the second proportion parameter field are used to define the weight relationship between different precision stages in subsequent supply matching. The reference field of the integer precision stage benchmark record corresponds to the integer precision stage benchmark record, and the reference field of the floating-point precision stage benchmark record corresponds to the floating-point precision stage benchmark record, so that the benchmark identifier of the hybrid benchmark target computing unit can trace back the first execution stage data block and the second execution stage data block respectively.

[0195] Preferably, after generating the hybrid benchmark target computing unit benchmark identifier, a benchmark integrity check is performed on the hybrid benchmark target computing unit benchmark identifier. The benchmark integrity check includes checking whether the hybrid benchmark target computing unit benchmark identifier contains the integer precision standard computing unit benchmark identifier, checking whether the hybrid benchmark target computing unit benchmark identifier contains the floating-point precision standard computing unit benchmark identifier, checking whether the first proportion parameter corresponds to the first execution stage data block, checking whether the second proportion parameter corresponds to the second execution stage data block, checking whether the integer precision stage benchmark record corresponds to the reference field of the integer precision stage benchmark record, and checking whether the floating-point precision stage benchmark record corresponds to the reference field of the floating-point precision stage benchmark record. After the benchmark integrity check passes, the hybrid benchmark target computing unit benchmark identifier is used as the target computing unit benchmark identifier, so that when the dual-benchmark standard computing unit sequence is subsequently retrieved, matching can be performed according to the integer precision stage benchmark record and the floating-point precision stage benchmark record respectively, instead of classifying the hybrid precision task into a single computing unit benchmark caliber.

[0196] Preferably, the hybrid benchmark target computing unit benchmark identifier, after serving as the target computing unit benchmark identifier, continues to participate in the retrieval of the computing power supply expectation parameter set. Specifically, based on the integer precision stage benchmark record reference field in the hybrid benchmark target computing unit benchmark identifier, the integer precision computing power supply expectation parameter matching the integer precision standard computing unit benchmark identifier is retrieved from the dual-benchmark standard computing unit sequence; based on the floating-point precision stage benchmark record reference field in the hybrid benchmark target computing unit benchmark identifier, the floating-point precision computing power supply expectation parameter matching the floating-point precision standard computing unit benchmark identifier is retrieved from the dual-benchmark standard computing unit sequence. Subsequently, the integer precision computing power supply expectation parameter is marked with a stage weight based on the first proportion parameter, and the floating-point precision computing power supply expectation parameter is marked with a stage weight based on the second proportion parameter, so that the computing power supply expectation parameter set can retain the supply matching relationship of different execution stages in the hybrid precision task. The supply matching relationship is then used to generate the supply ladder distribution sequence, so that the supply ladder distribution sequence can simultaneously reflect the supply data of the integer precision calculation stage and the floating-point precision calculation stage.

[0197] Preferably, in one scenario, if the task to be processed corresponding to the task processing request message first executes the quantization inference stage and then the half-precision floating-point calculation stage, then the quantization inference stage is parsed into the first execution stage data block, and the half-precision floating-point calculation stage is parsed into the second execution stage data block. The first execution stage data block is associated with the integer precision standard computing power unit benchmark identifier through the first computational load and the first proportion parameter, and the second execution stage data block is associated with the floating-point precision standard computing power unit benchmark identifier through the second computational load and the second proportion parameter. Thus, the hybrid benchmark target computing power unit benchmark identifier can simultaneously express the integer precision calculation requirements of the quantization inference stage and the floating-point precision calculation requirements of the half-precision floating-point calculation stage. When generating the supply ladder distribution sequence subsequently, comparable supply data can be organized based on the integer precision standard computing power unit benchmark identifier and the floating-point precision standard computing power unit benchmark identifier, respectively. The supply data continues to enter the bidirectional data constraint iteration operation to generate continuous iterative output results corresponding to the hybrid precision task.

[0198] Preferably, the hybrid benchmark target computing power unit benchmark identifier is further written into the measurement benchmark field of the historical sequence database. When the measurement benchmark field is written to the hybrid benchmark target computing power unit benchmark identifier, it simultaneously retains the integer precision stage benchmark record, the floating-point precision stage benchmark record, the first proportion parameter, and the second proportion parameter. This allows for differentiation between single integer precision tasks, single floating-point precision tasks, and hybrid precision tasks when subsequently extracting historical benchmark price value sequences or performing moving weighted average processing operations. Through this writing method, the hybrid benchmark target computing power unit benchmark identifier not only participates in the current period's computing power supply expected parameter set retrieval and bidirectional data constraint iteration operations, but also continues to participate in subsequent historical sequence database queries and prediction interval value generation, preventing the benchmark source of the hybrid precision task from being lost in subsequent data processing.

[0199] Optionally, after the step of triggering the next data loop processing process based on the next sequence node value in the supply ladder distribution sequence, provided that the deviation value does not fall within the preset convergence interval, the method further includes:

[0200] Record the cumulative number of times the data loop processing process is executed;

[0201] Determine whether the cumulative execution count has reached the preset execution count limit parameter;

[0202] Under abnormal triggering conditions where the cumulative execution count reaches the preset execution count limit parameter, the bidirectional data constraint iteration operation is stopped;

[0203] Retrieve the benchmark price value broadcast in the previous period and extract the initial sequence node value in the supply ladder distribution sequence;

[0204] A weighted average calculation is performed on the benchmark price value broadcast in the previous period and the initial sequence node value to extract the weighted average calculation result as the compensation benchmark price value, and then the compensation benchmark price value is used to replace the benchmark price value in the current period.

[0205] Preferably, when the deviation value does not fall within the preset convergence interval, after triggering the next data cycle processing process based on the next sequence node value in the supply ladder distribution sequence, the iteration status record corresponding to the current data cycle processing process is first read. The iteration status record includes the current sequence node value, the next sequence node value, the first node response parameter, the second node request parameter, the deviation value, the target computing power unit benchmark identifier, and the round identifier of the data cycle processing process. The round identifier of the data cycle processing process is used to distinguish the cycle order that the bidirectional data constraint iteration operation has been executed in this cycle. The current sequence node value and the next sequence node value are used to express the traversal position of the supply ladder distribution sequence in two adjacent data cycle processing processes. The deviation value is used to express the degree of difference between the first node response parameter and the second node request parameter. Based on the round identifier of the data loop processing process, the data loop processing process is sequentially accumulated to obtain the cumulative execution count value, so that the cumulative execution count value can maintain the same iteration source as the traversal process of the supply ladder distribution sequence, the change process of the deviation value, and the computing power unit caliber corresponding to the target computing power unit benchmark identifier.

