Flow-optimized remote positioning data uploading method

Abstract

The invention provides a remote positioning data uploading method for flow optimization, and relates to the technical field of intelligent traffic, and the method comprises the steps: collecting a network quality parameter in a vehicle driving process, transmitting the network quality parameter to a vehicle-mounted terminal, adjusting the size of a cache window according to the network signal intensity, caching the data, and uploading the data to a cloud end; acquiring continuous positioning points of the automobile, performing curve fitting, and replacing a mode of positioning the automobile position by original coordinates of the automobile with a mode of uploading curve parameters; in a low-speed or straight-line driving scene, simply calculating and uploading a difference value with previous data; the technical key points are as follows: by optimizing data transmission and compression modes, efficient data uploading can also be realized under the condition of lower bandwidth, the requirement on hardware configuration of the vehicle-mounted terminal is reduced, and the hardware cost of the vehicle-mounted terminal is reduced, so that assistance is provided for large-scale popularization of intelligent transportation equipment, and rapid development of the whole industry is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation, and specifically to a method for uploading remote positioning data with optimized traffic. Background Art

[0002] An intelligent transportation system effectively integrates advanced scientific and technological means (information technology, computer technology, data communication technology, sensor technology, electronic control technology, automatic control theory, operations research, artificial intelligence, etc.) into transportation, service control, and vehicle manufacturing, strengthening the connection among vehicles, roads, and users, thereby forming a comprehensive transportation system that ensures safety, improves efficiency, enhances the environment, and saves energy.

[0003] In the era of the booming development of intelligent vehicles, the upload of vehicle positioning data to the cloud is crucial for realizing effective vehicle monitoring, intelligent driving assistance, and numerous location-based service applications. Existing vehicle positioning data upload technologies mainly rely on the full-volume transmission mode, which has problems such as high bandwidth pressure and transmission costs, data loss caused by unstable signals, redundant data occupying bandwidth, and the contradiction between real-time performance and resource consumption. Specifically as follows:

[0004] (1) Since intelligent vehicles are equipped with a large number of advanced sensors, a large amount of data is generated per unit time. Among them, in order to meet the real-time requirements, positioning data often needs to be uploaded at a frequency of seconds or even milliseconds. Although the traditional full-volume upload scheme can ensure data integrity, it consumes huge traffic costs. For large-scale vehicle fleets, the annual storage cost alone may be as high as hundreds of millions of yuan, which undoubtedly brings a heavy economic burden to enterprises and severely restricts the large-scale promotion and development of related businesses;

[0005] (2) In special scenarios such as tunnels, overpasses, and urban canyons, satellite signals are prone to interruption or attenuation. When relying on base station positioning, its accuracy will drop significantly, and it is necessary to switch to WiFi or UWB positioning. However, the traditional system lacks an effective dynamic caching mechanism, and during the signal switching process, data loss is extremely likely to occur. This not only affects the accurate tracking of vehicle positions but also has a serious negative impact on various applications based on location data, reducing the quality and reliability of services;

[0006] (3) When the vehicle is stationary or moving at a low speed, the continuously uploaded positioning data is highly redundant. This waste of network resources further exacerbates the network transmission pressure, resulting in frequent network congestion during peak data transmission periods, seriously affecting the efficiency and timeliness of data upload.

[0007] (4) Traditional compression algorithms, such as ZIP or GZIP, although having certain effects in data compression, require high computing resources. However, in-vehicle terminals need to process numerous tasks with extremely high real-time requirements simultaneously, such as autonomous driving perception, vehicle status monitoring, etc. Excessive resources occupied by the compression algorithm will seriously affect the execution efficiency of other key tasks and may even pose a threat to driving safety. Summary of the Invention

[0008] In order to overcome the obvious deficiencies of the existing vehicle positioning data upload technology in multiple aspects such as data volume, signal stability, redundant data processing, and resource balance, the embodiments of this application provide a method for optimizing traffic in remote positioning data upload. By optimizing data transmission and compression methods, efficient data upload can be achieved even under low-bandwidth conditions, reducing the requirements for the hardware configuration of in-vehicle terminals, which is beneficial to reducing the hardware costs of in-vehicle terminals, thereby providing assistance for the large-scale popularization of intelligent transportation devices and promoting the rapid development of the entire industry.

