A Vehicle Ad-Hoc Network Resource Allocation Method and System Based on Federated Learning

Abstract

The present invention provides a vehicle networking resource allocation method and system based on federated learning. The method includes: continuously attempting to predict non-standard communication situations based on the driving profile of a target vehicle in the vehicle networking; when the attempt to predict is successful, determining a resource allocation strategy for the non-standard communication situation based on a pre-trained federated learning model; and allocating corresponding communication resources from the communication resource reserve pool of the vehicle networking to the target vehicle based on the resource allocation strategy. Based on the driving profile of the target vehicle in the vehicle networking, the present invention continuously attempts to predict non-standard communication situations. When the attempt to predict is successful, a resource allocation strategy for the non-standard communication situation is determined based on a pre-trained federated learning model, and corresponding communication resources are allocated from the communication resource reserve pool of the vehicle networking to the target vehicle, fully ensuring that the target vehicle can obtain extremely high-quality communication support when a non-standard communication situation occurs and avoiding affecting the vehicles in the vehicle networking.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle - to - everything (V2X) communication, and particularly to a method and system for resource allocation in vehicle - to - everything based on federated learning. Background Art

[0002] When vehicles communicate in a vehicle - to - everything network, in addition to some standard communication scenarios (such as a vehicle obtaining the status of a traffic signal, etc.), there may also be non - standard communication scenarios (such as when an accident occurs on the road and information is sent to inform vehicles approaching the accident location).

[0003] However, non - standard communication scenarios have extremely high requirements for communication quality. Without high - quality communication support after a non - standard communication scenario occurs, it may have a greater impact on vehicles (such as a vehicle approaching the accident location that could have adjusted its route to bypass but did not receive the accident reminder in time and entered the section blocked due to the accident).

[0004] Therefore, there is an urgent need for a solution. Summary of the Invention

[0005] One of the objectives of the present invention is to provide a method for resource allocation in vehicle - to - everything based on federated learning. Based on the driving profile of a target vehicle in the vehicle - to - everything network, continuously attempt to predict non - standard communication scenarios. When a prediction is successful, based on a pre - trained federated learning model, determine the resource allocation strategy for the non - standard communication scenario. Based on the resource allocation strategy, allocate corresponding communication resources from the communication resource reserve pool in the vehicle - to - everything network to the target vehicle, fully ensuring that the target vehicle can obtain extremely high - quality communication support when a non - standard communication scenario occurs, and avoiding affecting the vehicles in the vehicle - to - everything network.

[0006] A method for resource allocation in vehicle - to - everything based on federated learning provided by an embodiment of the present invention includes:

[0007] Continuously attempt to predict non - standard communication scenarios based on the driving profile of a target vehicle in the vehicle - to - everything network;

[0008] When a prediction is successful, based on a pre - trained federated learning model, determine the resource allocation strategy for the non - standard communication scenario; wherein the resource allocation strategy is a strategy indicating how to allocate communication resources to the target vehicle suitable for the non - standard communication scenario;

[0009] Based on the resource allocation strategy, allocate corresponding communication resources from the communication resource reserve pool equipped in the vehicle - to - everything network to the target vehicle; wherein there are a large number of communication resources to be allocated in the communication resource reserve pool.

[0010] Optionally, the method for resource allocation in vehicle - to - everything based on federated learning further includes:

[0011] Obtain the resource allocation history of the communication resource reserve pool within the first time period; wherein, the end time of the first time period is the time when the remaining unallocated resource amount of the communication resource reserve pool does not exceed the first resource amount threshold, and the duration of the first time period is the standard duration value;

[0012] Perform a feature map representation on the resource allocation history to obtain a feature map;

[0013] Identify the trigger events on the feature map; wherein, the event types of the trigger events exist in the standard type library, and the correlation degree between the trigger events and the non-standard communication situations does not exceed the correlation degree threshold;

[0014] Generate a policy generation template based on the event type corresponding to the trigger event, and generate a resource allocation optimization policy according to the trigger event;

[0015] Based on the resource allocation optimization policy, optimize the resource allocation of the communication resource reserve pool within the second time period; wherein, the second time period is after and adjacent to the first time period; the duration of the second time period has a positive correlation with the influence degree corresponding to the event type of the trigger event.