[0206] Preferably, the cumulative execution count is not a separately recorded loop count, but a data field written into the anomaly convergence monitoring record and used in the subsequent determination of the preset execution count limit parameter. The anomaly convergence monitoring record includes the cumulative execution count, the preset execution count limit parameter, the deviation value corresponding to the most recent data loop processing process, the current sequence node value corresponding to the most recent data loop processing process, and the target computing unit baseline identifier. The preset execution count limit parameter is configured before the bidirectional data constraint iteration operation starts, and it limits the maximum number of data loop processing processes allowed to execute within the same refresh cycle. After the anomaly convergence monitoring record is formed, it continues to be used to determine whether the cumulative execution count has reached the preset execution count limit parameter, ensuring that the destination of the cumulative execution count can be directly linked to the stopping condition of the bidirectional data constraint iteration operation.

[0207] Preferably, the configuration process of the preset execution count limit parameter can be pre-set based on the target computing unit benchmark identifier, the length of the supply ladder distribution sequence, and the adaptive refresh cycle. For quantization inference tasks corresponding to integer precision standard computing unit benchmark identifiers, if the adaptive refresh cycle is short, the preset execution count limit parameter can be configured as a lower round limit. The lower round limit is used to reduce the broadcast delay caused by multiple traversals of the supply ladder distribution sequence within the same refresh cycle. For high-precision computing tasks corresponding to floating-point precision standard computing unit benchmark identifiers, if the supply ladder distribution sequence contains many supply-side sequence nodes, the preset execution count limit parameter can be configured as a higher round limit. The higher round limit is used to enable multiple sequence node values ​​in the supply ladder distribution sequence to participate in the round-by-round judgment of the deviation value. Through the above configuration method, the preset execution count limit parameter is not an isolated threshold, but rather, together with the target computing unit benchmark identifier, the supply ladder distribution sequence, and the adaptive refresh cycle, it defines the executable range of the bidirectional data constraint iteration operation.

[0208] Preferably, the specific implementation process for determining whether the cumulative execution count has reached the preset execution count limit parameter is as follows: The cumulative execution count and the preset execution count limit parameter are read from the abnormal convergence monitoring record. The cumulative execution count and the preset execution count limit parameter are compared using integers of the same type to generate an execution count determination result. If the execution count determination result indicates that the cumulative execution count has not reached the preset execution count limit parameter, then the next data loop processing process is triggered based on the next sequence node value in the supply ladder distribution sequence, and the deviation value generated by the next data loop processing process is written back to the abnormal convergence monitoring record. If the execution count determination result indicates that the cumulative execution count has reached the preset execution count limit parameter, then an abnormal triggering condition record is generated. The abnormal triggering condition record is used to indicate that the bidirectional data constraint iteration operation in this cycle has reached the upper limit of the allowed data loop processing process, but the deviation value corresponding to the most recent data loop processing process has not yet fallen into the preset convergence interval.

[0209] Preferably, the specific implementation process for stopping the bidirectional data constraint iteration operation under the abnormal trigger condition that the cumulative execution count reaches the preset execution count upper limit parameter is as follows: First, read the target computing unit benchmark identifier, the deviation value corresponding to the most recent data loop processing process, and the current sequence node value corresponding to the most recent data loop processing process from the abnormal trigger condition record; then, write the abnormal trigger condition record into the current cycle iteration termination record. The current cycle iteration termination record is used to mark that the bidirectional data constraint iteration operation in this cycle has not formed a convergence value through the preset convergence interval, and is used to prevent the selection of the next sequence node value in the supply ladder distribution sequence. After the current cycle iteration termination record is formed, the bidirectional data constraint iteration operation no longer continues to update the first node response parameter and the second node request parameter, but instead enters the process of generating the compensation benchmark price value, so that the subsequent benchmark price value replacement processing in the non-converged state has a clear trigger source.

[0210] Preferably, when retrieving the benchmark price value broadcast in the previous cycle, the broadcast record of the previous cycle that matches the benchmark identifier of the target computing power unit is first retrieved from the broadcast scheduling record or historical sequence database. The previous cycle broadcast record includes the benchmark price value broadcast in the previous cycle, the timestamp information corresponding to the previous cycle, the benchmark identifier of the target computing power unit corresponding to the previous cycle, and the broadcast status identifier corresponding to the previous cycle. The benchmark identifier of the target computing power unit is used to verify that the benchmark price value broadcast in the previous cycle and the non-converged bidirectional data constraint iteration operation in the current cycle belong to the same computing power unit. The timestamp information is used to confirm that the benchmark price value broadcast in the previous cycle belongs to valid broadcast data within the adjacent refresh cycle. The broadcast status identifier is used to confirm that the benchmark price value broadcast in the previous cycle has been broadcast to the preset node cluster through the network interface. Through the above retrieval and verification, the benchmark price value broadcast in the previous cycle can serve as a stable reference source for the compensation benchmark price value, rather than being arbitrarily read from different computing power unit or invalid broadcast records.

[0211] Preferably, when extracting the initial sequence node values ​​from the supply tiered distribution sequence, the sorting status record corresponding to the supply tiered distribution sequence is first read. The sorting status record indicates that the set of expected computing power supply parameters has been sorted in ascending order to form the supply tiered distribution sequence. Based on the sorting status record, the initial sequence node values ​​are extracted from the starting supply-side sequence nodes of the supply tiered distribution sequence, and the initial sequence node values ​​are written into the initial supply record for this period. The initial supply record for this period includes the initial sequence node values, the target vehicle node identifier corresponding to the initial sequence node values, the target computing power unit reference identifier corresponding to the initial sequence node values, and the starting tier position of the initial sequence node values ​​in the supply tiered distribution sequence. The initial sequence node values ​​are used to express the starting reference data in the supply-side sequence nodes of this period. The initial supply record for this period is used to retain the node source, computing power unit caliber source, and starting tier position source of the initial sequence node values, and continues to participate in subsequent weighted average calculation operations.