[0009] The technical solution adopted by the embodiments of this application to solve its technical problems is as follows:

[0010] A method for optimizing traffic in remote positioning data upload, including collecting network quality parameters during the driving of a vehicle and transmitting them to the in-vehicle terminal to adjust the size of the cache window according to the network signal strength, caching the data and uploading it to the cloud;

[0011] Collecting continuous positioning points of the vehicle, performing curve fitting, and replacing the original coordinates of the vehicle with the form of uploading curve parameters for vehicle positioning; in the scenario of low-speed or straight-line driving, simply calculate and upload the difference from the previous data.

[0012] In a possible implementation manner, it includes a data caching module and a data compression module applied to this data upload method. The data caching module includes a dynamic cache window adjustment unit and a hierarchical caching unit, and the data compression module includes a trajectory fitting compression unit and a context-aware compression unit; after the data caching module and the data compression module complete the processing of vehicle driving data, they are uploaded to the cloud.

[0013] In a possible implementation manner, the network quality parameters include network signal strength and packet loss rate. The dynamic cache window adjustment unit sets a threshold interval N for the network signal strength. When the network signal strength is less than the minimum value in the threshold interval N, the cache window is enlarged and local caching is started; when the network signal strength is greater than the maximum value in the threshold interval N, the cache window is reduced and the data transmission frequency is increased.

[0014] In a possible implementation, the hierarchical cache unit constructs a three-level architecture of in-vehicle terminal, roadside unit (RSU), and edge server for hierarchical caching. The hierarchical caching consists of two methods: local caching and edge caching. The local cache stores the uncompressed original positioning data to retain high-precision data. The edge cache stores the compressed trajectory feature parameters.

[0015] In a possible implementation, in the edge cache, by using an asynchronous federated learning model, the upload priority of high-value data is predicted. In building the model, let θ ∈ R D represent the global model parameters, initialized as θ0. The client k locally trains the model θ t for E rounds to obtain the updated parameters implemented by stochastic gradient descent (SGD):

[0016] where η is the learning rate, is the local loss function, k is the client, and the local dataset of client k is D k , with the data volume being n k ;

[0017] Meanwhile, the update of each client carries a timestamp t, indicating the global model version (or local iteration number) corresponding to this update.

[0018] In a possible implementation, in the edge cache, for the aggregation of asynchronous data, two methods are used: weighted asynchronous averaging or ignoring outdated updates. When the server maintains the global model and receives the update from client k at timestamp t, it is updated according to the following formula:

[0019]

[0020] where α is the aggregation weight, which is related to the client data volume n k or the update frequency (e.g., );

[0021] When introducing the time decay factor γ(t, t′) (t′ is the current timestamp of the server, and γ decreases as the time difference increases), the aggregation formula is:

[0022]

[0023] In a possible implementation, the trajectory fitting and compression unit adopts a trajectory fitting and compression method. For continuous positioning points, curve fitting is performed by means of Bezier curves, and then the curve parameters are uploaded. The general formula of the Bezier curve is: Given n + 1 control points P0, P1, …, P n, the coordinates of any point P(t) on the curve are:

[0024]

[0025] where is the combination number.

[0026] In a possible implementation, when using the trajectory fitting compression method to process continuous positioning points, a differential coding and incremental update strategy is adopted. For vehicle positioning in low-speed or straight-line driving scenarios, only the differences from the previous data are uploaded, that is, the longitude difference, latitude difference, and timestamp difference.

[0027] In a possible implementation, the context-aware compression unit dynamically selects two compression algorithms, lossy and lossless, to compress the data according to different scenarios.

[0028] In a possible implementation, during the process of data caching and uploading to the cloud, a breakpoint resumption method is adopted. During a network interruption, the control unit of the in-vehicle terminal will record the position information of the data block currently being uploaded, calculate the hash value using the hash algorithm, and compare it with the hash value returned by the receiving party for data verification; the operation process of the hash value algorithm includes preprocessing, initializing the hash value, block processing, core operation, and output.

