Data processing method and device, electronic equipment, system and storage medium

Abstract

The embodiment of the invention provides a data processing method and device, electronic equipment, a system and a storage medium. The method comprises the following steps: receiving acquisition data streams generated by roadside sensing equipment in a sensing area; based on the transmission time delay between each roadside sensing device and each edge cloud node in the edge cloud platform, obtaining a first edge cloud node satisfying a first transmission condition, the first transmission condition comprising the lowest transmission time delay; distributing the collected data flow to a plurality of computing power units under the first edge cloud node for data preprocessing; performing data fusion on the preprocessed data to form target fusion data; and transmitting the target fusion data to an auxiliary decision making system of the road user. Before data transmission, the node with the lowest transmission delay is selected as the target node for data processing, so that the delay caused by data transmission is reduced, preprocessing and data fusion are completed at the edge cloud nodes, multiple independent and dispersed data can be integrated, and the data volume is reduced.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a data processing method and device, electronic equipment, system and storage medium. Background Art

[0002] In the scenario of road data collection and processing, various roadside sensing devices (such as radars and cameras) send the collected data to the central cloud platform for unified data processing. Specifically, the central cloud platform time-aligns the data from different sensing units and fuses the time-aligned data together to form more accurate environmental information.

[0003] The roadside sensing equipment sends the collected data directly to the central cloud platform. The transmission link is long and takes a lot of time, resulting in a large data transmission delay when the data is sent from the roadside to the central cloud platform, reducing the timeliness of the provision of environmental information. Road users (such as vehicles) cannot obtain environmental information in a timely manner, affecting the road user experience. Summary of the invention

[0004] The embodiments of the present application provide a data processing method and device, an electronic device, a system and a storage medium. Through this method, the collected data of the roadside sensing device can be sent to the edge cloud node where the transmission time between the roadside sensing devices is the lowest, and the data preprocessing and data fusion are performed through the edge cloud node. The data amount is reduced through data fusion, and the fused data is then sent to the road user's auxiliary decision-making system through the edge cloud platform computing device, thereby reducing the data transmission delay and improving the data processing efficiency.

[0005] In the first aspect, the embodiment of the present application provides a data processing method, including: receiving the collected data stream generated by each roadside sensing device in the sensing area; based on the transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform, obtaining a first edge cloud node that meets the first transmission condition, wherein the first transmission condition includes the lowest transmission delay; allocating the collected data stream to multiple computing units under the first edge cloud node for data preprocessing; fusing the preprocessed data to form target fusion data; and transmitting the target fusion data to the auxiliary decision system of the road user. Through this scheme, on the one hand, before data transmission, by selecting the node with the lowest transmission delay as the target node for data processing, that is, the first edge cloud node, the delay caused by data transmission is reduced, and then the data preprocessing and data fusion are completed at the first edge cloud node after the collected data stream is obtained, replacing the scheme that takes a long transmission time to transmit to the central cloud platform for data processing, thereby reducing the time required for data transmission; on the other hand, completing the preprocessing and data fusion at the edge cloud node can integrate multiple independent and scattered data, reduce the amount of data, and can be transmitted once when performing target fusion data transmission, further reducing the data transmission delay and improving data processing efficiency.

[0006] In some possible implementations, the acquisition of the first edge cloud node that meets the first transmission condition based on the transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform includes: pre-acquiring the transmission data volume, network bandwidth and network delay of each roadside sensing device in the sensing area, wherein the network delay is the network delay between the roadside sensing device and the edge cloud node; calculating the ratio of the transmission data volume to the network bandwidth, and summing it with the network delay to obtain the transmission delay; configuring the node with the lowest transmission delay with the current roadside sensing device among multiple edge cloud nodes as the first edge cloud node. By knowing the transmission delay between each roadside sensing device and each edge cloud node in advance, when selecting a node to perform data processing (i.e., preprocessing and data fusion), the node with the lowest delay can be selected based on user needs to perform data processing, thereby improving data processing efficiency by reducing data transmission delay.

[0007] In some possible implementations, the roadside sensing device includes a camera and / or a radar; wherein each of the collected data streams generated by the camera or the radar is configured as an independent computing task; the allocation of the collected data streams to multiple computing units under the first edge cloud node for data preprocessing includes: allocating multiple computing tasks to multiple computing units under the first edge cloud node according to a load balancing mechanism; the computing unit extracts features based on the received computing tasks to obtain first feature data. The efficiency of task completion is improved by performing feature extraction in parallel by multiple computing units under the edge cloud node, and the computing efficiency and response speed are improved by load balancing of multiple computing units during the task allocation stage.

