A trajectory mapping method, server, client and system

Abstract

The trajectory mapping method, the server, the client and the system provided by the embodiment of the present application have the following beneficial effects: global model parameters are randomly sent to multiple clients; reference model parameters of the multiple clients after iterative training and noise disturbance are received; reference model parameters of the multiple clients are aggregated to obtain new global model parameters; the new global model parameters are sent to the multiple clients, so that the multiple clients perform a new round of iterative training. The present application performs noise disturbance on model parameters, increases the difference between disturbed data and original data, and achieves stronger privacy protection effect.

Description

Technical Field

[0001] This invention relates to the field of computers, and more specifically to a trajectory mapping method, server, client, and system. Background Technology

[0002] Crowdsourced mapping methods based on trajectory data mainly fall into three categories: clustering, trajectory merging, and kernel density estimation. Clustering-based methods first generate a set of road endpoints using a clustering algorithm based on geospatial distance and directional similarity, and then connect these road endpoints into road segments using trajectories. Trajectory merging-based methods scan trajectories sequentially and, for each trajectory, either merge it into an existing road segment or generate a new road segment if merging is not possible. Kernel density estimation-based methods perform kernel density estimation on the point cloud and then extract the map.

[0003] The above crowdsourcing mapping methods often involve publishing and executing spatiotemporal tasks while exposing the information and physical location of the tasks and staff, which can easily expose users' real data. Privacy protection is an important factor in crowdsourcing mapping.

[0004] Therefore, how to effectively protect privacy is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] To overcome the shortcomings of existing trajectory mapping methods in terms of poor privacy protection, this invention proposes a trajectory mapping method, server, client, and system.

[0006] To achieve the above objectives, according to a first aspect of the present invention, an embodiment of the present invention provides a trajectory mapping method applied to a server, the method comprising the following steps:

[0007] Randomly send global model parameters to multiple clients;

[0008] Receive the reference model parameters after iterative training and noise perturbation by the multiple clients;

[0009] A new global model parameter is obtained by aggregating reference model parameters from multiple clients;

[0010] The new global model parameters are sent to the multiple clients so that the multiple clients can perform the next round of iterative training.

[0011] Optionally, the new global model parameters are obtained by aggregating the reference model parameters of multiple clients, and are calculated using the following formula:

[0012]

[0013] Among them, W t+1W represents the global model parameters in round t+1, which is the new model parameter. t Let m represent the global model parameters in round t, and ΔW represent the number of clients. i This represents the reference model parameters returned after the i-th client iterative training and noise perturbation.

[0014] Optionally, sending the new global model parameters to the plurality of clients includes:

[0015] Calculate the validation metrics for the current global model;

[0016] Calculate the deviation of the validation metrics across multiple rounds;

[0017] When the deviation exceeds a preset threshold, new global model parameters are sent to the multiple clients.

[0018] According to a second aspect of the present invention, embodiments of the present invention also provide another trajectory mapping method, applied to a client, comprising:

[0019] Update the local model parameters based on the global model parameters received from the server;

[0020] The collected trajectory data is feature extracted and aligned with the real road network map to obtain a trajectory raster feature map. The trajectory raster feature map is then input into the model corresponding to the local model parameters for iterative training.

[0021] Calculate the difference in local model parameters before and after iterative training, and perform a pruning operation;

[0022] Based on the privacy budget, dataset size, and number of clients participating in training, noise perturbation is applied to the cropped difference to obtain reference model parameters, which are then sent to the server for aggregation calculation.

[0023] Optionally, the difference between the local model parameters before and after the calculation iteration is performed, and a pruning operation is calculated using the following formula:

[0024]

[0025] in, This represents the difference after cropping. This represents the difference before cropping, and C represents the preset cropping threshold. The Euclidean norm represents the difference before clipping.

[0026] Optionally, the reference model parameters are obtained by perturbing the cropped difference with noise based on the privacy budget, dataset size, and number of clients participating in training, and calculated using the following formula:

[0027]

[0028] σ=kLΔS / ε

[0029]

[0030] in, Indicates the reference model parameters. N(0,σ) represents the difference after clipping. 2 ) indicates that the mean is 0 and the variance is σ. 2 The distribution is Gaussian, k is the privacy constant, C is the preset pruning threshold, ε is the privacy budget, δ is the sensitivity, ΔS is the sensitivity, L is the ratio of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

[0031] According to a third aspect of the present invention, embodiments of the present invention also provide a server, comprising:

[0032] The sending module is used to randomly send global model parameters to multiple clients;

[0033] The receiving module is used to receive the reference model parameters after iterative training and noise perturbation by the multiple clients;

[0034] The aggregation module is used to aggregate reference model parameters from multiple clients to obtain new global model parameters.