[0029] The beneficial effects of this application are as follows:

[0030] First, in this solution, by innovatively adopting data compression technologies such as trajectory fitting compression, differential coding, and incremental update, the data traffic cost is significantly reduced. This not only saves a large amount of funds for enterprises, greatly relieving the cost pressure in data storage and transmission, but also enables the business based on vehicle positioning data to be carried out on a large scale at a lower cost, which is conducive to the overall development and innovation of the industry;

[0031] Second, in this solution, the dynamic caching strategy plays a key role when the network signal is unstable. By dynamically adjusting the cache window size according to the network quality, combined with the hierarchical caching strategy and the breakpoint resumption and data verification mechanisms, the data loss rate during network interruption is greatly reduced. It can ensure that in a complex communication environment, vehicle positioning data can still be uploaded to the cloud completely and accurately, providing reliable data support for applications such as vehicle monitoring and intelligent driving, and improving the stability and availability of related services;

[0032] Thirdly, in this solution, the lightweight compression method adopted can ensure effective data compression while occupying little CPU resources. When the in-vehicle terminal processes the positioning data upload task, it will not overly affect other real-time tasks. At the same time, this method can ensure the real-time nature of the positioning data upload, meet the strict requirements for data timeliness in the intelligent transportation field, and ensure the efficient operation of various applications during vehicle driving.

[0033] Fourthly, in this solution, by optimizing the data transmission and compression methods, efficient data upload can be achieved even under low-bandwidth conditions, reducing the requirements for the hardware configuration of the in-vehicle terminal, which is beneficial to reducing the hardware cost of the in-vehicle terminal, thereby providing assistance for the large-scale popularization of intelligent transportation devices and promoting the rapid development of the entire industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a cache flow chart of a method for uploading remote positioning data with optimized traffic according to the present invention under network signal changes;

[0035] Figure 2 It is an associated flow chart of vehicle positioning data compression in a method for uploading remote positioning data with optimized traffic according to the present invention;

[0036] Figure 3 It is a system framework diagram of a method for uploading remote positioning data with optimized traffic according to the present invention;

[0037] Figure 4 It is a schematic flow chart of hierarchical caching in a method for uploading remote positioning data with optimized traffic according to the present invention;

[0038] Figure 5 It is a framework diagram of a data compression method in a method for uploading remote positioning data with optimized traffic according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The technical solutions in the embodiments of the present application are to solve the problems in the above-mentioned background technology, and the general idea is as follows:

[0040] Embodiment 1:

[0041] This embodiment introduces the data caching of a method for uploading remote positioning data with optimized traffic. Specifically, refer to Figures 1 - 5 As shown, it includes collecting network quality parameters during the driving of the vehicle and transmitting them to the in-vehicle terminal to adjust the size of the cache window according to the network signal strength, caching the data and uploading it to the cloud;

[0042] As Figure 3As shown in the figure, the remote positioning data method includes a data caching module and a data compression module applied to the data uploading method. The data caching module includes a dynamic caching window adjustment unit (which adjusts the caching window size in real time and dynamically according to key indicators such as network quality, such as 4G / 5G signal strength and packet loss rate) and a hierarchical caching unit;

[0043] Among them, the network quality parameters include network signal strength (4G / 5G) and packet loss rate, etc. The dynamic caching window adjustment unit sets a threshold range N for the network signal strength (the signal strength threshold N can be 30%-70%). When the network signal strength is less than the minimum value in the threshold range N, it means that the network condition is poor. At this time, the caching window is enlarged (to 10 seconds), and local caching is started; when the network signal strength is greater than the maximum value in the threshold range N, it indicates that the network condition is good. At this time, the caching window is reduced (to 1 second). In this way, data transmission can be carried out more frequently, and the timeliness of data can be guaranteed in the form of improving the data transmission frequency;

[0044] Specifically, first, the vehicle-mounted communication module (such as DSRC / V2X) collects network quality parameters such as network signal strength and packet loss rate in real time (such as by periodically sending probe packets to the base station or surrounding communication nodes, obtaining feedback information to calculate the packet loss rate, and at the same time reading the received signal strength indication value (RSSI) to determine the signal strength);

[0045] Then, the collected network quality parameters are transmitted to the control unit on the vehicle-mounted terminal. The control unit makes a judgment according to the preset rules. When the signal strength is greater than 70%, the control unit sends an instruction to the caching module to set the caching window to 1 second. Within this 1 second, the positioning data obtained by the positioning module is temporarily stored in the caching module. Once the caching window time ends, the caching module packs the data within this 1 second and prepares to upload;