[0008] In some possible implementations, the allocation of multiple computing tasks to multiple computing units under the first edge cloud node according to the load balancing mechanism includes: when the load capacity of the first edge cloud node is less than the computing power required for the multiple computing tasks, in the order of the transmission delay from low to high, according to the load balancing mechanism, some computing tasks are allocated to multiple computing units under the second edge cloud node for data preprocessing until the multiple computing tasks are allocated, wherein the transmission delay of the second edge cloud node is higher than the transmission delay of the first edge cloud node. Insufficient computing power of the current node may lead to increased processing delay. In order to reduce the impact of increased delay, some tasks are allocated to other nodes with similar transmission delays for data preprocessing, and tasks are also allocated according to the load balancing mechanism, thereby improving computing efficiency and ensuring that the overall data processing time does not exceed the preset time requirements.

[0009] In some possible implementations, after the multiple computing tasks are assigned, the method further includes: when a computing unit to which the computing tasks have been assigned fails and cannot provide computing power, assigning the corresponding computing tasks to be reallocated to an idle computing unit in the current edge cloud node. By reallocating the computing tasks of the failed computing unit to the idle computing unit in the current node, the completion rate of the computing tasks is ensured.

[0010] In some possible implementations, data fusion is performed on the preprocessed data to form target fusion data, including: aggregating all first feature data related to the perception area to the first edge cloud node and performing time alignment; performing weighted average fusion on all first feature data that have been time aligned to form the target fusion data. When multiple nodes perform computing tasks related to the same perception area, the first feature data extracted by the multiple nodes can be aggregated to the first edge cloud node with the lowest transmission delay to the roadside perception device, and after all the first feature data are time aligned and weighted average fused by the first edge cloud node to obtain the target fusion data, the target fusion data is transmitted to the auxiliary decision system of the road user using the shortest transmission time, that is, the data transmission delay can be further reduced in the process of information feedback.

[0011] In some possible implementations, after the pre-processed data is fused to form the target fused data, it also includes: calculating the cycle duration of the target fused data, wherein the cycle duration includes the sum of the data transmission time, the data pre-processing time and the data fusion time; calculating the system throughput, wherein the system throughput includes the processing task completion rate per unit time; and / or calculating the resource utilization of each edge cloud node, wherein the resource utilization is the ratio of the actual used resources to the total resources of the node. The system performance is evaluated through the three dimensions of the cycle duration of the target fused data, the system throughput and the node resource utilization, and then adaptive improvements can be made according to the above evaluation results, so as to further meet the needs of users and improve the user experience.

[0012] In a second aspect, an embodiment of the present application further provides a data processing device, comprising: a processor and a memory, wherein the memory is used to store at least one instruction, and when the instruction is loaded and executed by the processor, the data processing method provided in the first aspect is implemented.

[0013] In the third aspect, an embodiment of the present application also provides a data processing system, which includes: multiple roadside sensing devices and an edge cloud platform; the multiple roadside sensing devices are used to collect road information in the sensing area and generate a collected data stream; the edge cloud platform includes: a data buffer unit, which is used to temporarily store the collected data stream generated by the multiple roadside sensing devices in the sensing area; a task allocation unit, which is used to obtain a first edge cloud node that meets a first transmission condition based on the transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform, wherein the first transmission condition includes a minimum transmission delay; multiple edge cloud nodes, each of which includes multiple computing power units, wherein each edge cloud node has its own data buffer unit; wherein the multiple computing power units under the first edge cloud node are used to perform data preprocessing on the received collected data stream, perform data fusion on the preprocessed data to form target fusion data, and transmit the target fusion data to the environmental information system.

[0014] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the data processing method provided in the first aspect.

[0015] In a fifth aspect, an embodiment of the present application further provides a computer program product, including a computer program or instructions, which, when executed by a processor, implements the data processing method provided in the first aspect.

