Intelligent carrying equipment data interaction method and system based on edge cloud unloading

Abstract

The invention discloses an intelligent carrying equipment data interaction method and system based on edge cloud unloading, and relates to the technical field of data transmission, and the method comprises the steps: collecting multi-dimensional original data on carrying equipment at a logistics site; performing preprocessing and feature modeling on the data to obtain a feature vector of the carrying equipment; normalizing the feature vector to obtain a standardized feature vector; formulating a data unloading strategy according to the logistics information; uploading the standardized feature vector to a cloud according to a data unloading strategy; performing data processing on the standardized feature vector in the cloud computing center to complete logistics scheduling of the carrying equipment; and modifying the data unloading strategy according to the real-time operation state of the carrying equipment. According to the method, the data consistency is improved through data normalization, and heterogeneous data are uniformly accessed to the same system; a data processing position is determined by constructing a multi-target unloading decision model. Weighted fusion and an abnormal compensation mechanism are combined, high-credibility data are output, and the whole process is constrained through formulation to ensure that the system performance reaches the standard.

Description

Technical Field

[0001] This invention relates to the field of data transmission technology, and in particular to a data interaction method and system for intelligent transportation equipment based on edge-cloud offloading. Background Technology

[0002] With the intelligent upgrading of logistics parks, various transportation equipment such as trains, trucks, forklifts, unmanned vehicles, and drones have been widely used in scenarios such as trunk transportation, short-distance transshipment, and low-altitude inspection. The data streams generated by these devices, including location, speed, load, and image data, are the core support for park scheduling, route planning, and safety monitoring. Edge-cloud collaboration technology, as a key means to solve the contradiction between limited computing power on the device side and high latency in the cloud, has become the mainstream architectural direction for data interaction in logistics parks by combining edge node preprocessing with centralized cloud scheduling.

[0003] Current edge-cloud coordination and offloading solutions for multiple devices in logistics parks still face many challenges, such as: prominent data heterogeneity (different devices have vastly different data formats, dimensions, and real-time requirements, lacking a unified normalization method and making it difficult to integrate into the same processing system); crude offloading decisions (existing solutions only consider latency or energy consumption without considering multiple constraints such as edge computing power, communication bandwidth, and device distance, resulting in low resource utilization); insufficient protocol compatibility (devices use diverse native protocols, leading to low conversion efficiency and easy data loss); unbalanced edge load (a single edge node bears too many tasks, increasing processing latency and reducing system stability); and low data fusion accuracy (there is a lack of effective handling mechanisms when multiple devices have conflicting data from the same source, resulting in excessive errors in the fusion results and failing to support precise scheduling). Summary of the Invention

[0004] This invention provides a data interaction method and system for intelligent transportation equipment based on edge-cloud offloading, which solves the problems of prominent data heterogeneity, coarse offloading decision-making, insufficient protocol compatibility, edge load imbalance, and low data fusion accuracy in the prior art.

[0005] On one hand, embodiments of the present invention provide a data interaction method for intelligent transportation equipment based on edge-cloud offloading, including: Collect multi-dimensional raw data from the transportation equipment at the logistics site; The multi-dimensional raw data is preprocessed to obtain multi-dimensional data; The feature vector of the carrier equipment is obtained by performing feature modeling on the multi-dimensional data. The feature vectors are normalized to obtain standardized feature vectors; A data unloading strategy is formulated based on the logistics information of the transport equipment; The standardized feature vector is uploaded to the cloud computing center according to the data offloading strategy. The standardized feature vectors are processed in the cloud computing center to obtain a logistics scheduling plan. The logistics scheduling of the transport equipment is completed according to the logistics scheduling plan. The data unloading strategy is modified based on the real-time operating status of the transport equipment.

[0006] In one possible implementation, the multi-dimensional raw data includes trunk line transportation equipment data, short-distance transfer equipment data, and low-altitude inspection equipment data.

[0007] In one possible implementation, the preprocessing is data cleaning, data completion, and format standardization; The data cleaning includes outlier removal and noise filtering; The noise filtering involves applying moving average filtering to numerical data and median filtering to image data.