[0016] Optionally, the performing a feature map representation on the resource allocation history to obtain a feature map includes:

[0017] Parse multiple historical items and corresponding generation times in the resource allocation history;

[0018] Based on the generation time, set each historical item on the time axis;

[0019] Use the time axis after all historical items are set as the historical time axis;

[0020] Determine multiple content clusters from the historical time axis; wherein, each content cluster contains more than the number threshold of historical items, and the maximum axial distance between two adjacent historical items in each content cluster does not exceed the distance threshold, and there are more than the pair threshold of historical items in each content cluster with the same standard content relationship;

[0021] Traverse each content cluster in turn;

[0022] Each time during traversal, create a feature slice of the traversed content cluster; wherein, all historical items in the traversed content cluster are associated and displayed on the feature slice;

[0023] Map the feature slice to the multi-modal map of the vehicle network;

[0024] After traversing all content clusters, use the multi-modal map after all feature slices are set as the feature map.

[0025] Optionally, the vehicle network resource allocation method based on federated learning further includes:

[0026] When the remaining unallocated resource amount in the communication resource reserve pool exceeds the second resource amount threshold for a duration exceeding the duration threshold, non-standard communication guidance is performed on the vehicle-to-everything (V2X) network;

[0027] Among them, performing non-standard communication guidance on the V2X network includes:

[0028] Performing event inquiries on the V2X network based on an event inquiry template;

[0029] When the V2X network accepts the inquiry and replies with a target event set, preprocess the target event set;

[0030] Generate a non-standard communication guidance template based on the preprocessed target event set;

[0031] Perform non-standard communication guidance on the V2X network based on the non-standard communication guidance template.

[0032] Optionally, the preprocessing of the target event set includes:

[0033] Traverse each target event in the target event set in sequence;

[0034] Each time during traversal, obtain the communication requirement degree of the traversed target event;

[0035] After traversing all target events, arrange all target events in descending order according to their respective communication requirement degrees to obtain an event sequence;

[0036] Intercept a local sequence from the event sequence; wherein, the local sequence is the front part of the event sequence, and the sum of the required resource amounts corresponding to the event types of each target event in the local sequence is closest to the third resource amount threshold;

[0037] Determine the preprocessed target event set based on the local sequence.

[0038] Optionally, the obtaining of the communication requirement degree of the traversed target event includes:

[0039] Analyze the communication requirement type of the traversed target event; wherein, the communication requirement type includes: active requirement and passive requirement;

[0040] When the communication requirement type is an active requirement, the communication requirement degree of the traversed target event is counted as the product of the first credibility of the related party of the traversed target event and the active weight;

[0041] When the communication requirement type is a passive requirement, the communication requirement degree of the traversed target event is counted as the product of the second credibility of the generating party of the traversed target event and the passive weight.

[0042] An vehicle-to-everything (V2X) network resource allocation system based on federated learning provided by an embodiment of the present invention includes:

[0043] A prediction module, configured to continuously attempt to predict non-standard communication scenarios based on the driving profile of a target vehicle in a vehicle-to-everything (V2X) network;

[0044] A decision-making module, configured to, when a prediction is successfully made, determine a resource allocation strategy for the non-standard communication scenario based on a pre-trained federated learning model; wherein, the resource allocation strategy is a strategy indicating how to allocate communication resources to the target vehicle that is suitable for the non-standard communication scenario;

[0045] An allocation module, configured to allocate corresponding communication resources from a communication resource reserve pool equipped in the V2X network to the target vehicle based on the resource allocation strategy; wherein, there are a large number of communication resources to be allocated in the communication resource reserve pool.

[0046] Optionally, the V2X network resource allocation system based on federated learning further includes:

[0047] An optimization module, configured to:

[0048] Obtain the resource allocation history of the communication resource reserve pool within a first time period; wherein, the end time of the first time period is the time when the remaining unallocated resource amount in the communication resource reserve pool does not exceed a first resource amount threshold, and the duration of the first time period is a standard duration value;

[0049] Perform a feature map representation on the resource allocation history to obtain a feature map;

[0050] Identify a trigger event on the feature map; wherein, the event type of the trigger event exists in a standard type library, and the correlation degree between the trigger event and the non-standard communication scenario does not exceed a correlation degree threshold;

[0051] Generate a resource allocation optimization strategy according to the trigger event based on a policy generation template corresponding to the event type of the trigger event;

[0052] Optimize the resource allocation of the communication resource reserve pool within a second time period based on the resource allocation optimization strategy; wherein, the second time period is after and adjacent to the first time period; the duration of the second time period has a positive correlation with the influence degree corresponding to the event type of the trigger event.