[0035] The iteration module is used to send new global model parameters to the multiple clients so that the multiple clients can perform a new round of iterative training.

[0036] According to a fourth aspect of the present invention, embodiments of the present invention also provide a client, comprising:

[0037] The update module is used to update the local model parameters based on the global model parameters received from the server.

[0038] The training module is used to extract features from the collected trajectory data and align them with the real road network map to obtain a trajectory raster feature map. The trajectory raster feature map is then input into the model corresponding to the local model parameters for iterative training.

[0039] The pruning module is used to calculate the difference in local model parameters before and after iterative training and to perform pruning operations.

[0040] The noise perturbation module is used to perturb the cropped difference with noise according to the privacy budget, dataset size and number of clients participating in training, to obtain reference model parameters, and send the reference model parameters to the server for aggregation calculation.

[0041] Optionally, the noise disturbance module calculates the reference model parameters using the following formula:

[0042]

[0043] σ=kLΔS / ε

[0044]

[0045] in, Indicates the reference model parameters. N(0,σ) represents the difference after clipping. 2 ) indicates that the mean is 0 and the variance is σ. 2 The distribution is Gaussian, k is the privacy constant, C is the preset pruning threshold, ε is the privacy budget, δ is the sensitivity, ΔS is the sensitivity, L is the ratio of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

[0046] According to a fifth aspect of the present invention, embodiments of the present invention also provide a trajectory mapping system, the trajectory mapping system including the server described in the above embodiments and the client described in the above embodiments.

[0047] As described above, the trajectory mapping method, server, client, and system provided by this invention have the following beneficial effects: randomly sending global model parameters to multiple clients; receiving reference model parameters after iterative training and noise perturbation by the multiple clients; aggregating the reference model parameters from the multiple clients to obtain new global model parameters; and sending the new global model parameters to the multiple clients to enable the multiple clients to perform a new round of iterative training. This invention perturbs the model parameters with noise, increasing the difference between the perturbed data and the original data, thus achieving a stronger privacy protection effect. Attached Figure Description

[0048] Figure 1 This is a schematic flowchart of a trajectory mapping method provided in an embodiment of the present invention;

[0049] Figure 2 This is a schematic flowchart of an iterative control method provided in an embodiment of the present invention;

[0050] Figure 3 This is a flowchart illustrating another trajectory mapping method provided in an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of the structure of a server for an execution trajectory mapping method provided in an embodiment of the present invention;

[0052] Figure 5 This is a schematic diagram of the structure of a client for an execution trajectory mapping method provided in an embodiment of the present invention;

[0053] Figure 6 This is a schematic diagram of the structure of a trajectory mapping system provided in an embodiment of the present invention. Detailed Implementation

[0054] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0055] Please see Figures 1 to 6 It should be noted that the illustrations provided in this embodiment are only schematic representations of the basic concept of the present invention. Therefore, the illustrations only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0056] This invention, based on a server-client interconnected communication architecture, employs federated learning technology on top of deep learning. A model for aggregation operations is deployed on the server side, while a deep learning model is deployed on the client side for trajectory data processing and training. This invention will detail the method for trajectory mapping performed by the server and client using the above scenario as an example. It should be noted that the models deployed on the server and client are not limited to those based on federated learning technology; any other model for constructing trajectory maps should fall within the scope of this invention.

[0057] See Figure 1 This is a flowchart illustrating a trajectory mapping method provided in an embodiment of the present invention, as shown below. Figure 1 As shown in the figure, this embodiment of the invention illustrates the flow of the server executing the trajectory mapping method.

[0058] Step S101: Randomly send global model parameters to multiple clients.

[0059] In one exemplary embodiment, the server can connect to multiple clients simultaneously; that is, the server can send global model parameters to all connected clients at the same time, or it can send global model parameters to some clients. For ease of control, this invention randomly selects m clients as participants in this round of federated learning. Specifically, the server initializes the global model for trajectory mapping and sends the initialized global model parameters to the clients, enabling the clients to train the model based on these global model parameters and return the corresponding results.

[0060] Step S102: Receive the reference model parameters after iterative training and noise perturbation by the multiple clients.