[0008] In one possible implementation, the feature modeling involves extracting the three-dimensional core features of the vehicle to obtain a feature vector; The three-dimensional core features include device status features, motion features, and environmental interaction features; The equipment status characteristics include health, load rate, and operational stability; The motion characteristics include instantaneous acceleration, trajectory deviation, and driving efficiency; The environmental interaction characteristics include device spacing, communication quality, and environmental risks.

[0009] In one possible implementation, the normalization involves mapping the feature vector to a unified interval using an adaptation method and then outputting a standardized feature vector.

[0010] In one possible implementation, the data unloading strategy is formulated based on the logistics information of the transport equipment by formulating the data unloading strategy according to real-time logistics information and data transmission parameters.

[0011] In one possible implementation, the objective function of the data offloading strategy is shown in the following equation: ; in, It is to minimize the optimization objective function. The total delay T is respectively total Total energy consumption E total Weighting coefficients; total delay T total Total energy consumption E total satisfy Constraints; in logistics scenarios, real-time performance is prioritized.

[0012] In one possible implementation, the data offloading strategy is divided by constraints, as shown in the following equation: ; Where st is subjectto, used to introduce the constraint conditions, S i，j X is the amount of data of type j of device i; i，j It is the unloading of decision variables, X i，j The variable X can take values ​​of 0 or 1. i，j =1 indicates data edge processing; X i，j =0 indicates cloud processing; i is the device type index, where i=1~5 correspond to: 1=forklift, 2=autonomous vehicle, 3=train, 4=truck, 5=drone; j is the data type index, where j=1~n corresponds to different data types of the device. This indicates that the summation is performed on the device type index i from 1 to 5. This indicates that the summation is performed on data type index j from 1 to n; It is the threshold for edge computing power utilization. B is the bandwidth utilization threshold, and B is the communication bandwidth. It is edge computing power.

[0013] On the other hand, embodiments of the present invention also provide a data interaction system for intelligent transportation equipment based on edge-cloud offloading, the system being used to implement the above-described method, including: The perception layer is used to collect multi-dimensional raw data on the transportation equipment at the logistics site; An edge layer is used to preprocess the multi-dimensional raw data to obtain multi-dimensional data; perform feature modeling on the multi-dimensional data to obtain the feature vector of the transport equipment; normalize the feature vector to obtain a standardized feature vector; formulate a data unloading strategy based on the logistics information of the transport equipment; and upload the standardized feature vector to the cloud computing center according to the data unloading strategy. The cloud layer is used to process the standardized feature vectors in the cloud computing center to obtain a logistics scheduling plan; to complete the logistics scheduling of the transport equipment according to the logistics scheduling plan; and to modify the data unloading strategy according to the real-time operating status of the transport equipment.

[0014] The data interaction method and system for intelligent transportation equipment based on edge-cloud offloading in this invention have the following advantages: (1) This application improves data consistency by normalizing data and integrates heterogeneous data into the same system; and determines the data processing location by constructing a multi-objective unloading decision model.

[0015] (2) Combining weighted fusion and anomaly compensation mechanisms, output highly reliable data, and ensure system performance meets standards through formulaic constraints throughout the process. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating the data interaction method for intelligent transportation equipment based on edge-cloud offloading provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of the intelligent transportation equipment data interaction system based on edge-cloud offloading provided in the embodiments of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Figure 1 A flowchart illustrating a data interaction method for intelligent transportation equipment based on edge-cloud offloading provided in an embodiment of the present invention; the embodiment of the present invention provides a data interaction method for intelligent transportation equipment based on edge-cloud offloading, including: Collect multi-dimensional raw data from the transportation equipment at the logistics site; The multi-dimensional raw data is preprocessed to obtain multi-dimensional data; The feature vector of the carrier equipment is obtained by performing feature modeling on the multi-dimensional data. The feature vectors are normalized to obtain standardized feature vectors; A data unloading strategy is formulated based on the logistics information of the transport equipment; The standardized feature vector is uploaded to the cloud computing center according to the data offloading strategy. The standardized feature vectors are processed in the cloud computing center to obtain a logistics scheduling plan. The logistics scheduling of the transport equipment is completed according to the logistics scheduling plan. The data unloading strategy is modified based on the real-time operating status of the transport equipment.