[0212] Preferably, before performing a weighted average calculation on the benchmark price value broadcast in the previous period and the initial sequence node value, a caliber verification is performed on the benchmark price value broadcast in the previous period and the initial sequence node value. This caliber verification includes verifying whether the target computing unit benchmark identifier corresponding to the benchmark price value broadcast in the previous period is consistent with the target computing unit benchmark identifier corresponding to the initial sequence node value, and verifying whether both the benchmark price value broadcast in the previous period and the initial sequence node value are scalar data that can be used for numerical calculations. After the caliber verification passes, the benchmark price value broadcast in the previous period is written into the historical reference field, the initial sequence node value is written into the current supply reference field, and a weighted average calculation is performed on the historical reference field and the current supply reference field according to a preset compensation weight configuration record to extract the weighted average calculation result. The compensation weight configuration record is used to limit the participation ratio of the benchmark price value and the initial sequence node value of the previous period broadcast in the non-convergence scenario, so that the weighted average calculation result retains the reference role of the previous period broadcast data corresponding to the historical reference field and the reference role of the current period supply ladder distribution sequence corresponding to the current period supply reference field.

[0213] Preferably, the compensation weight configuration record can be pre-configured based on the adaptive refresh cycle, the cumulative execution count, and the direction of change of the deviation value. If the adaptive refresh cycle is short and the deviation value gradually decreases during multiple data cycle processing, the compensation weight configuration record can increase the participation ratio of the initial sequence node value to reflect the influence of the supply-side sequence nodes on the compensation benchmark price value in this cycle. If the deviation value fluctuates significantly during multiple data cycle processing, or the supply-side sequence nodes in the supply ladder distribution sequence change rapidly, the compensation weight configuration record can increase the participation ratio of the benchmark price value broadcast in the previous cycle to reduce the fluctuation impact of non-converged data on the broadcast result in this cycle. The compensation weight configuration record and the abnormal convergence monitoring record jointly participate in the weighted average calculation operation, enabling the weighted average calculation result to simultaneously correlate the iteration state of this cycle and the broadcast state of the previous cycle.

[0214] Preferably, after the weighted average calculation result is generated, the weighted average calculation result, the benchmark price value broadcast in the previous period, the initial sequence node value, the compensation weight configuration record, the abnormal triggering condition record, and the target computing power unit benchmark identifier are jointly written into the compensation benchmark price record. The compensation benchmark price record is used to explain the source of the weighted average calculation result, including the non-convergence triggering source provided by the abnormal triggering condition record, the broadcast data source of the previous period provided by the previous period broadcast record, the starting source of the supply-side sequence nodes of the current period provided by the initial supply record of the current period, and the configuration source of the compensation weight configuration record. The compensation benchmark price record continues to participate in the generation of the broadcast scheduling record of the current period, ensuring that the weighted average calculation result has a traceable data link before being used as the compensation benchmark price value. Through the compensation benchmark price record, the weighted average calculation result can establish a direct link with the abnormal triggering conditions of the bidirectional data constraint iterative operation.

[0215] Preferably, after extracting the weighted average calculation result as the compensation benchmark price value, the compensation benchmark price value replaces the benchmark price value of the current period. Specifically, the benchmark price value of the current period is first read from the current period's benchmark price field and written to the replaced benchmark price record. Then, the compensation benchmark price value is written to the current period's benchmark price field, and a compensation status identifier is written to the current period's benchmark price field. The replaced benchmark price record indicates that the current period failed to form a broadcastable benchmark price value within the preset convergence interval, and the compensation status identifier indicates that the compensation benchmark price value is used for broadcasting in the current period. The compensation status identifier is further written to the broadcast scheduling record, enabling the preset node cluster to identify that the current period's benchmark price value originates from the compensation benchmark price record, rather than from continuous iterative output results that have already fallen within the preset convergence interval when receiving broadcast data for the current period.

[0216] Preferably, when performing the above-mentioned non-convergence compensation processing in the parking area, if multiple target vehicle nodes in the parking area enter or leave the preset parking physical area within a short period of time, causing the sequence node values ​​of the supply tiered distribution sequence to continuously change during multiple data loop processing processes, and the deviation value still does not fall into the preset convergence interval before reaching the preset execution count limit parameter, the abnormal trigger condition record stops the bidirectional data constraint iteration operation and triggers the reading of the benchmark price value broadcast in the previous period and the initial sequence node value of the current period. The benchmark price value broadcast in the previous period is written into the historical reference field, and the initial sequence node value of the current period is written into the current supply reference field. The historical reference field and the current supply reference field are weighted and averaged by the compensation weight configuration record to form the compensation benchmark price value. After the compensation benchmark price value replaces the benchmark price value of the current period, it can continue to be broadcast to the preset node cluster through the network interface, avoiding the loss of broadcast scheduling records due to non-convergence of iteration within a single refresh cycle.

[0217] Preferably, after the compensation benchmark price value replaces the benchmark price value of the current period, the compensation benchmark price record can also be written into the historical sequence database. When writing the compensation benchmark price record, the historical sequence database simultaneously saves the compensation status identifier, the cumulative execution count, the preset execution count limit parameter, the benchmark price value broadcast in the previous period, the initial sequence node value, and the target computing power unit benchmark identifier. The compensation status identifier is used to distinguish between the benchmark price value obtained through the preset convergence interval and the compensation benchmark price value obtained through weighted average calculation. The cumulative execution count and the preset execution count limit parameter are used to explain the non-convergence execution conditions of the current period. The benchmark price value broadcast in the previous period and the initial sequence node value are used to explain the two types of value sources of the compensation benchmark price value. When the historical sequence database is subsequently used to extract historical benchmark price value sequences or generate prediction interval values, it can identify the data records of the compensation source based on the compensation status identifier, enabling the historical sequence processing process to retain compensation information in the non-convergence state of the current period.

[0218] Optionally, after the step of broadcasting the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh period, the method further includes:

[0219] The benchmark price value, along with the corresponding timestamp information, is written into the historical sequence database;

[0220] Extract the historical benchmark price value sequence within a preset backtracking time period from the historical sequence database, and perform a moving weighted average processing operation on the historical benchmark price value sequence to obtain the moving average parameter;

[0221] Calculate the data distribution slope parameter corresponding to the supply ladder distribution sequence, and perform a fusion processing operation on the data distribution slope parameter and the moving average parameter to generate a prediction interval value;

[0222] The predicted interval values ​​are sent to the computing power request node to instruct the computing power request node to plan the task access timing.