[0053] Optionally, when the optimization module performs a feature map representation on the resource allocation history to obtain a feature map, it includes:

[0054] Parse multiple historical items and corresponding generation times in the resource allocation history;

[0055] Set each historical item on a time axis based on the generation time;

[0056] Use the time axis after all historical items are set as the historical time axis;

[0057] Determine multiple content clusters from the historical timeline; wherein, each content cluster contains more than a threshold number of historical items, and the maximum axial distance between two adjacent historical items in each content cluster does not exceed a spacing threshold, and there are more than a logarithmic threshold pair of historical items in each content cluster having the same standard content relationship;

[0058] Traverse each content cluster in sequence;

[0059] Each time during traversal, create a feature slice of the traversed content cluster; wherein, all historical items in the traversed content cluster are associated and displayed on the feature slice;

[0060] Map the feature slice to the multimodal map of the vehicle networking;

[0061] After traversing all content clusters, use the multimodal map after all feature slices are set as the feature map.

[0062] Optionally, the vehicle networking resource allocation system based on federated learning further includes:

[0063] A guiding module, configured to:

[0064] When the remaining unallocated resource amount in the communication resource reserve pool exceeds a second resource amount threshold for a duration exceeding a duration threshold, perform non-standard communication guidance on the vehicle networking;

[0065] Wherein, performing non-standard communication guidance on the vehicle networking includes:

[0066] Perform event inquiry on the vehicle networking based on an event inquiry template;

[0067] When the vehicle networking accepts the inquiry and replies with a target event set, preprocess the target event set;

[0068] Generate a non-standard communication guidance template based on the preprocessed target event set;

[0069] Perform non-standard communication guidance on the vehicle networking based on the non-standard communication guidance template.

[0070] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in the written specification and the drawings.

[0071] The following further describes the technical solutions of the present invention in detail through the drawings and embodiments. Description of the Drawings

[0072] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, but do not constitute a limitation to the present invention. In the accompanying drawings:

[0073] Figure 1 It is a schematic diagram of a vehicle networking resource allocation method based on federated learning in an embodiment of the present invention;

[0074] Figure 2 It is a schematic diagram of a vehicle networking resource allocation system based on federated learning in an embodiment of the present invention. Detailed implementation manners

[0075] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to explain and illustrate the present invention, and are not used to limit the present invention.

[0076] An embodiment of the present invention provides a vehicle networking resource allocation method based on federated learning, as Figure 1 shown, including:

[0077] S1. Continuously attempt to predict non-standard communication situations based on the driving profile of the target vehicle in the vehicle network;

[0078] In step S1, the driving profile includes: speed, direction, driving trajectory, surrounding environment information, etc.; when predicting non-standard communication situations, based on the driving profile, predict whether the target vehicle will generate non-standard communication situations in the future. For example, if the driving trajectory of the vehicle has been deviating from the normal road and the surrounding environment information of the vehicle indicates that there are a large number of construction vehicles for road repair operations, it means that the road the vehicle is passing through may be under construction, then the predicted non-standard communication situation is that the road may be under construction, and information needs to be released to inform the vehicles that will pass through the construction location soon;

[0079] S2. When the attempt to predict is successful, based on the pre-trained federated learning model, determine the resource allocation strategy for the non-standard communication situation; wherein, the resource allocation strategy is a strategy suitable for the non-standard communication situation indicating how to allocate communication resources to the target vehicle;

[0080] In step S2, the federated learning model is pre-trained in advance using a large amount of communication resource allocation experiences for different non-standard communication situations, so that the pre-trained federated learning model can determine the resource allocation strategy for the non-standard communication situation; the resource allocation experience can be the operation process records of historical experts manually allocating communication resources for non-standard communication situations;

[0081] S3. Based on the resource allocation strategy, allocate corresponding communication resources from the communication resource reserve pool equipped in the vehicle network to the target vehicle; wherein, there are a large number of communication resources to be allocated in the communication resource reserve pool.

[0082] In step S3, the vehicle networking is equipped with a communication resource reserve pool, and there are a large number of communication resources to be allocated in the communication resource reserve pool; the communication resources can be bandwidth, channels, processing capabilities, etc.; after the resource allocation strategy is determined, the corresponding communication resources are allocated to the target vehicle from the communication resource reserve pool based on the resource allocation strategy. For example, a new channel is allocated for the target vehicle to publish information to inform the vehicles that will pass by the construction location soon.