[0016] Through the above technical scheme, before data transmission, by selecting the node with the lowest transmission delay as the target node for data processing, that is, the first edge cloud node, the delay caused by data transmission is reduced, and then the data preprocessing and data fusion are completed at the first edge cloud node after the collected data stream is obtained, which replaces the scheme of transmitting data to the central cloud platform for processing which takes a long time, thereby reducing the time required for data transmission; on the other hand, completing the preprocessing and data fusion at the edge cloud node can integrate multiple independent and scattered data, reduce the amount of data, and when performing target fusion data transmission, it can be done through a single data transmission, which further reduces the data transmission delay and improves the data processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0018] Figure 1 A schematic diagram of an application scenario architecture provided for an embodiment of the present application;

[0019] Figure 2 A schematic diagram of a data processing method flow chart provided for one embodiment of the present application;

[0020] Figure 3 A schematic diagram of the structure of an edge cloud platform is provided for one embodiment of the present application;

[0021] Figure 4 A schematic diagram of a computing task allocation execution flow provided for an embodiment of the present application;

[0022] Figure 5 A schematic diagram of the structure of a data processing device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] Figure 1 A schematic diagram of an application scenario architecture provided for an embodiment of the present application.

[0025] Reference Figure 1 As shown, the road information collection scenario may include a roadside perception device 10 and an edge cloud platform 20.

[0026] In some embodiments, a variety of roadside perception devices 10 installed on both sides of the road or at traffic intersections will continuously collect a large amount of data. The various roadside perception devices 10 may include a plurality of cameras 10a and a plurality of radars 10b, wherein the cameras 10a may perform video monitoring of the road and generate a large amount of video stream data; the radars 10b may perform three-dimensional detection of the road and generate a large amount of point cloud stream data.

[0027] In some embodiments, a large amount of data collected by the roadside sensing device 10 needs to be processed in real time, and the data collected by the roadside sensing device can be processed by the edge cloud platform 20, wherein the edge cloud platform 20 may include multiple edge cloud nodes, and multiple edge cloud nodes with the shortest or less than a certain time period between the roadside sensing device and the edge cloud node can be selected to process the collected data, thereby reducing the data transmission delay and improving the efficiency of data processing. Then, the data processed by the edge cloud platform 20 is uploaded to the auxiliary decision system of the road user, so that the vehicles on the road can obtain the environmental information of the relevant location in time through the auxiliary decision system, and ensure the timeliness and accuracy of the vehicle driving auxiliary decision.

[0028] The present application also provides a data processing method, which can be applied to Figure 1 The edge cloud platform 20 in the illustrated architecture can cache the collected data on the edge cloud platform 20 side for data processing after the roadside sensing device 10 completes data collection. That is, the collected data is processed through edge cloud platform calculation, which reduces the time consumed in data processing, thereby reducing data transmission delay and improving data processing efficiency.

[0029] The data processing method provided by the present application is described in detail below with reference to the accompanying drawings.

[0030] Figure 2 A flow chart of a data processing method provided for one embodiment of the present application.

[0031] Reference Figure 2 As shown, the method may include the following steps:

[0032] S201: The edge cloud platform obtains a first edge cloud node that meets a first transmission condition based on the transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform, wherein the first transmission condition includes the lowest transmission delay.

[0033] In some embodiments, the collected data streams generated by each roadside perception device in the perception area can be temporarily stored in the edge cloud platform, and the data streams are collected by the edge cloud nodes in the edge cloud platform for data processing.

[0034] A sensing area may be provided with multiple cameras and multiple radars. In some application scenarios, the sensing area may be a traffic location such as a traffic intersection or a middle section of a road. This application does not limit the specific location of the sensing area.

[0035] In some embodiments, information about the amount of data generated per second by each camera and each radar in the perception area may be acquired in advance.

[0036] Among them, A cameras can be set in a perception area, and the video bit rate of each camera is f (kbps). Then the calculation formula for the video stream data volume V generated by A cameras per second in the area corresponding to the current MEC is: V = (f / 8) * A. For example, there are 10 720P cameras in the area corresponding to an MEC, and the video bit rate of each camera is 2Mbps (i.e. 2000kbps). The video stream data generated by the 10 720P cameras per second is: V = (2000 / 8) * 10, and V = 2500kbps (i.e. 2.5Mbps) is calculated.