[0223] Preferably, after broadcasting the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh cycle, a historical price write record is first generated. This historical price write record includes the benchmark price value, the timestamp information, the target computing power unit benchmark identifier, the total supply value corresponding to the supply tiered distribution sequence, the total demand value corresponding to the task processing request message, and the total number of available vehicle nodes for the current cycle. The timestamp information is used to mark the time position when the benchmark price value is formed and enters the broadcast scheduling record. The target computing power unit benchmark identifier is used to mark the integer precision standard computing power unit caliber, floating-point precision standard computing power unit caliber, or hybrid benchmark target computing power unit benchmark identifier corresponding to the benchmark price value. The total supply value is used to express the number of standard computing power units that can enter the supply tiered distribution sequence within the current cycle. The total demand value is used to express the number of standard computing power units that the computing power request node still needs to access within the current cycle. The total number of available vehicle nodes is used to express the number of target vehicle nodes that are in a network-connected state and can participate in the subsequent task access timing planning within the current cycle. Through the historical price writing record, the benchmark price value is not written into the historical sequence database in the form of an isolated value, but is written into the historical sequence database together with the target computing unit benchmark identifier, supply tiered distribution sequence, task processing request message and the total number of available vehicle nodes for the current period.

[0224] Preferably, when writing the historical price record into the historical sequence database, the historical price record is first subjected to field integrity verification. This field integrity verification includes verifying the existence of the benchmark price value, verifying whether the timestamp information corresponds to the broadcast scheduling record of the current period, verifying whether the target computing unit benchmark identifier corresponds to the benchmark price value, and verifying whether the total supply value, the total demand value, and the total number of available vehicle nodes all originate from the supply tiered distribution sequence, the task processing request message, and the available vehicle node statistics record of the current period. After the field integrity verification passes, the historical price record is written into the historical sequence database according to the chronological order of the timestamp information. The historical sequence database retains the time sequence field, the target computing unit benchmark identifier field, and the supply and demand status field for each historical price write record. The supply and demand status field includes the total supply value, the total demand value, and the total number of available vehicle nodes. This ensures that when extracting the historical benchmark price value sequence, the historical price write records can be read according to the same computing unit caliber, the same time range, and the same supply and demand status field, thus avoiding mixing historical benchmark price values ​​under different computing unit calibers into the same historical benchmark price value sequence.

[0225] Preferably, before extracting the historical benchmark price value sequence within a preset backtracking time period from the historical sequence database, a historical sequence query record is generated based on the task access timing prediction requirements of the computing power request node. The historical sequence query record includes the preset backtracking time period, the target computing power unit benchmark identifier, the preset stationing physical area, and the prediction tier identifier. The task access timing prediction requirements originate from the task access timing requirements in the task processing request message, and these requirements trigger the generation of the historical sequence query record. The preset backtracking time period limits the time range for reading historical price write records from the historical sequence database. The target computing power unit benchmark identifier limits the reading result to be within the same computing power unit caliber as the current task processing request message. The preset stationing physical area limits the reading result to be consistent with the preset stationing physical area corresponding to the current broadcast. The prediction tier identifier distinguishes between shorter time range predictions, medium time range predictions, and longer time range predictions. After extracting historical price write records from the historical sequence database based on the historical sequence query record, the extracted historical price write records are sorted according to the timestamp information to form the historical benchmark price value sequence.

[0226] Preferably, after the historical benchmark price value sequence is formed, a homology check is performed on the historical benchmark price value sequence. The homology check includes checking whether each historical benchmark price value in the historical benchmark price value sequence corresponds to the same target computing power unit benchmark identifier, checking whether each historical benchmark price value in the historical benchmark price value sequence belongs to the same preset stationing physical region, and checking whether each historical benchmark price value in the historical benchmark price value sequence has corresponding timestamp information. After the homology check passes, the historical benchmark price value sequence continues to enter the moving weighted average processing operation; if the homology check finds that the target computing power unit benchmark identifier of a certain historical benchmark price value is inconsistent with the target computing power unit benchmark identifier corresponding to the current task processing request message, then the historical benchmark price value is removed from the historical benchmark price value sequence, and the removal result is written into the historical sequence filtering record. The historical sequence filtering record continues to be stored in the historical sequence database, so that the source of the moving average parameter and the current task processing request message maintain the same computing power unit caliber.

[0227] Preferably, when performing a moving weighted average processing operation on the historical benchmark price value sequence, the historical benchmark price value sequence is first divided into near-end time segments and far-end time segments according to the timestamp information. Then, a preset moving weight configuration record is read. The moving weight configuration record is used to configure the participation ratio of the near-end time segment and the far-end time segment in the moving weighted average processing operation. The historical benchmark price value corresponding to the near-end time segment is closer to the broadcast time of the current period, and the historical benchmark price value corresponding to the far-end time segment is used to provide a historical fluctuation reference within the preset backtracking time period. The moving weight configuration record usually makes the near-end time segment have a higher participation ratio than the far-end time segment, so that the moving weighted average processing operation can better reflect the recent changes in the entry and exit of parked vehicles, the total number of available vehicle nodes, and the total number of pending task queues. After weighting and summing the historical benchmark price values ​​in the near-end time segment and the far-end time segment, the moving average parameter is obtained. The moving average parameter is used to express the time weighting center reference of the historical benchmark price values ​​within the preset backtracking time period.

[0228] Preferably, after the moving average parameter is generated, the moving average parameter, the historical benchmark price sequence, the moving weight configuration record, and the target computing unit benchmark identifier are jointly written into the moving average calculation record. The moving average calculation record is used to indicate which preset backtracking time period the moving average parameter originates from, which target computing unit benchmark identifier is used, and which set of moving weight configuration records is used for calculation. The moving average calculation record continues to participate in subsequent fusion processing operations, enabling the moving average parameter to serve as a central reference with time, scope, and weight sources when generating the prediction interval values, rather than simply a historical average value.