[0083] This application continuously attempts to predict non-standard communication situations based on the driving profile of the target vehicle in the vehicle networking. When the prediction is successful, based on the pre-trained federated learning model, the resource allocation strategy for the non-standard communication situation is determined. Based on the resource allocation strategy, the corresponding communication resources are allocated from the communication resource reserve pool of the vehicle networking to the target vehicle, fully ensuring that the target vehicle can obtain extremely high-quality communication support when a non-standard communication situation occurs, and avoiding affecting the vehicles in the vehicle networking.

[0084] In one embodiment, the vehicle networking resource allocation method based on federated learning further includes:

[0085] S4. Obtain the resource allocation history of the communication resource reserve pool in the first time period; where the end time of the first time period is the time when the remaining unallocated resource amount of the communication resource reserve pool does not exceed the first resource amount threshold, and the duration of the first time period is the standard duration value;

[0086] In step S4, the remaining unallocated resource amount is the number of communication resources remaining unallocated in the communication resource reserve pool; the first resource amount threshold is the quantity threshold of the communication resources, which can be set in advance by technicians; when the remaining unallocated resource amount does not exceed the first resource amount threshold, it means that the number of communication resources remaining unallocated in the communication resource reserve pool is small, and resource allocation optimization of the communication resource reserve pool is required; the standard duration value can be, for example, half an hour; the resource allocation history is the number of communication resources allocated by the communication resource reserve pool in the first time period, the vehicles to which the resources are allocated, and the non-standard communication situations generated by the allocated vehicles, etc.

[0087] S5. Represent the resource allocation history as a feature map to obtain a feature map;

[0088] S6. Identify the trigger events on the feature map; where the event types of the trigger events exist in the standard type library, and the correlation degree between the trigger events and the non-standard communication situations does not exceed the correlation degree threshold;

[0089] In step S6, the triggering event is an event that can be used as a basis for generating a resource allocation optimization strategy; there are a large number of standard types of events in the standard type library that can be used as a basis for generating a resource allocation optimization strategy, such as: allocating communication resources to all vehicles that publish the same road traffic accident prompt information, etc.; the correlation degree represents the degree of content correlation between the triggering event and the non-standard communication situation; the correlation degree threshold is the threshold of the correlation degree; when the correlation degree between the triggering event and the non-standard communication situation does not exceed the correlation degree threshold, it means that the degree of content correlation between the triggering event and the non-standard communication situation is small; ensuring that the event type of the triggering event exists in the standard type library and that the correlation degree between the triggering event and the non-standard communication situation does not exceed the correlation degree threshold can not only make the triggering event a basis for generating a resource allocation optimization strategy, but also prevent the system from re-identifying highly correlated events of the non-standard communication situation and generating unnecessary resource allocation optimization strategies;

[0090] S7. Based on the policy generation template corresponding to the event type of the triggering event, generate a resource allocation optimization strategy according to the triggering event;

[0091] In step S7, there is a policy generation template corresponding to the event type. The policy generation template is a template for the system to generate a resource allocation optimization strategy according to the triggering event. The resource allocation optimization strategy is a strategy that indicates how to optimize the resource allocation of the communication resource reserve pool based on the triggering event. For example, if the triggering event is to allocate communication resources to all vehicles that publish the same road traffic accident prompt information, the generated resource allocation optimization strategy is to allocate communication resources only to the vehicle closest to the accident location among these vehicles, and suspend the allocation of the remaining vehicles;

[0092] S8. Based on the resource allocation optimization strategy, optimize the resource allocation of the communication resource reserve pool in the second time period; wherein, the second time period is after the first time period and adjacent to the first time period; the duration of the second time period is positively correlated with the influence degree corresponding to the event type of the triggering event.

[0093] In step S8, after generating the resource allocation optimization strategy, optimize the resource allocation of the communication resource reserve pool in the second time period based on the resource allocation optimization strategy; the event type of the triggering event corresponds to an influence degree, and the influence degree represents the degree of influence of the triggering event on the communication resource reserve pool, which can be preset by technical personnel in advance; the greater the influence degree, the greater the degree of influence of the triggering event on the communication resource reserve pool, and the longer the time required to optimize the resource allocation of the communication resource reserve pool. Therefore, the duration of the second time period is positively correlated with the influence degree corresponding to the event type of the triggering event. The specific positive correlation relationship can be a multiple relationship greater than 1, or can be set by technical personnel in advance.