[0037] Among them, B radars are set in the area corresponding to one MEC, and each radar generates P MB of point cloud stream data per second. Exemplarily, the point frequency of the radar is 1.2MHZ, that is, it can scan 1200000 points per second, and the amount of data for each point needs to be determined according to the actual data format and the information contained. If the amount of data for each point is a 32-bit floating point number to represent the X, Y, and Z coordinates, plus an 8-bit byte of reflection intensity, then each point requires at least (32+8) / 8=5.5 bytes, so the amount of data per second is 1200000 points multiplied by 5.5 bytes, which is equal to about 6600KB≈6.6Mb / s. That is, 10 1.2MHZ radars generate 66Mb of data per second.

[0038] In some embodiments, in order to reduce the transmission delay of the collected data stream transmitted by the roadside sensing device, the collected data stream can be transmitted to the edge cloud platform for data processing. The transmission link between the roadside sensing device and the edge cloud platform is shorter than the transmission link between the roadside sensing device and the central cloud platform, thereby reducing the transmission delay.

[0039] In some embodiments, the edge cloud platform may include multiple edge cloud nodes. In some embodiments, the edge cloud nodes included in the edge cloud platform may be specifically a multi-access edge cloud computing server (MEC), hereinafter referred to as MEC. To further reduce the data transmission delay, a node with the lowest transmission delay among multiple edge cloud nodes may be selected to receive the collected data stream transmitted by the roadside sensing device and perform data processing.

[0040] Figure 3 A structural diagram of an edge cloud platform is provided for one embodiment of the present application.

[0041] Reference Figure 3As shown, the edge cloud platform 20 may include multiple edge cloud nodes 20a and a task allocation unit 20b. Each edge cloud node 20a includes a data buffer unit, and the collected data stream generated by each roadside sensing device 10 in the sensing area can be transmitted through the network to the data buffer unit in the selected edge cloud node 20a with the lowest data transmission delay, and then the edge cloud node can perform data processing on the received collected data stream. The data buffer unit is used to temporarily store the data stream to avoid blockage during the processing and ensure that data is not lost. The storage time of the data buffer unit can be calculated by the following formula: storage time = data buffer unit capacity / data flow. For example, if the buffer capacity is 500MB and the data flow per second is 60MB, the storage time of the data buffer unit is 500 / 60≈8.33 seconds.

[0042] In some embodiments, the edge cloud platform 20 can obtain a first edge cloud node that meets the first transmission condition based on the transmission delay between each roadside perception device and each edge cloud node in the edge cloud platform, and the first transmission condition includes the lowest transmission delay. Specifically, the data transmission time T required for each roadside perception device (multiple cameras and multiple radars) to transmit the generated data stream (video stream data and point cloud stream data) to the data buffer 20a of each edge cloud node through a network connection can be obtained in advance.

[0043] In some embodiments, after obtaining information on the amount of data generated per second by each camera and each radar in the perception area, the data transmission time T required for each camera and each radar to send the generated collected data stream to the edge cloud platform 20 can also be pre-calculated based on the information on the amount of data generated per second.

[0044] In some embodiments, the data transmission time T can be determined based on the amount of transmitted data V0, the network bandwidth B, and the transmission delay t. Specifically, the corresponding calculation formula may include: T = (V0 / B) + t. For example, if the data volume of the video stream generated by a camera is 60MB, the network bandwidth is 1000MB / s, and the transmission delay between the camera and the edge cloud platform 20 is 10 milliseconds, then the data transmission time is 60 / 1000+0.01=0.07 seconds.

[0045] In some embodiments, the data transmission time (i.e., transmission delay) between each roadside sensing device and each edge cloud node in the sensing area can be calculated by the above calculation method. For example, the edge cloud platform includes three edge cloud nodes, namely MEC node A, MEC node B, and MEC node C. The transmission delays between multiple roadside sensing devices in the sensing area and the three edge cloud nodes (MEC nodes) are calculated by transmission delay as follows:

[0046] Transmission delay with MEC node A = 15ms;

[0047] Transmission delay to MEC Node B = 25ms;

[0048] Transmission delay with MEC node C = 20ms.

[0049] By comparing the transmission delays, the node with the lowest transmission delay is MEC node A, so it can be determined that the first edge cloud node that meets the first transmission condition is MEC node A.

[0050] In some embodiments, after obtaining the data delay between each roadside sensing device and each edge cloud node, the task allocation unit 20b of the edge cloud platform can perform data caching and computing task allocation based on the transmission delay.

[0051] S202: The first edge cloud node receives a collection data stream generated by each roadside sensing device in the sensing area.