[0020] The multi-dimensional raw data includes trunk line transportation equipment data, short-distance transfer equipment data, and low-altitude inspection equipment data.

[0021] The preprocessing includes data cleaning, data completion, and format standardization. The data cleaning includes outlier removal and noise filtering; The noise filtering involves applying moving average filtering to numerical data and median filtering to image data.

[0022] The feature modeling involves extracting the three-dimensional core features of the vehicle to obtain a feature vector. The three-dimensional core features include device status features, motion features, and environmental interaction features; The equipment status characteristics include health, load rate, and operational stability; The motion characteristics include instantaneous acceleration, trajectory deviation, and driving efficiency; The environmental interaction characteristics include device spacing, communication quality, and environmental risks.

[0023] The normalization is to map the feature vector to a unified interval through an adaptation method and then output a standardized feature vector.

[0024] The data unloading strategy is formulated based on the logistics information of the transport equipment.

[0025] The objective function for optimizing the data offloading strategy is shown in the following equation: ; in, It is to minimize the optimization objective function. The total delay T is respectively total Total energy consumption E total Weighting coefficients; total delay T total Total energy consumption E total satisfy Constraints; in logistics scenarios, real-time performance is prioritized.

[0026] The data offloading strategy is divided by constraints, which are shown in the following formula: ; Where st is subjectto, used to introduce the constraint conditions, S i，j X is the amount of data of type j of device i; i，j It is the unloading of decision variables, X i，j The variable X can take values ​​of 0 or 1. i，j =1 indicates data edge processing; Xi，j =0 indicates cloud processing; i is the device type index, where i=1~5 correspond to: 1=forklift, 2=autonomous vehicle, 3=train, 4=truck, 5=drone; j is the data type index, where j=1~n corresponds to different data types of the device. This indicates that the summation is performed on the device type index i from 1 to 5. This indicates that the summation is performed on data type index j from 1 to n; It is the threshold for edge computing power utilization. B is the bandwidth utilization threshold, and B is the communication bandwidth. It is edge computing power.

[0027] For example, the entire solution is divided into three layers: the perception layer, the local layer, and the cloud layer.

[0028] The perception layer is used for data acquisition and feature modeling. It collects raw data from multiple devices and extracts three core features: real-time performance, data volume, and reliability, providing a foundation for subsequent processing.

[0029] First, feature vectors of the transportation equipment are obtained through feature modeling; the modeling formula is as follows: ; in, (Average real-time threshold for all devices). (Average data volume across all devices); S i，j T is the amount of data of type j of type i devices. i,j It is the real-time threshold for the j-th type of data from the i-th type of device. is the initial credibility weight of data of type j for type i device (determined by device hardware precision and data reliability), i is the device type index (i=1~5 correspond to: 1=forklift, 2=unmanned vehicle, 3=train, 4=truck, 5=drone). j is the data type index (j=1~n correspond to different data types of the device, such as location, speed, weight, etc.).

[0030] The output three-dimensional feature vector F i,j It can intuitively reflect the "urgency" of the data (T). i,j / T is smaller, indicating urgency), "processing cost" (S i,j The larger the / S, the higher the cost), and the "credibility" (ω) i,j (The larger the size, the more reliable it is), providing a basis for offloading decisions and data fusion.