[0229] Preferably, before calculating the data distribution slope parameter corresponding to the supply tiered distribution sequence, the sorting status record corresponding to the supply tiered distribution sequence, the sequence node values ​​in the supply tiered distribution sequence, and the target vehicle node identifier corresponding to each sequence node value are read first. The sorting status record is used to indicate that the supply tiered distribution sequence has been formed by the set of expected computing power supply parameters in ascending order. The sequence node values ​​are used to express the supply-side numerical position of each target vehicle node under the same target computing power unit reference identifier. The target vehicle node identifier is used to retain the node source of each sequence node value. Based on the sorting status record and the target vehicle node identifier, a tiered position analysis is performed on the change amplitude between adjacent sequence node values ​​to obtain the data distribution slope parameter. The data distribution slope parameter is used to express the rate of change of the supply tiered distribution sequence in the ascending direction, thereby reflecting whether the supply-side sequence nodes are concentrated or discretely distributed in the current period.

[0230] Preferably, the data distribution slope parameter is not directly determined by the value of a single sequence node, but is formed by the step change relationship between the values ​​of multiple adjacent sequence nodes. Specifically, the values ​​of two adjacent sequence nodes are first read according to the ascending order of the supply step distribution sequence. Then, the change in value between the two adjacent sequence nodes is calculated, and this change in value is correlated with the step position change relationship between the two adjacent sequence nodes to form an adjacent step change record. Multiple adjacent step change records are then summarized to obtain the data distribution slope parameter. The adjacent step change record is used to explain the source of the data distribution slope parameter. The data distribution slope parameter continues to participate in subsequent fusion processing operations, enabling the predicted interval value to simultaneously reflect the time change reference of the historical benchmark price value sequence and the current supply distribution reference of the supply step distribution sequence.

[0231] Preferably, after the data distribution slope parameter is generated, a current supply and demand trend record is generated by combining the total supply value and the total demand value. The current supply and demand trend record includes the total supply value, the total demand value, the total number of available vehicle nodes, the total number of pending task queues, and the data distribution slope parameter; wherein, the total supply value and the total number of available vehicle nodes are used to express the current supply-side scale, the total demand value and the total number of pending task queues are used to express the current request-side scale, and the data distribution slope parameter is used to express the step change state of the supply-side sequence nodes. The current supply and demand trend record continues to be processed together with the moving average calculation record in the fusion processing operation, so that the fusion processing operation can read both the moving average parameter corresponding to the historical benchmark price value sequence and the data distribution slope parameter corresponding to the current supply step distribution sequence.

[0232] Preferably, when performing the fusion processing operation on the data distribution slope parameter and the moving average parameter, the moving average parameter is first used as a prediction center reference, and then the prediction center reference is subjected to interval expansion processing according to the data distribution slope parameter to generate a lower limit value and an upper limit value of the prediction interval. If the data distribution slope parameter indicates that the sequence node values ​​in the supply ladder distribution sequence change relatively gently, the interval expansion processing adopts a narrower interval expansion range; if the data distribution slope parameter indicates that the sequence node values ​​in the supply ladder distribution sequence change relatively steeply, the interval expansion processing adopts a wider interval expansion range. The lower limit value and the upper limit value of the prediction interval together form the prediction interval value, so that the prediction interval value uses both the moving average parameter as a historical time reference and the data distribution slope parameter as a current supply distribution reference.

[0233] Preferably, the fusion processing operation can also adjust the generation method of the prediction interval value according to the prediction level identifier. For shorter time range predictions, the fusion processing operation prioritizes the moving average parameter and the data distribution slope parameter in the current supply and demand trend record, so that the prediction interval value reflects more the changes in historical benchmark price values ​​and supply ladder distribution sequence near the current refresh cycle; for medium time range predictions, the fusion processing operation further reads historical records of the same time period, which are used to express the changes in historical benchmark price values ​​within the same week attribute and similar time segments, and performs deviation correction processing on the historical records of the same time period and the moving average parameter; for longer time range predictions, the fusion processing operation reads historical intraday curve records, which are used to express the changes in benchmark price values ​​in different time segments in multiple historical cycles. With the participation of the prediction level identifier, the historical records of the same time period, and the historical intraday curve records, the prediction interval value can correspond to the time-series planning requirements of different tasks.

[0234] Preferably, after the predicted interval value is generated, the predicted interval value, the lower limit of the predicted interval value, the upper limit of the predicted interval value, the moving average parameter, the data distribution slope parameter, the target computing unit benchmark identifier, and the predicted tier identifier are jointly written into the predicted interval generation record. The predicted interval generation record is used to explain the source of the predicted interval value, including the time-weighted source corresponding to the historical benchmark price value sequence, the current supply distribution source corresponding to the supply ladder distribution sequence, and the prediction window source corresponding to the predicted tier identifier. The predicted interval generation record is further written into the historical sequence database, so that when predicting interval values ​​are generated again subsequently, existing predicted interval generation records can be read to continuously track changes in predicted interval values ​​over different periods.

[0235] Preferably, before sending the predicted interval value to the computing power request node, a predicted interval sending record is first generated. This record includes the predicted interval value, the predicted interval generation record, the node address information of the computing power request node, the task processing request message identifier, and the target computing power unit reference identifier. The task processing request message identifier is used to associate the predicted interval value with the corresponding task processing request message. The node address information of the computing power request node is used to determine the sending target of the predicted interval value. The target computing power unit reference identifier instructs the computing power request node to interpret the predicted interval value according to an integer precision standard computing power unit caliber, a floating-point precision standard computing power unit caliber, or a hybrid reference target computing power unit reference identifier. After field integrity verification, the predicted interval sending record is sent to the computing power request node through the network interface.

[0236] Preferably, after receiving the predicted interval value, the computing power request node generates a task access timing planning record based on the predicted interval value and the target computing power unit reference identifier. The task access timing planning record includes the task processing request message identifier, the predicted interval value, the target computing power unit reference identifier, candidate access time segments, and task access timing status; wherein, the candidate access time segments express the time range within which the computing power request node can initiate target scheduling instructions, and the task access timing status records whether the computing power request node has completed task access timing planning based on the predicted interval value. The task access timing planning record continues to be associated with subsequent target scheduling instructions, ensuring that the predicted interval value is not merely a query result but can participate in the subsequent processing of the computing power request node's planned task access timing.