[0052] In some embodiments, the video stream data generated by each camera in the perception area and the point cloud stream data generated by each radar can be transmitted to the data buffer unit of the first edge cloud node through the network.

[0053] S203: The task allocation unit allocates the collected data stream to multiple computing units under the first edge cloud node for data preprocessing.

[0054] In some embodiments, the collected data stream of each camera in the perception area and the collected data stream generated by each radar are configured as independent computing tasks. The task allocation unit can allocate multiple computing tasks to multiple computing units under the first edge cloud node according to the load balancing mechanism. After receiving the computing tasks, the multiple computing units under the first edge cloud node can perform feature extraction through multiple computing tasks in a parallel processing manner and obtain the first feature data.

[0055] In some embodiments, within the selected first edge cloud node, computing tasks can be assigned to multiple computing units under the first edge cloud node according to the load conditions of the computing units. Furthermore, the load of the computing units that receive computing tasks can be monitored to ensure optimal utilization of computing resources. The load condition can be evaluated by the following formula:

[0056] Load = (Video feature data volume + Point cloud feature data volume) / Computing power

[0057] The calculation formula of load balancing index (LBI) is:

[0058] LBI=(Σ load_i) / N

[0059] Where N is the number of computing units in the node, and load_i is the load of the i-th computing unit. Monitor the load balancing indicators in real time to ensure the load balance of the computing units to improve computing efficiency and system response speed.

[0060] S204: Fusing the preprocessed data to form target fused data.

[0061] In some implementations, after multiple computing units under the first edge cloud node complete feature extraction and obtain first feature data, the first edge cloud node can also time-align all first feature data related to the perception area, and perform weighted average fusion of all time-aligned first feature data to form target fusion data.

[0062] In some embodiments, the method of time-aligning all first feature data may include synchronizing the timestamps of all first feature data by adjustment. Specifically, linear interpolation may be used to align data with different timestamps. The formula is: synchronized timestamp = (timestamp 1 + timestamp 2 + ... + timestamp n) / n.

[0063] In some embodiments, a specific method of weighted average fusion may include fusing all first feature data using the following fusion formula:

[0064] Fusion feature = (weight 1*feature 1+weight 2*feature 2+...+weight n*feature n) / (weight 1+weight 2+...+weight n).

[0065] In some embodiments, after the fusion is completed, the consistency and accuracy of the fused feature data can also be checked to ensure that there is no loss or conflict. The consistency of the data is ensured by comparing feature data from different sources. In one embodiment, the formula for checking data consistency is: consistency detection value = absolute value (fusion feature 1-fusion feature 2).

[0066] S205: Transmit the target fusion data to the auxiliary decision-making system of the road user.

[0067] In some embodiments, after the target fusion data is acquired, the target fusion data may be transmitted to an auxiliary decision system, so that road users (such as vehicles and terminals on the road) may obtain environmental information on the road in a timely manner through the auxiliary decision system.

[0068] Figure 4 A schematic diagram of the computing task allocation execution flow provided for one embodiment of the present application.

[0069] Reference Figure 4 As shown, the computing task allocation execution process may include the following steps:

[0070] S401: Based on the transmission delay between each roadside sensing device and each edge cloud node, determine the first edge cloud node with the lowest transmission delay.

[0071] In some embodiments, the specific implementation of S401 can be Figure 2 The implementation method of S201 in the illustrated embodiment is the same or similar and will not be described in detail.

[0072] S402: Determine whether the computing power of the first edge cloud node meets the computing power requirements of multiple computing tasks.

[0073] In some embodiments, in some application scenarios, the amount of data in the collected data stream generated by the roadside sensing device may be large, resulting in the computing power provided by multiple computing power units under an edge cloud node being unable to meet its computing power requirements. Therefore, after determining the first edge cloud node and before allocating computing tasks, it is possible to determine whether the computing power of the first edge cloud node meets the computing power requirements of multiple computing tasks.