[0031] Then, the data is normalized to eliminate the differences in units and formats of heterogeneous data, converting it into standardized data within a unified range to adapt to the same system processing. The normalization of numerical data is achieved through an improved min-max normalization formula, which is as follows: ; in, These are the two extreme values ​​of all devices for the j-th type of data (e.g., 0~100km / h for all devices' "speed data"), with U=10 and L=0 representing a unified output range; It is the normalized result (standardized data) of the j-th type of data for the i-th type of device. It is the raw data of the j-th type of the i-th type of device (such as position, speed, weight, etc.); It is the maximum value of all devices for the j-th type of data (such as the maximum value of "speed data" for all devices). is the minimum value of all devices for the j-th type of data (such as the minimum value of "speed data" for all devices), U is the upper limit of the normalized unified output range (fixed value U=10), and L is the lower limit of the normalized unified output range (fixed value L=0). The formula for special image data processing (applicable to UAV inspection images) is as follows: ; Pixel values ​​are normalized to the 0-1 range to align with the dimensions of numerical data; where... It is the normalized pixel data of the drone inspection image. This refers to the raw image data (pixel values) from the drone inspection; 255 is the upper limit of pixel values ​​for a digital image under normal conditions. For example, forklift position D... 1,1 =30m (D) j=1,min=0m D j=1,max=100m After normalization This eliminates the dimensional differences in data from different devices, ensuring the effectiveness of subsequent fusion and processing.

[0032] At the edge layer, the optimization of edge-cloud decision-making is to minimize total latency and energy consumption while satisfying edge computing power and communication bandwidth constraints, and to determine the data processing location (edge / cloud).

[0033] The objective function of the data offloading strategy is shown in the following equation: ; in, It is to minimize the optimization objective function. The total delay T is respectively total Total energy consumption E total Weighting coefficients; total delay T total Total energy consumption Etotal satisfy Constraints; in logistics scenarios, real-time performance is prioritized.

[0034] The data offloading strategy is divided by constraints, which are shown in the following formula: ; Among them, S i，j X is the data volume (in KB) of the j-th type of data for the i-th type of device; i，j It is unloading decision variables (X) i，j =1 indicates that the data is processed at the edge; X i，j =0 indicates cloud processing), i is the device type index (i=1~5 correspond to: 1=forklift, 2=autonomous vehicle, 3=train, 4=truck, 5=drone). j is the data type index (j=1~n correspond to different data types of the device, such as location, speed, weight, etc.). It is the threshold for edge computing power utilization. It is the bandwidth utilization threshold; It is edge computing power.

[0035] The formula for calculating the total delay is as follows: ; in, (Distance delay factor), L i It is the distance between the device and the edge node, in meters. It is edge computing power. B represents cloud computing power, and B represents communication bandwidth.

[0036] The formula for calculating total energy consumption is as follows: ; Among them, edge processing energy consumption Transmission energy consumption Cloud processing energy consumption For example, high real-time data (drone obstacle location, T...) i,j =50ms) will be allocated to edge processing (x i,j =1), low real-time data (truck remaining fuel, T i,j =500ms) allocated to the cloud (x i,j =0), to achieve optimal resource allocation.

[0037] At the edge layer, cross-device protocol adaptation involves converting the native protocols of different devices into the system-wide unified protocol (Protobuf) to ensure seamless data access. The protocol conversion matrix is ​​shown below: ; Line: Native protocol, 1=CAN (train / truck), 2=MQTT (drone), 3=HTTP (truck remote monitoring), 4=ROS (unmanned vehicle / forklift); Column: Unified protocol, only column 1 = Protobuf is 1; The protocol conversion efficiency constraint formula is as follows: ; Among them, P prot =(x,y) is the protocol conversion identifier function, P prot In the expression =(x,y), x represents the native protocol (corresponding to the classification of "rows" in the transformation matrix: 1=CAN, 2=MQTT, 3=HTTP, 4=ROS), and y represents the unified protocol (corresponding to the classification of "columns" in the transformation matrix: only 1=Protobuf format). It sums the values ​​of the native protocol type x (x=1~4). This represents the amount of device data using the native protocol x, and η is the protocol conversion efficiency, which must meet a constraint of ≥99%. The matrix ensures that all mainstream protocols can be converted to a unified protocol, and the efficiency constraint guarantees no data loss during transmission, adapting to the needs of multiple devices accessing the campus.