[0061] According to the above steps, the server further receives reference model parameters returned by m clients. The reference model parameters are model parameters generated by each client after receiving the server's global model parameters, and after training iteration and adding noise perturbation locally.

[0062] Step S103: Aggregate the reference model parameters from multiple clients to obtain new global model parameters.

[0063] Furthermore, the server processes the reference model parameters returned from m clients. In practice, the reference model parameters are aggregated to obtain new global model parameters, which can be calculated using the following formula:

[0064]

[0065] Among them, W t+1 W represents the global model parameters in round t+1, which is the new model parameter. t Let m represent the global model parameters in round t, and ΔW represent the number of clients. i This represents the reference model parameters returned after the i-th client iterative training and noise perturbation.

[0066] Step S104: Send the new global model parameters to the multiple clients so that the multiple clients can perform the next round of iterative training.

[0067] After the aggregation operation in step S103, the server sends the new global model parameters back to the m clients for the next round of iterative training. It should be noted that the number of training rounds is not limited in this embodiment of the invention; it can be any number of rounds depending on the actual trajectory mapping needs.

[0068] In practice, trajectory mapping typically requires multiple rounds of training. To ensure mapping efficiency, a termination of training decision step can be included before executing step S104. (See [link to relevant documentation]). Figure 2The figure shows a flowchart of an iterative control method provided in an embodiment of the present invention. As shown in the figure, the process of the server controlling the number of iterations of the global model is as follows:

[0069] Step S1041: Calculate the validation metrics for the current global model.

[0070] After each iteration, once the server receives the corresponding reference model parameters from m clients, it can further calculate the validation metrics for the new global model parameters obtained after aggregation. These validation metrics may include one or more of the following: average precision, recall, and F1 score.

[0071] In practical implementation, the specific definition of the verification indicators is as follows:

[0072] Precision = TP / (TP + FP)

[0073] Recall = TP / (TP + FN)

[0074] F1=(2×Precision×Recall) / (Precision+Recall)

[0075] Where Precision represents the average precision, Recall represents the recall rate, TP is the positive sample predicted by the model as positive, FN is the negative sample predicted by the model as negative, TN is the negative sample predicted by the model as positive, and FP is the positive sample predicted by the model as negative.

[0076] Step S1042: Calculate the deviation of the verification index across multiple rounds.

[0077] To ensure computational accuracy, in practice, the deviation of validation metrics across five rounds can be further calculated. For example, the deviations in average precision, recall, and F1 score across five rounds can be calculated. This allows for further observation of whether the validation metrics tend to stabilize over the five iterations.

[0078] Step S1043: When the deviation is greater than a preset threshold, the new global model parameters are sent to the multiple clients.

[0079] When the deviation exceeds a preset threshold, it indicates that the validation metrics have still changed significantly after 5 rounds of iteration, and the global model has not yet converged. Further iterations are needed, in which case the new global model is sent to m clients for the next round of iteration. Conversely, if the deviation is less than or equal to the threshold, it means that the validation metrics have not changed significantly after 5 rounds of iteration. Therefore, the global model can be considered stable and converged, and iteration can be stopped. The current global model parameters are then output as the final result.

[0080] As described in the above embodiments, the trajectory mapping method provided by this invention involves a server randomly sending global model parameters to multiple clients; receiving reference model parameters after iterative training and noise perturbation from the multiple clients; aggregating the reference model parameters from the multiple clients to obtain new global model parameters; and sending the new global model parameters to the multiple clients to enable them to perform a new round of iterative training. This invention perturbs the model parameters with noise, increasing the difference between the perturbed data and the original data, thus achieving a stronger privacy protection effect.

[0081] See Figure 3 Figure 1 is a flowchart illustrating another trajectory mapping method provided in an embodiment of the present invention. As shown in the figure, the process of the client executing the trajectory mapping method is as follows:

[0082] Step S201: Update the local model parameters according to the global model parameters sent by the server.

[0083] According to the description of the above embodiment of the server-executed trajectory mapping method, m randomly selected clients receive global model parameters from the server. The embodiment of the present invention specifically describes the operation steps of the selected clients on the global model parameters.

[0084] The client also deploys a local model with corresponding local model parameters. After receiving the global model parameters sent by the server, the client further updates the local model parameters based on the global model parameters, thereby ensuring the consistency of model data information.