[0094] In an embodiment of the present invention, the resource allocation history within a suitable first time period is obtained, represented as a feature map, a trigger event on the feature map is identified, a template is generated based on a policy corresponding to the event type of the trigger event, a resource allocation optimization policy is generated according to the trigger event, and resource allocation optimization is performed on the communication resource reserve pool within a suitable second time period based on the resource allocation optimization policy, fully ensuring the resource allocation ability of the communication resource reserve pool, greatly improving the rationality of resource allocation of the communication resource reserve pool, and further improving the applicability of the system.

[0095] In one embodiment, the representing the resource allocation history as a feature map to obtain a feature map includes:

[0096] S51. Analyze multiple historical items in the resource allocation history and their corresponding generation times;

[0097] In step S51, there are multiple historical items in the resource allocation history, and the corresponding generation time of a historical item is the time when the historical item was generated in history;

[0098] S52. Based on the generation time, set each historical item on the time axis;

[0099] In step S52, when setting a historical item, find the position on the event axis corresponding to the generation time on the time axis and set the historical item at this time axis position;

[0100] S53. Use the time axis after all historical items are set as the historical time axis;

[0101] S54. Determine multiple content clusters from the historical time axis; wherein, each content cluster contains more than a number threshold of historical items, and the maximum on-axis spacing between any two adjacent historical items in each content cluster does not exceed a spacing threshold, and there are more than a logarithm threshold of pairs of historical items in each content cluster having the same standard content relationship;

[0102] In step S54, the number threshold can be, for example, 5; the maximum on-axis spacing is the maximum of the on-axis spacings between any two adjacent historical items in the same content cluster, and the on-axis spacing represents the generation time interval between adjacent historical items; the spacing threshold can be 5 minutes; the logarithm threshold can be, for example, 3; the standard content relationship means that there is a causal relationship, a geographical correlation relationship, etc. between the two historical items in a pair of historical items; that there are more than a logarithm threshold of pairs of historical items in the same content cluster having the same standard content relationship can ensure a relatively high overall correlation degree among the historical items in the same content cluster; in this way, it is necessary to set the historical items in the determined content cluster on the same feature slice;

[0103] S55. Traverse each content cluster in turn;

[0104] S56. Each time during traversal, create a feature slice of the content cluster traversed; wherein, all historical items in the content cluster traversed are associated and displayed on the feature slice.

[0105] In step S56, the feature slice is a virtual slice on which all historical items in the content cluster traversed are associated and displayed. The so-called associated display means that when each historical item is displayed, an association mark is made, for example: making an associated connection, etc.

[0106] S57. Map the feature slice to the multimodal map of the vehicle network.

[0107] In step S57, the multimodal map of the vehicle network can be a three-dimensional map of the cities covered by the vehicle network, on which different modalities of information are set, such as: road information, traffic status information, etc. When setting the feature slice on the multimodal map, find the map area covered by the geographical locations of each historical item on the feature slice on the multimodal map, and set the feature slice on this map area.

[0108] S58. After traversing each content cluster, use the multimodal map after all feature slices are set as the feature map.

[0109] When the embodiment of the present invention represents the resource allocation history with a feature map, content clusters are introduced, feature slices are created based on the content clusters, and the representation of the feature map is completed based on the feature slices and the multimodal map of the vehicle network. All necessary associated historical items are available on the same feature slice, which is convenient for subsequent identification of trigger events, greatly improving the rationality of the feature map representation and the working efficiency of the system.

[0110] In one embodiment, the vehicle network resource allocation method based on federated learning further includes:

[0111] S9. When the remaining unallocated resource amount in the communication resource reserve pool exceeds the second resource amount threshold for a duration exceeding the duration threshold, perform non-standard communication guidance on the vehicle network.

[0112] In step S9, the second resource amount threshold is the quantity threshold of communication resources, which can be set in advance by technicians; the duration threshold can be, for example, 1 hour. When the remaining unallocated resource amount exceeds the second resource amount threshold for a duration exceeding the duration threshold, it indicates that the utilization rate of the communication resource reserve pool is low, and thus non-standard communication guidance needs to be performed on the vehicle network.