[0038] The formula for load balancing at the edge layer is as follows: ; in, It is the load factor of the k-th edge node. It is the set of tasks for node k; 1.2 is the average load across all edge nodes, and 1.2 is the load balancing threshold. When the edge layer load factor is greater than 1.2, task migration is triggered, as shown in the following formula: ; Prioritize migrating tasks with "large data volume and low real-time requirements" to the node with the lowest load; among them, The target task needs to be migrated from the overloaded node. T is the computing power of the k-th edge node. i，j是 The real-time threshold for data of type j in type i device; It is the average real-time threshold for all devices.

[0039] The compensation for lost data is shown in the following formula: ; Where λ = 0.1 is the attenuation coefficient. It is the duration of data loss. D is the historical average of the j-th type of data. comp (t) is the compensation value after data loss at time t, Dfusion (t-1) is the weighted fusion result of multi-device data at time t-1.

[0040] After the data is uploaded to the cloud, the data fusion at the cloud layer is achieved through weighted fusion, and the weighted fusion formula is shown below: ; Among them, D fusion (t) is the weighted fusion result of data from multiple devices at time t. It is the initial credibility weight of the i-th type of device and the j-th type of data (determined by the device hardware accuracy and data reliability). It is the normalized result of the j-th type of data of the i-th type of device at time t (standardized data after removing dimensions). For example, "location of goods in the park" = forklift location × 0.8 (high confidence) + unmanned vehicle location × 0.2 (low confidence); When data conflicts occur ( (Time-triggered) The data conflict is handled using the data conflict handling formula, as shown below: ; Where k=5 is used to adjust the sensitivity. Devices with a larger deviation from the mean have a lower weight to reduce the impact of conflicts. It is a new credibility weight that is dynamically adjusted after data conflicts; It is the normalized average value of the j-th type of data (a benchmark used to determine whether the data conflict). It is the normalized result of the j-th type of data of the i1-th type of device (used to compare conflicts with the same source data of other devices). It is the normalized result of the j-th type of data for the i2-th type of device (used to compare data from the same source with other devices to avoid conflicts); e k k is the weight adjustment sensitivity (controls the adjustment range of conflicting data weights, fixed at κ=5). It is the data conflict threshold (the difference between normalized data exceeding this value is considered a conflict, fixed at δ=0.3). The error constraint during fusion is shown in the following formula: ; For numerical data, Regarding location data, ; It is the fusion error threshold; It is the actual value of the data at time t (such as the actual cargo location, the actual equipment speed, used to verify the accuracy of fusion / compensation).

[0041] After data fusion is completed in the cloud, the scheduling data is transmitted back to the edge layer scheduling system to realize actual business applications and complete the scheduling of transportation equipment. The data offloading strategy is then modified based on the real-time operating status of the transportation equipment.

[0042] Figure 2 This is a schematic diagram of the structure of an intelligent transportation device data interaction system based on edge-cloud offloading provided in an embodiment of this application. The embodiment of this invention provides a data interaction structure for intelligent transportation devices based on edge-cloud offloading, including: The perception layer is used to collect multi-dimensional raw data on the transportation equipment at the logistics site; An edge layer is used to preprocess the multi-dimensional raw data to obtain multi-dimensional data; perform feature modeling on the multi-dimensional data to obtain the feature vector of the transport equipment; normalize the feature vector to obtain a standardized feature vector; formulate a data unloading strategy based on the logistics information of the transport equipment; and upload the standardized feature vector to the cloud computing center according to the data unloading strategy. The cloud layer is used to process the standardized feature vectors in the cloud computing center to obtain a logistics scheduling plan; to complete the logistics scheduling of the transport equipment according to the logistics scheduling plan; and to modify the data unloading strategy according to the real-time operating status of the transport equipment.