[0085] Step S202: Extract features from the collected trajectory data and align it with the real road network map to obtain a trajectory raster feature map. Input the trajectory raster feature map into the model corresponding to the local model parameters for iterative training.

[0086] The client can also collect trajectory data locally. This trajectory data can be the location coordinates of the road network obtained through data acquisition devices, which will not be described in detail in this embodiment of the invention.

[0087] Specifically, the client extracts features from the collected trajectory data, aligns the trajectory raster feature map with the real road network map through segmentation, and inputs it into the local model for local training. Adaptive moment estimation is used for local optimization to solve the regularization optimization problem.

[0088] Step S203: Calculate the difference between the local model parameters before and after iterative training, and perform a pruning operation.

[0089] For each local model parameter, the client can calculate the difference between the data before and after the iteration, and then perform a cropping operation on that difference.

[0090] In practice, the client can perform a clipping operation on the difference using the following formula:

[0091]

[0092] in, This represents the difference after cropping. This represents the difference before cropping, and C represents the preset cropping threshold. The Euclidean norm represents the difference before clipping.

[0093] Step S204: Based on the privacy budget, dataset size, and number of clients participating in training, noise perturbation is applied to the cropped difference to obtain reference model parameters, and the reference model parameters are sent to the server for aggregation calculation.

[0094] To add noise perturbation to the difference, in an exemplary embodiment, the client applies noise perturbation to the difference, and the reference model parameters can be calculated using the following formula:

[0095]

[0096] σ=kLΔS / ε

[0097]

[0098] in, Indicates the reference model parameters. N(0,σ) represents the difference after clipping. 2 ) indicates that the mean is 0 and the variance is σ. 2 The distribution is Gaussian, k is the privacy constant, C is the preset pruning threshold, ε is the privacy budget, δ is the sensitivity, ΔS is the sensitivity, L is the ratio of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

[0099] The client sends the calculated reference model parameters back to the server, so that the server can collect the reference model parameters from all m clients and then perform subsequent operations such as aggregation calculation.

[0100] As described in the above embodiments, the trajectory mapping method provided by this invention involves the client updating its local model parameters based on the global model parameters received from the server; extracting features from collected trajectory data and aligning it with a real road network map to obtain a trajectory raster feature map; inputting the trajectory raster feature map into the model corresponding to the local model parameters for iterative training; calculating the difference between the local model parameters before and after iterative training and performing a pruning operation; and applying noise perturbation to the pruned difference based on the privacy budget, dataset size, and the number of clients participating in the training to obtain reference model parameters, which are then sent to the server for aggregation calculation. This invention adds noise protection to the intermediate results, preventing the problem of reverse-engineering the original data from the intermediate results and ensuring the privacy of crowdsourced data. Furthermore, this invention solves the problem of achieving a good balance between privacy budget and model performance, thereby improving the efficiency of trajectory mapping.

[0101] Through the description of the above method embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] Corresponding to the server-side trajectory mapping method embodiment provided by the present invention, the present invention also provides a server.

[0103] See Figure 4 This is a schematic diagram of the structure of a server for an execution trajectory mapping method provided in an embodiment of the present invention. As shown in the figure, the server includes:

[0104] Sending module 11 is used to randomly send global model parameters to multiple clients;

[0105] The receiving module 12 is used to receive the reference model parameters after iterative training and noise perturbation by the multiple clients;

[0106] Aggregation module 13 is used to aggregate reference model parameters from multiple clients to obtain new global model parameters;

[0107] The iteration module 14 is used to send new global model parameters to the multiple clients so that the multiple clients can perform a new round of iterative training.

[0108] In an exemplary embodiment, the aggregation module 13 aggregates reference model parameters from multiple clients to obtain new global model parameters, as shown in the following formula:

[0109]

[0110] Among them, W t+1 W represents the global model parameters in round t+1, which is the new model parameter. t Let m represent the global model parameters in round t, and ΔW represent the number of clients. i This represents the reference model parameters returned after the i-th client iterative training and noise perturbation.

[0111] In an exemplary embodiment, the iteration module 14 can also control the number of iteration rounds, specifically for: calculating the verification index of the current global model; calculating the deviation of the verification index of multiple rounds; and when the deviation is greater than a preset threshold, sending new global model parameters to the multiple clients.

[0112] Corresponding to the client-side trajectory mapping method embodiment provided by the present invention, the present invention also provides a client.