[0113] Among them, in S9, performing non-standard communication guidance on the vehicle network includes:

[0114] S91. Based on the event inquiry template, conduct event inquiry on the vehicle network.

[0115] In step S91, the event inquiry template is a template for the system to conduct event inquiries on the vehicle networking. When conducting event inquiries on the vehicle networking, it is asked whether there are events in the vehicle networking that may require non-standard communication;

[0116] S92. When the vehicle networking receives the inquiry and replies with a target event set, preprocess the target event set;

[0117] In step S92, when there are events in the vehicle networking that may require non-standard communication, a target event set will be replied. Each target event in the target event set is an event that may require non-standard communication;

[0118] S93. Generate a non-standard communication guidance template based on the preprocessed target event set;

[0119] In step S93, the non-standard communication guidance template is a template for the system to conduct non-standard communication guidance on the vehicle networking. For example: if a target event in the preprocessed target event set is an interactive communication request between drivers of different vehicles, the non-standard communication guidance template is to guide the drivers of different vehicles to conduct communication exchanges on important matters (such as: sharing road conditions);

[0120] S94. Conduct non-standard communication guidance on the vehicle networking based on the non-standard communication guidance template;

[0121] In step S94, after generating the non-standard communication guidance template, conduct non-standard communication guidance on the vehicle networking based on the non-standard communication guidance template;

[0122] In S92, the preprocessing of the target event set includes:

[0123] S921. Traverse each target event in the target event set in sequence;

[0124] S922. Each time during traversal, obtain the communication requirement degree of the traversed target event;

[0125] In step S922, the communication requirement degree of the target event represents the degree of the need for non-standard communication of the target event;

[0126] S923. After traversing all target events, arrange each target event in descending order according to its respective communication requirement degree to obtain an event sequence;

[0127] S924. Intercept a partial sequence from the event sequence; where the partial sequence is the front part of the event sequence, and the sum of the required resource amounts corresponding to the event types of each target event in the partial sequence is closest to the third resource amount threshold;

[0128] In step S924, the local sequence being the front part of the event sequence means that the local sequence is a certain proportion of the front part of the event sequence. The proportion is defined by the condition that the sum of the required resource amounts is closest to the third resource amount threshold. The third resource amount threshold is the quantity threshold of communication resources, which can be set in advance by technicians. Each event type corresponds to a required resource amount, and the required resource amount is the quantity of communication resources required for the target event of this event type to perform non-standard communication. Ensuring that the sum of the required resource amounts is closest to the third resource amount threshold can make the intercepted local sequence most suitable for the current communication traffic allocation support of the communication resource reserve pool.

[0129] S925. Determine the preprocessed target event set based on the local sequence.

[0130] In step S925, by integrating each target event in the local sequence, the preprocessed target event set is obtained.

[0131] In the embodiment of the present invention, when the utilization degree of the communication resource reserve pool is relatively low, non-standard communication guidance is carried out for the vehicle network, which improves the utilization degree of the communication resource reserve pool and enhances the experience of each vehicle user in the vehicle network. Secondly, the target event set is preprocessed, and based on the preprocessed target event set, a non-standard communication guidance template is generated, which greatly improves the rationality of generating the non-standard communication guidance template, and then improves the suitability of non-standard communication guidance for the vehicle network.

[0132] In one embodiment, in S922, obtaining the communication demand degree of the traversed target event includes:

[0133] S9221. Analyze the communication demand type of the traversed target event. Among them, the communication demand type includes: active demand and passive demand.

[0134] In step S9221, the communication demand type of the target event is divided into two types: active demand and passive demand. Active demand refers to an event in which the related party of the target event actively generates a non-standard communication demand. For example, when drivers between different vehicles perform interactive communication requests, the related party is the driver. Passive demand refers to an event in which the generating party of the target event passively generates a non-standard communication demand. For example, when the traffic information platform updates the progress information of a traffic accident, the generating party is the traffic information platform.

[0135] S9222. When the communication demand type is active demand, the communication demand degree of the traversed target event is calculated as the product of the first credibility of the related party of the traversed target event and the active weight.

[0136] In step S9222, the first credibility of the related party can be calculated as the sum of the total duration, frequency, etc. of the non-standard communications carried out by the related party in history; the active weight is a weight value that can be set in advance by technicians; when the communication requirement type is an active requirement, the communication requirement degree of the target event is calculated as the product of the first credibility of the related party of the target event and the active weight;

[0137] S9223. When the communication requirement type is a passive requirement, the communication requirement degree of the traversed target event is calculated as the product of the second credibility of the generating party of the traversed target event and the passive weight.