[0237] Preferably, when performing the above-mentioned historical sequence prediction processing within the parking area, the historical sequence database can store historical price write records within multiple refresh cycles, and the historical price write records form a traceable data order according to the timestamp information. If the total number of available vehicle nodes within the parking area continues to increase in the near-term time segment, the data distribution slope parameter corresponding to the supply ladder distribution sequence can usually reflect that the change in the supply-side sequence nodes tends to be gradual or that the number of supply-side sequence nodes is increasing; if the total number of pending task queues continues to increase in the near-term time segment, after the moving average parameter and the current supply and demand trend record participate in the fusion processing operation, the prediction interval value can reflect the impact of changes in the request-side data scale on the subsequent task access timing planning. Therefore, after receiving the prediction interval value, the computing power request node can select a more suitable candidate access time segment based on the prediction interval value and generate a target scheduling instruction based on the candidate access time segment.

[0238] Preferably, the historical sequence database further distinguishes and stores the benchmark price value obtained through the preset convergence interval and the compensation benchmark price value obtained through the compensation benchmark price record. Specifically, the historical price writing record includes a price source status identifier, which indicates whether the currently written benchmark price value originates from the continuous iteration output result or from the compensation benchmark price record. When extracting the historical benchmark price value sequence, the historical sequence query record can select whether to include the compensation source data record based on the price source status identifier. When the compensation source data record is included in the historical benchmark price value sequence, the moving weighted average processing operation configures a lower participation ratio for the compensation source data record based on the price source status identifier, ensuring that the moving average parameter retains the continuity of the compensation source data record while reducing the disturbance of the non-convergence period to the predicted interval value.

[0239] Preferably, after the predicted interval value is sent to the computing power request node, a prediction sending feedback record is also written to the historical sequence database. The prediction sending feedback record includes the prediction interval sending record, the receiving status of the computing power request node, the task access timing planning record, and the timestamp information; wherein, the receiving status of the computing power request node is used to record whether the predicted interval value has been received by the computing power request node, and the task access timing planning record is used to record whether the computing power request node has planned the task access timing based on the predicted interval value. The prediction sending feedback record continues to be stored together with the prediction interval generation record in the historical sequence database, so that when subsequently backtracking on the historical benchmark price value sequence, moving average parameter, data distribution slope parameter, and predicted interval value, it can be checked whether the predicted interval value has entered the task access timing planning processing of the computing power request node.

[0240] like Figure 3 As shown, this is an embodiment of a vehicle-mounted computing power real-time price index calculation device, characterized in that it includes:

[0241] The standard computing power conversion module is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area, query the preset computing power conversion mapping table based on the hardware specifications, extract the benchmark conversion coefficients corresponding to the hardware specifications, and then use the benchmark conversion coefficients to convert the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence.

[0242] The request message parsing module is used to receive the task processing request message sent by the computing power request node, parse the task processing request message to obtain the target computing power unit benchmark identifier, and extract the computing power acquisition threshold parameter based on the task processing request message.

[0243] The supply ladder generation module is used to retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit in the dual benchmark standard computing power unit sequence, and perform an ascending sorting operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0244] The bidirectional iterative pricing module is used to extract the sequence node values ​​in the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iterative operations on the computing power acquisition threshold parameter in sequence to obtain continuous iterative output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iterative operations falls within the preset convergence interval, the convergence value of the final iterative output is extracted as the benchmark price value.

[0245] The benchmark price broadcasting module is used to broadcast the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh cycle, so as to instruct the computing power requesting node to generate a target scheduling instruction.

[0246] like Figure 4 As shown, an electronic device according to an embodiment of this application includes a processor and a memory. The memory stores a computer program, and when the processor runs the computer program, it performs the following processing:

[0247] The hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area are obtained. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients.

[0248] The system receives a task processing request message sent by a computing power request node, parses the task processing request message to obtain the target computing power unit baseline identifier, and extracts the computing power acquisition threshold parameter based on the task processing request message.

[0249] Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0250] Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value.

[0251] According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the network interface to instruct the computing power request node to generate a target scheduling instruction.

[0252] like Figure 5 As shown, this is a vehicle according to an embodiment of the present application. The vehicle includes an on-board processor, an on-board communication interface, and an on-board memory. The on-board memory stores a real-time price index calculation program for on-board computing power. When the on-board processor runs the real-time price index calculation program for on-board computing power, it performs the following processing:

[0253] The vehicle communication interface is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients.

[0254] The system receives task processing request messages sent by computing power request nodes through the vehicle communication interface, parses the task processing request messages to obtain the target computing power unit baseline identifier, and extracts computing power acquisition threshold parameters based on the task processing request messages.

[0255] Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0256] Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value.

[0257] According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the vehicle communication interface to instruct the computing power request node to generate a target scheduling instruction.

[0258] like Figure 6 As shown, this is an embodiment of a vehicle-mounted computing power real-time price index calculation system, which includes a target vehicle node, a computing power request node, a price index calculation node, and a preset node cluster.

[0259] The target vehicle node is used to send the hardware specifications and available computing power of the on-board processor to the price index calculation node.

[0260] The computing power request node is used to send a task processing request message to the price index calculation node, and receive the benchmark price value broadcast by the price index calculation node or receive the benchmark price value forwarded by the preset node cluster, so as to generate a target scheduling instruction.

[0261] The price index calculation node is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area, query the preset computing power conversion mapping table based on the hardware specifications, extract the benchmark conversion coefficient corresponding to the hardware specifications, and then use the benchmark conversion coefficient to convert the available computing power of each target vehicle node into a dual benchmark standard computing power unit sequence.

[0262] The price index calculation node is also used to receive the task processing request message sent by the computing power request node, parse the task processing request message to obtain the target computing power unit benchmark identifier, and extract the computing power acquisition threshold parameter based on the task processing request message.

[0263] The price index calculation node is also used to retrieve the set of expected computing power supply parameters that match the target computing power unit benchmark identifier in the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence.

[0264] The price index calculation node is also used to extract the sequence node values ​​in the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, the convergence value of the final iteration output is extracted as the benchmark price value.

[0265] The price index calculation node is also used to broadcast the benchmark price value to the preset node cluster via a network interface according to a preset adaptive refresh cycle.