[0113] See Figure 5 Figure 1 is a schematic diagram of the structure of a client for an execution trajectory mapping method provided in an embodiment of the present invention. As shown in the figure, the client includes:

[0114] Update module 21 is used to update local model parameters based on the global model parameters received from the server;

[0115] Training module 22 is used to extract features from the collected trajectory data and align them with the real road network map to obtain a trajectory grid feature map, and input the trajectory grid feature map into the model corresponding to the local model parameters for iterative training;

[0116] The pruning module 23 is used to calculate the difference between the local model parameters before and after iterative training and to perform pruning operations.

[0117] The noise perturbation module 24 is used to perturb the pruned difference with noise according to the privacy budget, the dataset size and the number of clients participating in the training, to obtain reference model parameters, and send the reference model parameters to the server for aggregation calculation.

[0118] In an exemplary embodiment, the trimming module 23 calculates the difference between the local model parameters before and after the iteration and performs a trimming operation using the following formula:

[0119]

[0120] in, This represents the difference after cropping. This represents the difference before cropping, and C represents the preset cropping threshold. The Euclidean norm represents the difference before clipping.

[0121] In an exemplary embodiment, the noise perturbation module applies noise perturbation to the clipped difference to obtain reference model parameters, which are calculated using the following formula:

[0122]

[0123] σ=kLΔS / ε

[0124]

[0125] in, Indicates the reference model parameters. N(0,σ) represents the difference after clipping. 2 ) indicates that the mean is 0 and the variance is σ. 2 The distribution is Gaussian, k is the privacy constant, C is the preset pruning threshold, ε is the privacy budget, δ is the sensitivity, ΔS is the sensitivity, L is the ratio of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

[0126] Figure 6 This is a schematic diagram of the structure of a trajectory mapping system provided in an embodiment of the present invention, as shown below. Figure 6 As shown, the system includes a server 31 and multiple clients 32. The clients 32 are communicatively connected to the server 31, thereby enabling information interaction between the server 31 and the clients 32. Specifically, the server 31 performs the server-side trajectory mapping steps described in the above method embodiments, and the clients 32 perform the client-side trajectory mapping steps described in the above method embodiments, which will not be repeated here.

[0127] The above-described product can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0128] The electronic devices of this invention exist in various forms, including but not limited to:

[0129] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include: smartphones (e.g., iPhones), multimedia phones, feature phones, and low-end phones, etc.

[0130] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include PDAs, MIDs, and UMPCs, such as the iPad.

[0131] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes: audio and video players (such as iPods), handheld game consoles, e-books, as well as smart toys and portable car navigation devices.

[0132] (4) Server: A device that provides computing services. The components of a server include a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but because they need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0133] (5) Other electronic devices with data interaction functions.

[0134] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0135] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for apparatus or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The apparatus and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0136] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0137] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A trajectory mapping method characterized by, Applied to a server, the method comprises: Randomly sending global model parameters to a plurality of clients; Receiving reference model parameters after the plurality of clients iteratively train and are noise disturbed; Aggregating the reference model parameters of the plurality of clients to obtain new global model parameters; Sending the new global model parameters to the plurality of clients to enable the plurality of clients to perform next-round iterative training, including: updating local model parameters according to the received global model parameters sent by the server; performing feature extraction on collected trajectory data, aligning the trajectory data with a real road network map to obtain a trajectory grid feature map, inputting the trajectory grid feature map into a model corresponding to the local model parameters to perform iterative training; calculating a difference value of the local model parameters before and after the iterative training, and performing clipping operation; according to a privacy budget, a data set size, and a number of clients participating in training, noise disturbing the clipped difference value to obtain reference model parameters, and sending the reference model parameters to the server for aggregation calculation by the server; The clipping operation is calculated by the following formula: ， represents the difference after clipping, represents the difference before clipping, C represents a preset clipping threshold, represents the Euclidean norm of the difference before clipping; The reference model parameters are calculated by the following formula: ， wherein, denotes the reference model parameters, denotes the clipped difference, denotes a Gaussian distribution with mean 0 and variance k is a privacy constant, and C is a pre-set clipping threshold, is the privacy budget, is ΔS is the sensitivity, L is the proportion of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

2. The trajectory mapping method of claim 1, wherein, The aggregation of the reference model parameters of the plurality of clients to obtain the new global model parameters is calculated by the following formula: ， wherein, denotes the global model parameters of the t+1th round, is the new global model parameters, denotes the global model parameters of the tth round, m denotes the number of clients, denotes the reference model parameters returned by the ith client after iterative training and noise disturbance.