[0074] In some embodiments, the feature extraction task of each video stream and point cloud stream is regarded as an independent computing task. These tasks need to be allocated to the computing units of the MEC nodes according to the computing power and load. The computing time of feature extraction depends on the complexity of the task. The calculation formula is as follows:

[0075] Video feature extraction time = YOLO (Si)

[0076] Point cloud feature extraction time = PointNet (Pj)

[0077] Where Si is the data volume of the i-th video stream, and Pj is the data volume of the j-th point cloud stream. The computing power of each MEC node is represented by R_k, the required computing power for the video stream feature extraction task is W_vi, and the required computing power for the point cloud stream feature extraction task is W_pj. To ensure whether the computing power of the first edge cloud node meets the computing power requirements of multiple computing tasks, it can be determined by the following formula:

[0078] R_k≥W_vi+W_pj

[0079] Through the above computing power detection, if it is determined that the computing power of the first edge cloud node meets the task computing power requirement, execute S403.

[0080] S403: After allocating multiple computing tasks to multiple computing units under the first edge cloud node for data preprocessing, execute S408.

[0081] In some embodiments, the specific implementation of S403 can be Figure 2The implementation method of S203 in the illustrated embodiment is the same or similar and will not be described in detail here.

[0082] S408: The first edge cloud node performs data fusion.

[0083] In some embodiments, the specific implementation of S408 can be Figure 2 The implementation method of S204 in the illustrated embodiment is the same or similar and will not be described in detail here.

[0084] Through the above computing power detection, if it is determined that the computing power of the first edge cloud node cannot meet the computing power requirements of the task, execute S404.

[0085] S404: Allocate computing tasks according to the maximum computing power of the first edge cloud node.

[0086] In some embodiments, when the computing power of the first edge cloud node cannot meet the computing power requirements of multiple computing tasks, the computing tasks can be allocated to the first edge cloud node according to the maximum computing power of the first edge cloud node, and the load balancing indicators of each computing power unit under the first edge cloud node can be monitored in real time to ensure the load balancing of the computing unit, so as to improve the computing efficiency and system response speed. After receiving the computing task, the multiple computing power units under the first edge cloud node can perform feature extraction on the corresponding collected data stream and obtain the first feature data.

[0087] S405: Determine whether the computing power of the next adjacent edge cloud node meets the computing power required for the remaining computing tasks according to the order of transmission delay from low to high.

[0088] In some embodiments, when the computing power of the first edge cloud node cannot meet the computing power requirements of multiple computing tasks, part of the computing tasks are allocated to multiple computing power units under the second edge cloud node for data preprocessing until the multiple computing tasks are allocated, wherein the transmission delay of the second edge cloud node is higher than the transmission delay of the first edge cloud node. In a specific implementation method, the computing power of the next adjacent edge cloud node can be further determined by sorting the transmission delay from low to high to determine whether it meets the computing power required by the remaining computing tasks. For example,

[0089] Transmission delay with MEC node A = 15ms;

[0090] Transmission delay to MEC Node B = 25ms;

[0091] Transmission delay with MEC node C = 20ms.

[0092] By comparing the transmission delays, the node with the lowest transmission delay is MEC node A, so it can be determined that the first edge cloud node that meets the first transmission condition is MEC node A.

[0093] Based on the above transmission delay information, it can be further determined whether the computing power of MEC node C meets the computing power required by the remaining computing tasks.

[0094] Through computing power detection, if it is determined that the computing power of the selected edge cloud node (MEC node C) meets the computing power required for the remaining computing tasks, the currently selected edge cloud node (MEC node C) is determined as the second edge cloud node, and S406 is executed. If it does not meet the requirements, S407 is executed.

[0095] S406: Allocate the remaining computing tasks to the currently selected edge cloud node.

[0096] In some embodiments, when it is determined that the selected edge cloud node power meets the computing power required by the remaining computing tasks, the remaining computing tasks are allocated to multiple computing power units under the currently selected edge cloud node (i.e., the second edge cloud node) according to the load balancing mechanism, and the load balancing indicators of each computing power unit under the current edge cloud node are monitored in real time to ensure the load balancing of the computing units to improve computing efficiency and system response speed. After receiving the computing tasks, the multiple computing power units under the current edge cloud node can perform feature extraction on the corresponding collected data stream and obtain the first feature data.

[0097] S407: After allocating computing tasks according to the maximum computing power of the current edge cloud node, return to S405.