[0138] In step S9223, the second credibility of the generating party can be calculated as the sum of the total number of times and frequency of generating the target event by the generating party in history; the passive weight is a weight value that can be set in advance by technicians; when the communication requirement type is a passive requirement, the communication requirement degree of the target event is calculated as the product of the second credibility of the generating party of the target event and the passive weight.

[0139] An embodiment of the present invention provides a vehicle networking resource allocation system based on federated learning, as Figure 2 shown, including:

[0140] A prediction module 1, configured to continuously attempt to predict non-standard communication situations based on the driving profile of a target vehicle in the vehicle networking;

[0141] A decision-making module 2, configured to, when the attempt to predict is successful, make a decision on the resource allocation strategy for the non-standard communication situation based on a pre-trained federated learning model; wherein, the resource allocation strategy is a strategy indicating how to allocate communication resources to the target vehicle suitable for the non-standard communication situation;

[0142] An allocation module 3, configured to allocate corresponding communication resources to the target vehicle from the communication resource reserve pool equipped in the vehicle networking based on the resource allocation strategy; wherein, there are a large number of communication resources to be allocated in the communication resource reserve pool.

[0143] The vehicle networking resource allocation system based on federated learning further includes:

[0144] An optimization module, configured to:

[0145] Obtain the resource allocation history of the communication resource reserve pool in the first time period; wherein, the end moment of the first time period is the moment when the remaining unallocated resource amount of the communication resource reserve pool does not exceed the first resource amount threshold, and the duration of the first time period is a standard duration value;

[0146] Perform a feature map representation on the resource allocation history to obtain a feature map;

[0147] Identify trigger events on the feature map; among them, the event types of the trigger events exist in the standard type library, and the correlation degree between the trigger events and the non-standard communication situations does not exceed the correlation degree threshold;

[0148] Generate a policy generation template based on the policy corresponding to the event type of the trigger event, and generate a resource allocation optimization policy according to the trigger event;

[0149] Based on the resource allocation optimization policy, optimize the resource allocation of the communication resource reserve pool within the second time period; among them, the second time period is after the first time period and adjacent to the first time period; the duration of the second time period has a positive correlation with the influence degree corresponding to the event type of the trigger event.

[0150] The optimization module represents the resource allocation history as a feature map to obtain a feature map, including:

[0151] Parse multiple historical items and corresponding generation times in the resource allocation history;

[0152] Based on the generation time, set each historical item on the time axis;

[0153] Use the time axis after all historical items are set as the historical time axis;

[0154] Determine multiple content clusters from the historical time axis; among them, each content cluster contains more than the number threshold of historical items, and the maximum axial distance between two adjacent historical items in each content cluster does not exceed the distance threshold, and there are more than the logarithm threshold pairs of historical items in each content cluster with the same standard content relationship;

[0155] Traverse each content cluster in turn;

[0156] Each time during traversal, create a feature slice of the traversed content cluster; among them, all historical items in the traversed content cluster are associated and displayed on the feature slice;

[0157] Map the feature slice to the multi-modal map of the vehicle network;

[0158] After traversing each content cluster, use the multi-modal map after all feature slices are set as the feature map.

[0159] The vehicle network resource allocation system based on federated learning further includes:

[0160] A guidance module for:

[0161] When the remaining unallocated resource amount of the communication resource reserve pool exceeds the second resource amount threshold and lasts for a duration exceeding the duration threshold, perform non-standard communication guidance on the vehicle network;

[0162] Among them, performing non-standard communication guidance on the vehicle network includes:

[0163] Perform event inquiries on the Internet of Vehicles based on the event inquiry template;

[0164] When the Internet of Vehicles receives the inquiry and replies with the target event set, preprocess the target event set;

[0165] Generate a non-standard communication guidance template based on the preprocessed target event set;

[0166] Perform non-standard communication guidance on the Internet of Vehicles based on the non-standard communication guidance template.