[0043] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0044] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A data interaction method for intelligent transportation equipment based on edge-cloud offloading, characterized in that, include: Collect multi-dimensional raw data from the transportation equipment at the logistics site; The multi-dimensional raw data is preprocessed to obtain multi-dimensional data; The feature vector of the carrier equipment is obtained by performing feature modeling on the multi-dimensional data. The feature vectors are normalized to obtain standardized feature vectors; A data unloading strategy is formulated based on the logistics information of the transport equipment; The standardized feature vector is uploaded to the cloud computing center according to the data offloading strategy. The standardized feature vectors are processed in the cloud computing center to obtain a logistics scheduling plan. The logistics scheduling of the transport equipment is completed according to the logistics scheduling plan. The data unloading strategy is modified based on the real-time operating status of the transport equipment.

2. The data interaction method for intelligent transportation equipment based on edge-cloud offloading according to claim 1, characterized in that, The multi-dimensional raw data includes trunk line transportation equipment data, short-distance transfer equipment data, and low-altitude inspection equipment data.

3. The data interaction method for intelligent transportation equipment based on edge-cloud offloading according to claim 1, characterized in that, The preprocessing includes data cleaning, data completion, and format standardization. The data cleaning includes outlier removal and noise filtering; The noise filtering involves applying moving average filtering to numerical data and median filtering to image data.

4. The data interaction method for intelligent transportation equipment based on edge-cloud offloading according to claim 1, characterized in that, The feature modeling involves extracting the three-dimensional core features of the vehicle to obtain a feature vector. The three-dimensional core features include device status features, motion features, and environmental interaction features; The equipment status characteristics include health, load rate, and operational stability; The motion characteristics include instantaneous acceleration, trajectory deviation, and driving efficiency; The environmental interaction characteristics include device spacing, communication quality, and environmental risks.

5. The data interaction method for intelligent transportation equipment based on edge-cloud offloading according to claim 1, characterized in that, The normalization is to map the feature vector to a unified interval through an adaptation method and then output a standardized feature vector.

6. The data interaction method for intelligent transportation equipment based on edge-cloud offloading according to claim 5, characterized in that, The data unloading strategy is formulated based on the logistics information of the transport equipment.

7. The data interaction method for intelligent transportation equipment based on edge-cloud offloading according to claim 6, characterized in that, The objective function for optimizing the data offloading strategy is shown in the following equation: ； in, It is to minimize the optimization objective function. The total delay T is respectively total Total energy consumption E total Weighting coefficients; total delay T total Total energy consumption E total satisfy Constraints; in logistics scenarios, real-time performance is prioritized.

8. The data interaction method for intelligent transportation equipment based on edge-cloud offloading according to claim 1, characterized in that, The data offloading strategy is divided by constraints, which are shown in the following formula: ； Where st is subjectto, used to introduce the constraint conditions, S i，j X is the amount of data of type j of device i; i，j It is the unloading of decision variables, X i，j The variable X can take values ​​of 0 or 1. i，j =1 indicates data edge processing; X i，j =0 indicates cloud processing; i is the device type index, where i=1~5 correspond to: 1=forklift, 2=autonomous vehicle, 3=train, 4=truck, 5=drone; j is the data type index, where j=1~n corresponds to different data types of the device. This indicates that the summation is performed on the device type index i from 1 to 5. This indicates that the summation is performed on data type index j from 1 to n; It is the threshold for edge computing power utilization. B is the bandwidth utilization threshold, and B is the communication bandwidth. It is edge computing power.

9. A data interaction system for intelligent transportation equipment based on edge-cloud offloading, the system being used to implement the method described in any one of claims 1-8, characterized in that, include: The perception layer is used to collect multi-dimensional raw data on the transportation equipment at the logistics site; An edge layer is used to preprocess the original multi-dimensional data to obtain multi-dimensional data. Feature modeling is performed on the multi-dimensional data to obtain the feature vector of the transport equipment; the feature vector is normalized to obtain a standardized feature vector; a data unloading strategy is formulated based on the logistics information of the transport equipment; and the standardized feature vector is uploaded to the cloud computing center according to the data unloading strategy. The cloud layer is used to process the standardized feature vectors in the cloud computing center to obtain a logistics scheduling plan; to complete the logistics scheduling of the transport equipment according to the logistics scheduling plan; and to modify the data unloading strategy according to the real-time operating status of the transport equipment.

Images