[0098] In some embodiments, when it is determined that the power of the selected edge cloud node cannot meet the computing power required for the remaining computing tasks, the remaining computing tasks can be allocated to the current edge cloud node according to the maximum computing power of the current edge cloud node, and the load balancing indicators of each computing power unit under the current edge cloud node can be monitored in real time to ensure the load balancing of the computing unit, so as to improve the computing efficiency and system response speed. After receiving the computing task, the multiple computing power units under the current edge cloud node can perform feature extraction on the corresponding collected data stream and obtain the first feature data. Among them, after completing the task allocation for the current edge cloud node, it can return to S405 to continue to sort the transmission delay from low to high, and allocate the remaining computing tasks to the next adjacent edge cloud node until the remaining computing tasks are allocated. After each selected edge cloud node completes the feature extraction and obtains the first feature data, execute S408.

[0099] In some embodiments, when multiple edge cloud nodes jointly execute multiple computing tasks corresponding to a perception area, after the multiple edge cloud nodes complete feature extraction and obtain first feature data, the first feature data extracted by the multiple edge cloud nodes can be transmitted to the first edge cloud node with the lowest transmission delay to the roadside perception device in the perception area, and the fusion of all first feature data can be completed through the first edge cloud node, wherein the specific fusion implementation method may be the same or similar to the implementation method in S204, and will not be repeated here.

[0100] After completing the feature extraction, the results processed by each MEC node need to be fused to provide a global and unified view. By transmitting the target fusion data to the auxiliary decision-making system, road users can obtain environmental information in a timely manner through the auxiliary decision-making system, that is, they can have a global understanding of the environmental information around road users and improve user experience.

[0101] In some embodiments, when a computing power unit to which computing tasks have been assigned fails and is unable to provide computing power, the corresponding computing tasks to be reallocated are assigned to the idle computing power units in the current edge cloud node, thereby ensuring the completion rate of computing tasks by reallocating the computing tasks of the failed computing power units to the idle computing power units in the current node.

[0102] In some embodiments, after the pre-processed data is fused to form the target fused data, it also includes: calculating the cycle duration of the target fused data, wherein the cycle duration includes the sum of the data transmission time, the data pre-processing time and the data fusion time; calculating the system throughput, wherein the system throughput includes the processing task completion rate per unit time; and / or calculating the resource utilization of each edge cloud node, wherein the resource utilization is the ratio of the actual used resources to the total resources of the node. The system performance is evaluated through the three dimensions of the cycle duration of the target fused data, the system throughput and the node resource utilization, and then adaptive improvements can be made according to the above evaluation results, so as to further meet the needs of users and improve the user experience.

[0103] In some embodiments, after completing data fusion, a transmission delay impact assessment can also be performed. Specifically, a comprehensive assessment can be performed on the delay of each MEC node to ensure that the selected nodes and the load balancing of the computing unit will not have a negative impact on task processing during transmission delay. Calculate the transmission delay between MEC nodes to confirm whether it will affect business execution. Exemplarily, the delay between the roadside sensing device and MEC node A is 15ms, the delay with node B is 25ms, and the delay with node C is 20ms. When the computing power of node A is insufficient, consider allocating part of the task to node B. At this time, it is necessary to calculate the additional delay in transmitting data from the roadside sensing device to node B, that is, (25ms-15ms=10ms). By detecting and evaluating the impact of the additional delay on the data delay, it can be ensured that the overall processing time of the task meets the system requirements.

[0104] Figure 5 A schematic diagram of the structure of a data processing device provided in an embodiment of the present application.

[0105] Reference Figure 5 As shown, the data processing device may include a processor 100 and a memory 200, and the memory 200 is used to store at least one instruction, and when the instruction is loaded and executed by the processor 100, the data processing method provided by any embodiment of the present application is implemented.

[0106] An embodiment of the present application also provides a data processing system, which includes: multiple roadside sensing devices and an edge cloud platform; the multiple roadside sensing devices are used to collect road information in the sensing area and generate a collected data stream; the edge cloud platform includes: a data buffer unit, which is used to temporarily store the collected data stream generated by the multiple roadside sensing devices in the sensing area; a task allocation unit, which is used to obtain a first edge cloud node that meets a first transmission condition based on the transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform, wherein the first transmission condition includes a minimum transmission delay; multiple edge cloud nodes, each of which includes multiple computing units, wherein each edge cloud node has its own data buffer unit; wherein the multiple computing units under the first edge cloud node are used to perform data preprocessing on the received collected data stream, perform data fusion on the preprocessed data to form target fusion data, and transmit the target fusion data to an environmental information system.

[0107] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the data processing method provided by any embodiment of the present application is implemented.