[0266] The preset node cluster is used to receive the benchmark price value broadcast by the price index calculation node, and send the benchmark price value to the computing power request node to instruct the computing power request node to generate the target scheduling instruction.

[0267] The above Figures 3-6 For an exemplary description, please refer to the above. Figure 1 This will not be elaborated upon here.

Claims

1. A method for calculating the real-time price index of vehicle-mounted computing power, characterized in that, Includes the following steps: The hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area are obtained. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients. The system receives a task processing request message sent by a computing power request node, parses the task processing request message to obtain the target computing power unit baseline identifier, and extracts the computing power acquisition threshold parameter based on the task processing request message. Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence. Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value. According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the network interface to instruct the computing power request node to generate a target scheduling instruction.

2. The on-vehicle computing power real-time price index calculation method according to claim 1, wherein The step of converting the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficient includes: Extract the integer precision computing power index of the target vehicle node, multiply the integer precision computing power index using the first coefficient in the benchmark conversion coefficient to obtain the integer precision standard computing power unit value, and establish a correspondence between the integer precision standard computing power unit value and the integer precision standard computing power unit benchmark identifier; Extract the floating-point precision computing power index of the target vehicle node, multiply the floating-point precision computing power index using the second coefficient in the reference conversion coefficient to obtain the floating-point precision standard computing power unit value, and establish a correspondence between the floating-point precision standard computing power unit value and the floating-point precision standard computing power unit reference identifier. The integer precision standard computing power unit value, the integer precision standard computing power unit reference identifier, the floating point precision standard computing power unit value, and the floating point precision standard computing power unit reference identifier are all written into the dual reference standard computing power unit sequence.

3. The on-vehicle computing power real-time price index calculation method according to claim 1, wherein The step of sequentially performing bidirectional data constraint iteration operations on the computing power threshold parameter using the sequence node values ​​includes: Set the computing power acquisition threshold parameter as the first constraint condition, and set the current sequence node value in the sequence node values ​​as the second constraint condition; In a single data loop processing process, the first constraint is applied to the supply ladder distribution sequence to filter out a subset of available nodes that meet the first constraint, and the corresponding first node response parameter is calculated based on the subset of available nodes. Apply the second constraint to the first node response parameter to calculate the corresponding second node request parameter; The response parameters of the first node are compared with the request parameters of the second node, and the deviation value of the comparison output is used as the iteration index of the current data loop processing process, and the iteration index is used as the basis for determining whether the deviation value falls into the preset convergence interval. If the deviation value does not fall within the preset convergence interval, the next data loop processing process is triggered based on the value of the next sequence node in the supply ladder distribution sequence.

4. The on-vehicle computing power real-time price index calculation method according to claim 1, wherein Before the step of broadcasting the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh cycle, the method further includes: Collect the total number of available vehicle nodes in the preset parking physical area that are in a network connected state, and count the total number of pending task queues to which the task processing request message belongs; When the total number of available vehicle nodes is greater than the first preset node number threshold, or when the total number of pending tasks in the queue is greater than the first preset queue number threshold, the adaptive refresh cycle is set to the first cycle duration parameter. When the total number of available vehicle nodes is lower than the second preset node number threshold and the total number of pending tasks in the queue is lower than the second preset queue number threshold, the adaptive refresh cycle is set to the second cycle duration parameter. The time interval for broadcasting the benchmark price value to the preset node cluster via the network interface is controlled according to the first cycle duration parameter or the second cycle duration parameter.

5. The on-vehicle computing power real-time price index calculation method according to claim 1, wherein Before the step of obtaining the hardware specification characteristics of the on-board processors of the target vehicle nodes distributed within the preset parking physical area, the method further includes: Obtain the spatial location attribute information of the preset parking physical area, match the corresponding computing power supply stability coefficient based on the spatial location attribute information, and obtain the credibility weight coefficient corresponding to the preset parking physical area. The computing power supply stability coefficient is multiplied by the credibility weight coefficient to obtain the node collaboration comprehensive weight value of the preset stationed physical area; After obtaining the benchmark price value, the node collaborative comprehensive weight value is multiplied by the benchmark price value to obtain the regional weighted price adjustment parameter; The regional weighted price adjustment parameter is used as an alternative broadcast result after regional weighting of the benchmark price value, and the regional weighted price adjustment parameter is broadcast to the preset node cluster through the network interface.

6. The on-vehicle computing power real-time price index calculation method according to claim 5, wherein The step of obtaining the spatial location attribute information of the preset parking physical area includes: Collect the geographic coordinate point dataset of the preset parking physical area, input the geographic coordinate point dataset into the preset point of interest location parsing program to output the location classification label, and use the location classification label as the spatial location attribute information; In the absence of the location classification label output by the point of interest location parsing program, the statistical data on vehicle dwell time distribution in the preset parking physical area within a historical set period is retrieved. Based on the statistical data on vehicle dwell time distribution, an alternative location classification label is determined, and the alternative location classification label is used to replace the original location classification label. The alternative location classification label is then used as the spatial location attribute information.

7. The on-vehicle computing power real-time price index calculation method according to claim 1, wherein In the step of parsing the task processing request message to obtain the target computing unit reference identifier, if the parsed and extracted target computing unit reference identifier contains mixed precision indication information, the method further includes: Extract the first execution stage data block and the second execution stage data block from the task processing request message; The first computational load corresponding to the data block in the first execution stage and the second computational load corresponding to the data block in the second execution stage are calculated respectively. A total computing load is generated based on the first computing load and the second computing load. A first percentage parameter of the first computing load in the total computing load is calculated based on the total computing load. A second percentage parameter of the second computing load in the total computing load is calculated based on the total computing load. The reference identifier for the standard computing power unit with integer precision is assigned according to the first proportion parameter, and the reference identifier for the standard computing power unit with floating-point precision is assigned according to the second proportion parameter. Based on the integer precision standard computing power unit reference identifier, the first proportion parameter, the floating point precision standard computing power unit reference identifier, and the second proportion parameter combination, a hybrid reference target computing power unit reference identifier is generated, and the hybrid reference target computing power unit reference identifier is used as the target computing power unit reference identifier.