[0046] At the same time, when the signal strength is less than 30%, on the one hand, the control unit sends an instruction to the caching module to expand the caching window to 10 seconds; on the other hand, it starts the local compression module. Within the 10-second caching window time, the positioning data continues to be stored in the caching module, and at the same time the local compression module compresses the cached data according to a specific compression algorithm (such as a lightweight compression algorithm optimized for the characteristics of positioning data). The compressed data continues to be stored in the caching module and waits to be uploaded after the network condition improves;

[0047] Such as Figure 3 As shown in the figure, the hierarchical caching unit constructs a three-level architecture of vehicle-mounted terminal, roadside unit (RSU), and edge server for hierarchical caching. The hierarchical caching consists of two methods: local caching and edge caching;

[0048] Among them, the local cache stores the uncompressed original positioning data to retain high-precision data. The local cache of the vehicle terminal uses high-speed flash memory (such as SSD) as the storage medium to ensure fast storage and reading of the uncompressed original positioning data. Real-time positioning obtains the positioning information of the vehicle, including detailed data such as longitude, latitude, altitude, speed, and timestamp, and directly stores it in the local cache. The management of the local cache adopts a strategy of First In First Out (FIFO) combined with importance marking (for some special positioning data, such as the positioning information when the vehicle enters a specific area (such as accident-prone areas, construction areas, etc.), it is marked as important data and preferentially retained in the local cache to avoid being squeezed out of the cache by the FIFO strategy);

[0049] Secondly, the Road Side Unit (RSU) and the edge server together constitute the edge cache layer. When the data in the local cache is ready to be uploaded, it is first transmitted to the RSU. The built-in processing module of the RSU performs preliminary analysis and processing on the data, and uses curve fitting algorithms (such as Bezier curve fitting or polynomial fitting) to process continuous positioning points, and extracts curve feature parameters (such as the control point coordinates of the curve, polynomial coefficients, etc.). These compressed trajectory feature parameters are stored in the cache of the RSU;

[0050] At the same time, the RSU uses an asynchronous federated learning model to evaluate and predict the importance of the data, and generates an upload priority label for each data block. When the edge server has idle resources, the RSU transfers the data in the cache to the edge server for further storage and processing according to the upload priority;

[0051] Specifically, by using an asynchronous federated learning model in the edge cache to predict the upload priority of high-value data, it can ensure that important data is uploaded first, improving the effectiveness and pertinence of data transmission. In constructing the model, let θ ∈ R D represent the global model parameters, initialized as θ0, and the client k locally trains the model θ t for E rounds to obtain the updated parameter implemented by Stochastic Gradient Descent (SGD):

[0052] where η is the learning rate, is the local loss function, k is the client, and the local dataset of client k is D k with the data volume of n k ;

[0053] At the same time, the update of each client carries a timestamp t, indicating the global model version (or local iteration number) corresponding to this update.

[0054] When aggregating asynchronous data in the edge cache using either weighted asynchronous averaging or ignoring outdated updates, and maintaining a global model on the server, when receiving an update from client k at timestamp t,

[0055] update according to the following formula:

[0056]

[0057] where α is the aggregation weight, which is related to the amount of client data n k or the update frequency (e.g., );

[0058] When introducing a time decay factor γ(t, t′) (t′ is the current timestamp of the server, and γ decreases as the time difference increases), the aggregation formula is:

[0059]

[0060] In some examples, when caching and uploading data to the cloud, the method of resuming interrupted transfer is adopted. During a network interruption, the control unit of the vehicle terminal will record the position information of the data block currently being uploaded, calculate the hash value using the hash algorithm, and compare it with the hash value returned by the receiving party for data verification;

[0061] The operation process of the hash value algorithm includes preprocessing, initializing the hash value, block processing, core operation, and output;

[0062] Among them, preprocessing requires first padding the input message so that its length is 448 modulo 512. The padding method is to add a 1 and several 0s after the message, and then represent the original message length (in bits) as a 64-bit unsigned integer and add it to the end of the padded message. At this time, the length of the message is an integer multiple of 512;

[0063] Initializing the hash value requires using 8 32-bit constants (hash initial values), denoted as h0, h1, h2, h3, h4, h5, h6, h7 respectively (these constants are usually obtained by taking the first 32 bits of the decimal part of the square root of the first 8 prime numbers);

[0064] Block processing first divides the padded message into 512-bit blocks and processes each block in turn. For each 512-bit block, it is further divided into 16 32-bit words, and then expanded into 64 32-bit words through a series of expansion operations;

[0065] The core operation performs a series of complex logical operations on each expanded 32-bit word, including operations such as AND, OR, NOT, XOR, etc., as well as some shift operations.