3. The trajectory mapping method of claim 1, wherein, The sending of the new global model parameters to the plurality of clients comprises: Calculating a validation index of a current global model; Calculating a deviation of the validation indexes of a plurality of rounds; When the deviation is greater than a preset threshold, sending the new global model parameters to the plurality of clients.

4. A trajectory mapping method characterized by, Applied to a client, the method comprises: Updating local model parameters according to the received global model parameters sent by the server; Performing feature extraction on collected trajectory data, aligning the trajectory data with a real road network map to obtain a trajectory grid feature map, inputting the trajectory grid feature map into a model corresponding to the local model parameters to perform iterative training; Calculating a difference value of the local model parameters before and after the iterative training, and performing clipping operation, which is calculated by the following formula: ， wherein, represents the difference after clipping, represents the difference before clipping, C represents a preset clipping threshold, represents the Euclidean norm of the difference before clipping; According to a privacy budget, a data set size, and a number of clients participating in training, noise disturbing the clipped difference value to obtain reference model parameters, and sending the reference model parameters to the server for aggregation calculation by the server; the reference model parameters are calculated by the following formula: ， ， wherein, denotes the reference model parameters, denotes the clipped difference, denotes a Gaussian distribution with mean 0 and variance k is a privacy constant, and C is a pre-set clipping threshold, is the privacy budget, is ΔS is the sensitivity, L is the proportion of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

5. A server, characterized by Comprise: A sending module configured to randomly send global model parameters to a plurality of clients; A receiving module configured to receive reference model parameters after the plurality of clients iteratively train and are noise disturbed; An aggregation module configured to aggregate the reference model parameters of the plurality of clients to obtain new global model parameters; and A sending module configured to send the new global model parameters to the plurality of clients to enable the plurality of clients to perform next-round iterative training. The iteration module is configured to send new global model parameters to the plurality of clients to enable the plurality of clients to perform a new round of iterative training, and the client execution process comprises: updating local model parameters according to the received global model parameters sent by the server; performing feature extraction on collected trajectory data, aligning the trajectory data with a real road network map to obtain a trajectory grid feature map, inputting the trajectory grid feature map into a model corresponding to the local model parameters to perform iterative training, calculating a difference between the local model parameters before and after the iterative training, and performing clipping operation; performing noise perturbation on the clipped difference according to a privacy budget, a data set size and a number of clients participating in training to obtain reference model parameters, and sending the reference model parameters to the server for aggregated calculation by the server; The clipping operation is calculated by the following formula: ， represents the difference after clipping, represents the difference before clipping, C represents a preset clipping threshold, represents the Euclidean norm of the difference before clipping; The reference model parameters are calculated by the following formula: ， wherein, denotes the reference model parameters, denotes the clipped difference, denotes a Gaussian distribution with mean 0 and variance k is a privacy constant, and C is a pre-set clipping threshold, is the privacy budget, is ΔS is the sensitivity, L is the proportion of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

6. A client, characterized in that, The trajectory mapping system comprises the server of claim 5 and the client of claim 6. The updating module is configured to update local model parameters according to the received global model parameters sent by the server; The training module is configured to perform feature extraction on collected trajectory data, align the trajectory data with a real road network map to obtain a trajectory grid feature map, input the trajectory grid feature map into a model corresponding to the local model parameters to perform iterative training; The clipping module is configured to calculate a difference between the local model parameters before and after the iterative training, and perform clipping operation, and the clipping operation is calculated by the following formula: ， wherein, represents the difference after clipping, represents the difference before clipping, C represents a preset clipping threshold, represents the Euclidean norm of the difference before clipping; The noise perturbation module is configured to perform noise perturbation on the clipped difference according to a privacy budget, a data set size and a number of clients participating in training to obtain reference model parameters, and send the reference model parameters to the server for aggregated calculation by the server; and the reference model parameters are calculated by the following formula: ， wherein, denotes the reference model parameters, denotes the clipped difference, denotes a Gaussian distribution with mean 0 and variance k is a privacy constant, and C is a pre-set clipping threshold, is the privacy budget, is ΔS is the sensitivity, L is the proportion of clients participating in training to the total number of clients, and D is the size of the training dataset held by the client.

7. A trajectory mapping system characterized by, The trajectory mapping system comprises the server of claim 5 and the client of claim 6.

Images