[0108] An embodiment of the present application also provides a computer program product, including a computer program or instructions, which implements the data processing method provided in any embodiment of the present application when the computer program or instructions are executed by a processor.

[0109] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0110] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0111] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0113] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (Processor) to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0114] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A data processing method, characterized in that: The method comprises: Based on the transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform, obtaining a first edge cloud node that meets a first transmission condition, wherein the first transmission condition includes a minimum transmission delay; The first edge cloud node receives a collection data stream generated by each roadside sensing device in the sensing area; Allocating the collected data stream to multiple computing units under the first edge cloud node for data preprocessing; Perform data fusion on the preprocessed data to form target fused data; The target fusion data is transmitted to an auxiliary decision-making system of the road user.

2. The method according to claim 1, characterized in that The acquiring, based on the transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform, a first edge cloud node that meets the first transmission condition comprises: Pre-acquire the transmission data volume, network bandwidth and network delay of each roadside sensing device in the sensing area, wherein the network delay is the network delay between the roadside sensing device and the edge cloud node; Calculating a ratio of the amount of transmitted data to the network bandwidth, and summing the ratio with the network delay to obtain a transmission delay; The node with the lowest transmission delay with the current roadside sensing device among multiple edge cloud nodes is configured as the first edge cloud node.

3. The method according to claim 1, characterized in that: The roadside sensing device includes a camera and / or a radar; Wherein, each acquisition data stream generated by the camera or the radar is configured as an independent computing task; The allocating the collected data stream to multiple computing units under the first edge cloud node for data preprocessing includes: Allocate multiple computing tasks to multiple computing units under the first edge cloud node according to a load balancing mechanism; The computing unit performs feature extraction based on the received computing task to obtain first feature data.

4. The method according to claim 3, characterized in that The allocating multiple computing tasks to multiple computing units under the first edge cloud node according to the load balancing mechanism includes: When the load capacity of the first edge cloud node is less than the computing power required for the multiple computing tasks, part of the computing tasks are allocated to multiple computing power units under the second edge cloud node in the order of the transmission delay from low to high according to the load balancing mechanism for data preprocessing until the multiple computing tasks are allocated, wherein the transmission delay of the second edge cloud node is higher than the transmission delay of the first edge cloud node.

5. The method according to claim 4, characterized in that After the plurality of computing tasks are assigned, the method further includes: When a computing unit to which computing tasks have been assigned fails and cannot provide computing power, the corresponding computing tasks to be reallocated are assigned to the idle computing units in the current edge cloud node.

6. The method according to claim 3, characterized in that The pre-processed data is fused to form target fusion data including: Gathering all first feature data related to the sensing area to the first edge cloud node and performing time alignment; All time-aligned first feature data are weighted averaged and fused to form the target fused data.

7. The method according to any one of claims 1 to 6, characterized in that: After fusing the pre-processed data to form target fused data, the method further includes: Calculate the cycle time of forming the target fusion data this time, wherein the cycle time includes the sum of data transmission time, data preprocessing time and data fusion time; Calculating system throughput, wherein the system throughput includes a processing task completion rate per unit time; and / or The resource utilization of each edge cloud node is calculated, where the resource utilization is the ratio of actually used resources to the total resources of the node.

8. A data processing device, characterized in that: The device comprises: A processor and a memory, wherein the memory is used to store at least one instruction, and when the instruction is loaded and executed by the processor, the data processing method according to any one of claims 1 to 7 is implemented.

9. A data processing system, characterized in that: The system includes: a plurality of roadside sensing devices and an edge cloud platform; The multiple roadside sensing devices are used to collect road information in the sensing area and generate a collected data stream; The edge cloud platform includes: A data buffer unit, used for temporarily storing the collected data streams generated by the plurality of roadside sensing devices in the sensing area; A task allocation unit, configured to obtain a first edge cloud node that meets a first transmission condition based on a transmission delay between each roadside sensing device and each edge cloud node in the edge cloud platform, wherein the first transmission condition includes a minimum transmission delay; A plurality of edge cloud nodes, each of the edge cloud nodes comprising a plurality of computing units, wherein each edge cloud node has its own data buffer unit; The multiple computing units under the first edge cloud node are used to perform data preprocessing on the received collected data stream, fuse the preprocessed data to form target fusion data, and transmit the target fusion data to the environmental information system.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the data processing method according to any one of claims 1 to 7 is implemented.

Images