[0066] These operations involve the 8 hash values mentioned above and some fixed constants. The specific operation formulas are as follows:

[0067] First, define some auxiliary functions:

[0068]

[0069] Among them, ∧ is the bitwise AND, is the bitwise XOR, is the bitwise NOT;

[0070]

[0071] Among them, ROTR(n, x) represents rotating x to the right by n bits;

[0072] Σ1(x) represents ROTR(6, x) ⊕ ROTR(11, x) ⊕ ROTR(25, x);

[0073] σ0(x) represents ROTR(7, x) ⊕ ROTR(18, x) ⊕ SHR(3, x);

[0074] Among them, SHR(n, x) represents shifting x to the right by n bits and filling the high bits with 0;

[0075] σ1(x) represents ROTR(17, x) ⊕ ROTR(19, x) ⊕ SHR(10, x).

[0076] Then, for each group of 64 words W[i] (i ranges from 0 to 63), perform the following calculations:

[0077] T1 = h7 + Σ1(h4) + Ch(h4, h5, h6) + K[i] + W[i];

[0078] Among them, K[i] is a fixed constant related to the current round of processing;

[0079] T2 = Σ0(h0) + Maj(h0, h1, h2);

[0080] h7 = h6, h6 = h5, h5 = h4, h4 = h3 + T1, h3 = h2, h2 = h1, h1 = h0, h0 = T1 + T2.

[0081] The output is to concatenate the final 8 hash values h0, h1, h2, h3, h4, h5, h6, h7 after processing all groups, and the final 256-bit hash value is obtained, which can be used for data integrity verification to ensure that the data is not lost or damaged during transmission.

[0082] It should be noted that in complex scenarios such as the hierarchical positioning of overpasses, through multi-sensor fusion technology, that is, by combining the data of multiple sensors such as GPS, IMU (Inertial Measurement Unit), and wheel speed sensors (well-known technologies), and edge caching strategies, the positioning accuracy can be effectively improved, enabling the position of the vehicle in a complex environment to be determined more precisely, providing a more accurate data basis for applications such as intelligent driving and vehicle navigation, and enhancing the user experience and driving safety.

[0083] Embodiment 2:

[0084] Based on Embodiment 1, as Figures 2 to 5 shown, this embodiment introduces the data compression of the remote positioning data upload method for traffic optimization. By collecting continuous positioning points of the vehicle and performing curve fitting, the vehicle position is located by uploading the curve parameters in place of the original coordinates of the vehicle (in the scenario of low speed or straight driving, simply calculate and upload the difference from the previous data). After the data caching module and the data compression module complete the processing of the vehicle driving data, it is uploaded to the cloud;

[0085] As Figure 3 shown, the data compression module includes a trajectory fitting compression unit and a context-aware compression unit;

[0086] Among them, the trajectory fitting compression unit adopts the trajectory fitting (or polynomial fitting) compression method. For continuous positioning points, curve fitting is performed by means of Bezier curves, and then the curve parameters are uploaded;

[0087] The general formula of the Bezier curve is: Given n + 1 control points P0, P1,..., P n , the coordinates of any point P(t) on the curve are:

[0088]

[0089] Among them, is the combination number.

[0090] At the same time, when using the trajectory fitting compression method to process continuous positioning points, a difference coding and incremental update strategy is adopted. For the vehicle positioning in the low-speed or straight driving scenario, only the differences from the previous data are uploaded, that is, the longitude difference, latitude difference, and timestamp difference (for example, when the vehicle is driving at a constant speed on the highway, the data volume can be reduced by 80% using this strategy, effectively reducing the data transmission volume);

[0091] When the vehicle is in a low-speed or straight-line driving scenario, this difference change is relatively small. At this time, only these difference data are uploaded instead of the complete positioning data. For example, when the vehicle is moving slowly in a parking lot or driving at a constant speed in a straight line on a highway, adopting this method can reduce the data volume by 80%; and after the receiving end (such as a cloud server) receives the difference data, combined with the complete positioning data stored last time, the current actual positioning data can be restored through simple calculations to achieve incremental data update;