8. The on-vehicle computing power real-time price index calculation method according to claim 3, wherein After the step of triggering the next data loop processing process based on the next sequence node value in the supply ladder distribution sequence, provided that the deviation value does not fall within the preset convergence interval, the method further includes: Record the cumulative number of times the data loop processing process is executed; Determine whether the cumulative execution count has reached the preset execution count limit parameter; Under abnormal triggering conditions where the cumulative execution count reaches the preset execution count limit parameter, the bidirectional data constraint iteration operation is stopped; Retrieve the benchmark price value broadcast in the previous period and extract the initial sequence node value in the supply ladder distribution sequence; A weighted average calculation is performed on the benchmark price value broadcast in the previous period and the initial sequence node value. The weighted average calculation result is then extracted as the compensation benchmark price value, and the compensation benchmark price value is used to replace the benchmark price value in the current period.

9. The on-vehicle computing power real-time price index calculation method according to claim 1, wherein, After the step of broadcasting the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh cycle, the method further includes: The benchmark price value, along with the corresponding timestamp information, is written into the historical sequence database; Extract the historical benchmark price value sequence within a preset backtracking time period from the historical sequence database, and perform a moving weighted average processing operation on the historical benchmark price value sequence to obtain the moving average parameter; Calculate the data distribution slope parameter corresponding to the supply ladder distribution sequence, and perform a fusion processing operation on the data distribution slope parameter and the moving average parameter to generate a prediction interval value; The predicted interval values ​​are sent to the computing power request node to instruct the computing power request node to plan the task access timing.

10. An in-vehicle computing power real-time price index calculation device, characterized by, include: The standard computing power conversion module is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area, query the preset computing power conversion mapping table based on the hardware specifications, extract the benchmark conversion coefficients corresponding to the hardware specifications, and then use the benchmark conversion coefficients to convert the available computing power of each target vehicle node into a dual-benchmark standard computing power unit sequence. The request message parsing module is used to receive the task processing request message sent by the computing power request node, parse the task processing request message to obtain the target computing power unit benchmark identifier, and extract the computing power acquisition threshold parameter based on the task processing request message. The supply ladder generation module is used to retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit in the dual benchmark standard computing power unit sequence, and perform an ascending sorting operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence. The bidirectional iterative pricing module is used to extract the sequence node values ​​in the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iterative operations on the computing power acquisition threshold parameter in sequence to obtain continuous iterative output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iterative operations falls within the preset convergence interval, the convergence value of the final iterative output is extracted as the benchmark price value. The benchmark price broadcasting module is used to broadcast the benchmark price value to a preset node cluster via a network interface according to a preset adaptive refresh cycle, so as to instruct the computing power requesting node to generate a target scheduling instruction.

11. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing a computer program, and the processor performs the following processing when running the computer program: The hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area are obtained. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients. The system receives a task processing request message sent by a computing power request node, parses the task processing request message to obtain the target computing power unit baseline identifier, and extracts the computing power acquisition threshold parameter based on the task processing request message. Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence. Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value. According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the network interface to instruct the computing power request node to generate a target scheduling instruction.

12. A vehicle, characterized in that, The vehicle includes an on-board processor, an on-board communication interface, and an on-board memory. The on-board memory stores a real-time price index calculation program for on-board computing power. When the on-board processor runs the real-time price index calculation program for on-board computing power, it performs the following processing: The vehicle communication interface is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area. Based on the hardware specifications, a preset computing power conversion mapping table is queried to extract the benchmark conversion coefficients corresponding to the hardware specifications. Then, the available computing power of each target vehicle node is converted into a dual-benchmark standard computing power unit sequence using the benchmark conversion coefficients. The system receives task processing request messages sent by computing power request nodes through the vehicle communication interface, parses the task processing request messages to obtain the target computing power unit baseline identifier, and extracts computing power acquisition threshold parameters based on the task processing request messages. Retrieve the set of expected computing power supply parameters that match the benchmark identifier of the target computing power unit from the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence. Extract the sequence node values ​​from the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, extract the convergence value of the final iteration output as the benchmark price value. According to the preset adaptive refresh cycle, the benchmark price value is broadcast to the preset node cluster through the vehicle communication interface to instruct the computing power request node to generate a target scheduling instruction.

13. A vehicle-mounted computing power real-time price index calculation system, characterized in that, This includes target vehicle nodes, computing power request nodes, price index calculation nodes, and a pre-defined node cluster; The target vehicle node is used to send the hardware specifications and available computing power of the on-board processor to the price index calculation node. The computing power request node is used to send a task processing request message to the price index calculation node, and receive the benchmark price value broadcast by the price index calculation node or receive the benchmark price value forwarded by the preset node cluster, so as to generate a target scheduling instruction. The price index calculation node is used to obtain the hardware specifications of the on-board processors of the target vehicle nodes distributed in the preset parking physical area, query the preset computing power conversion mapping table based on the hardware specifications, extract the benchmark conversion coefficient corresponding to the hardware specifications, and then use the benchmark conversion coefficient to convert the available computing power of each target vehicle node into a dual benchmark standard computing power unit sequence. The price index calculation node is also used to receive the task processing request message sent by the computing power request node, parse the task processing request message to obtain the target computing power unit benchmark identifier, and extract the computing power acquisition threshold parameter based on the task processing request message. The price index calculation node is also used to retrieve the set of expected computing power supply parameters that match the target computing power unit benchmark identifier in the dual benchmark standard computing power unit sequence, and perform an ascending sort operation on the set of expected computing power supply parameters to generate a corresponding supply ladder distribution sequence. The price index calculation node is also used to extract the sequence node values ​​in the supply ladder distribution sequence, and use the sequence node values ​​to perform bidirectional data constraint iteration operations on the computing power acquisition threshold parameter in sequence to obtain continuous iteration output results. Under the trigger condition that the difference between the output results of two consecutive executions of the bidirectional data constraint iteration operations falls within the preset convergence interval, the convergence value of the final iteration output is extracted as the benchmark price value. The price index calculation node is also used to broadcast the benchmark price value to the preset node cluster via a network interface according to a preset adaptive refresh cycle. The preset node cluster is used to receive the benchmark price value broadcast by the price index calculation node, and send the benchmark price value to the computing power request node to instruct the computing power request node to generate the target scheduling instruction.

Images