[0092] Specifically, in the process of obtaining the positioning point data of the vehicle, every time a certain number of consecutive positioning points (such as 100) are obtained, the data compression module of the in-vehicle terminal starts the trajectory fitting compression process. The compression processing module automatically selects a suitable curve fitting algorithm according to the distribution characteristics of the positioning points. If the positioning points show a relatively smooth curve feature, the Bezier curve fitting is preferred; if the change of the positioning points more conforms to the polynomial law, the polynomial fitting is selected (for the positioning point data of a vehicle driving at a constant speed on a highway, the polynomial fitting may be more appropriate);

[0093] For a given set of data points {(x1,x1),(x2,y2),…,(xn,xn)}, the goal of polynomial fitting is to find an m-degree polynomial function f(x) = a0 + a1x + a1x 1 + a2x 2 + … + a m x m , such that this polynomial function fits these data points as well as possible in a certain sense. Here, m is the degree of the polynomial, and a0, a1, …, a m are the coefficients of the polynomial;

[0094] And the fitting method usually uses the least squares method to determine the coefficients of the polynomial. The basic idea of the least squares method is to minimize the sum of the squares of the errors between the fitting function and the data points, that is, for the given data points (x i ,y i ) and the fitting polynomial f(x), to minimize the following objective function:

[0095]

[0096] To find the coefficients a0, a1, …, a m that minimize S, take the partial derivatives of S with respect to respectively and set these partial derivatives equal to zero to obtain a system of linear equations, and solving this system of equations can obtain the coefficients of the polynomial;

[0097] At the same time, taking polynomial fitting as an example, the compression processing module calculates the polynomial coefficients that can best fit these positioning points through mathematical methods such as the least squares method. Suppose a quadratic polynomial y = ax2 If the coordinate relationship of the positioning points is fitted by \(y = ax^{2}+bx + c\), then only the three coefficients \(a\), \(b\), and \(c\) need to be uploaded, without uploading 100 original coordinate points, thus achieving a compression rate of up to 97%;

[0098] In addition, in order to compensate for the error of trajectory prediction in case of signal loss, etc., the compression processing module combines the Kalman filtering algorithm. The Kalman filtering algorithm uses the historical positioning data of the vehicle and the current motion state (such as speed, acceleration, etc.) to predict and correct the possibly lost positioning points, ensuring that the compressed data can accurately restore the driving trajectory of the vehicle in subsequent use;

[0099] In some examples, the context-aware compression unit dynamically selects two compression algorithms, lossy and lossless, to compress the data according to different scenarios;

[0100] Among them, in the urban road scenario, due to the high requirement for accuracy, lossless compression is adopted to retain centimeter-level accuracy; in the highway scenario, the requirement for accuracy is relatively low, and lossy compression is enabled to obtain a higher compression rate and reach meter-level accuracy, which can comprehensively improve the performance of uploading automotive positioning data;

[0101] Specifically, the scenario recognition module built in the vehicle terminal determines the scenario where the vehicle is located by fusing various sensor data. For example, it comprehensively analyzes by combining GPS data, map information, vehicle speed sensor data, and camera image data. When the vehicle speed is low and the surrounding environmental features show an urban street (such as recognizing buildings and traffic lights on both sides of the street through the camera), the scenario recognition module determines that the vehicle is in the urban road scenario; when the vehicle speed is high and the surrounding environment is an open area (such as determining as a highway through map information and GPS data), it is determined as the highway scenario;

[0102] According to the scenario recognition result, the compression control module of the vehicle terminal dynamically selects an appropriate compression algorithm. In the urban road scenario, since the requirement for positioning accuracy is extremely high and centimeter-level accuracy needs to be retained, a lossless compression algorithm such as arithmetic coding is selected to ensure that there is no information loss during the compression and decompression process. In the highway scenario, the requirement for accuracy is relatively low, and a certain degree of accuracy loss is allowed in exchange for a higher compression rate. At this time, a lossy compression algorithm such as the compression algorithm based on the discrete cosine transform (DCT) is enabled to achieve a higher compression rate while reaching meter-level accuracy.

[0103] Finally, it should be noted that: Obviously, the above embodiments are merely examples given to clearly illustrate the present invention, rather than limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. And the obvious changes or modifications derived therefrom still fall within the protection scope of the present invention.

Claims

1. A method for uploading remote positioning data with optimized traffic, characterized in that, Including: Collect network quality parameters during the driving of the vehicle, and transmit them to the in-vehicle terminal to adjust the size of the cache window according to the network signal strength, cache the data, and upload it to the cloud. Collect continuous positioning points of the vehicle, perform curve fitting, and replace the original coordinates of the vehicle with the form of uploading curve parameters to position the vehicle; in the scenario of low-speed or straight driving, simply calculate and upload the difference from the previous data.

2. The method for uploading remote positioning data with optimized traffic according to claim 1, characterized in that: Including a data cache module and a data compression module applied to this data upload method. The data cache module includes a dynamic cache window adjustment unit and a hierarchical cache unit. The data compression module includes a trajectory fitting compression unit and a context-aware compression unit. Among them, after the data cache module and the data compression module complete the processing of vehicle driving data, they are uploaded to the cloud.

3. The method for uploading remote positioning data with optimized traffic according to claim 2, wherein: The network quality parameters include network signal strength and packet loss rate. The dynamic cache window adjustment unit sets a threshold range N for the network signal strength. When the network signal strength is less than the minimum value in the threshold range N, the cache window is enlarged, and local caching is started; when the network signal strength is greater than the maximum value in the threshold range N, the cache window is reduced, and the data transmission frequency is increased.

4. The method for uploading remote positioning data with optimized traffic according to claim 2, wherein: The hierarchical cache unit constructs a three-level architecture of in-vehicle terminal, roadside unit (RSU), and edge server for hierarchical caching. The hierarchical caching consists of two methods: local caching and edge caching. Among them, the local cache stores the uncompressed original positioning data to retain high-precision data; the edge cache stores the compressed trajectory feature parameters.

5. The method for uploading remote positioning data with optimized traffic according to claim 4, wherein: In the edge cache, by using an asynchronous federated learning model, the upload priority of high-value data is predicted. In the construction of the model, let θ ∈ R D represent the global model parameters, which are initialized as θ0. The client k locally trains the model θ t for E rounds and obtains the updated parameters implemented by stochastic gradient descent (SGD): where η is the learning rate, is the local loss function, k is the client, and the local dataset of client k is D k , with the data volume being n k ; At the same time, the update of each client carries a timestamp t, indicating the global model version (or local iteration number) corresponding to this update.

6. The method for uploading remote positioning data with optimized traffic according to claim 5, wherein: For the aggregation of asynchronous data in the edge cache, two methods of weighted asynchronous averaging or ignoring outdated updates are used. When the server maintains the global model and receives the update from client k at timestamp t, Update according to the following formula: where α is the aggregation weight, which is related to the client data volume n k or the update frequency (such as ); When introducing a time decay factor γ(t, t′) (t′ is the current timestamp of the server, and γ decreases as the time difference increases), the aggregation formula is:

7. The method for uploading remote positioning data with optimized traffic according to claim 2, characterized in that: The trajectory fitting and compression unit adopts a trajectory fitting and compression method. For continuous positioning points, curve fitting is performed by means of a Bezier curve, and then the curve parameters are uploaded. The general formula of the Bezier curve is: Given n + 1 control points P0, P1, …, P n , the coordinates of any point P(t) on the curve are: Among them, is the combination number.

8. The method for uploading remote positioning data with optimized traffic according to claim 7, characterized in that: When using the trajectory fitting compression method to process continuous positioning points, a difference coding and incremental update strategy is adopted. For the positioning of the vehicle in the low-speed or straight driving scenario, only the differences from the previous data, namely longitude difference, latitude difference, and timestamp difference, are uploaded.

9. The method for uploading remote positioning data with optimized traffic according to claim 1, wherein: The context-aware compression unit dynamically selects two compression algorithms, lossy and lossless, to compress the data according to different scenarios.

10. A method for uploading remote positioning data with optimized traffic as described in claim 1, characterized in that: During the process of caching and uploading data to the cloud, the breakpoint resumption method is adopted. During the network interruption, the control unit of the in-vehicle terminal will record the position information of the data block currently being uploaded, calculate the hash value using the hash algorithm, and compare it with the hash value returned by the receiving party for data verification. Among them, the operation process of the hash value algorithm includes preprocessing, initializing the hash value, block processing, core operation, and output.

Images