[0167] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A method for allocating resources in an Internet of Vehicles based on federated learning, characterized in that: include: Based on the driving profile of the target vehicle in the Internet of Vehicles, we continue to try to predict non-standard communication situations, including: when an accident occurs on the road, information is released to inform vehicles that are about to pass the accident location, and when the road may be under construction, information needs to be released to inform vehicles that are about to pass the construction location; When trying to predict, based on the pre-trained federated learning model, decide on the resource allocation strategy for the non-standard communication situation; wherein the resource allocation strategy is a strategy that indicates how to allocate communication resources to the target vehicle in a suitable manner for the non-standard communication situation; Based on the resource allocation strategy, corresponding communication resources are allocated to the target vehicle from the communication resource reserve pool equipped by the Internet of Vehicles; wherein the communication resource reserve pool contains communication resources to be allocated; Also includes: Obtaining a resource allocation history of the communication resource reserve pool in a first time period; wherein the end time of the first time period is the time when the remaining unallocated resource amount of the communication resource reserve pool does not exceed the first resource amount threshold; Performing feature graph representation on the resource allocation history to obtain a feature graph; Identify a trigger event on the feature graph; wherein the event type of the trigger event exists in a standard type library, and the correlation between the trigger event and the non-standard communication situation does not exceed a correlation threshold; A policy generation template corresponding to the event type of the triggering event is used to generate a resource allocation optimization policy according to the triggering event; Based on the resource allocation optimization strategy, resource allocation optimization is performed on the communication resource reserve pool in the second time period; wherein the second time period is after the first time period and adjacent to the first time period; there is a positive correlation between the duration of the second time period and the impact degree corresponding to the event type of the triggering event, and the impact degree represents the degree of influence of the triggering event on the communication resource reserve pool.

2. The method for allocating Internet of Vehicles resources based on federated learning as claimed in claim 1, characterized in that: The step of representing the resource allocation history by a feature graph to obtain the feature graph includes: Parsing multiple history items in the resource allocation history and the corresponding generation time; Based on the time of generation, each historical item is set on the timeline; The timeline after all history items are set is used as the history timeline; Determine multiple content clusters on the historical time axis; wherein each content cluster contains more than a threshold number of historical items, and the maximum on-axis spacing between two adjacent historical items in each content cluster does not exceed a spacing threshold, and historical items in each content cluster that exceed a logarithmic threshold have the same standard content relationship; Traverse each content cluster in turn; Each time the traversal is performed, a feature slice of the traversed content cluster is created; wherein all historical items in the traversed content cluster are associated and displayed on the feature slice; Map feature slices onto a multimodal map of the connected vehicle network; After traversing each content cluster, the multimodal map after all feature slices are set is used as the feature map.

3. The method for allocating Internet of Vehicles resources based on federated learning as claimed in claim 1, characterized in that: Also includes: When the remaining unallocated resources in the communication resource reserve pool exceed the second resource amount threshold for a period of time exceeding the duration threshold, non-standard communication guidance is performed on the Internet of Vehicles; Among them, non-standard communication guidance for the Internet of Vehicles includes: Conduct event inquiries on the Internet of Vehicles based on event inquiry templates; When the IoV receives the query and replies with the target event set, the target event set is pre-processed; Generate a non-standard communication guidance template based on the preprocessed target event set; Based on the non-standard communication guidance template, non-standard communication guidance is carried out for the Internet of Vehicles.

4. The method for allocating Internet of Vehicles resources based on federated learning as claimed in claim 3, characterized in that: The preprocessing of the target event set includes: Traverse each target event in the target event set in turn; During each traversal, the communication demand degree of the traversed target event is obtained; After traversing each target event, arrange each target event from large to small according to their communication demand degree to obtain an event sequence; A local sequence is intercepted from the event sequence; wherein the local sequence is the front part of the event sequence, and the sum of the required resource quantities corresponding to the event types of each target event in the local sequence is closest to the third resource quantity threshold; Based on the local sequence, the preprocessed target event set is determined.

5. The method for allocating Internet of Vehicles resources based on federated learning as claimed in claim 4, characterized in that: The obtaining of the communication demand degree of the traversed target event includes: Analyze the communication demand type of the traversed target event; wherein the communication demand type includes: active demand and passive demand; When the communication demand type is an active demand, the communication demand degree of the traversed target event is calculated as the product of the first credibility of the related party of the traversed target event and the active weight; When the communication demand type is a passive demand, the communication demand degree of the traversed target event is calculated as the product of the second credibility of the generator of the traversed target event and the passive weight.

6. A vehicle network resource allocation system based on federated learning, characterized in that: include: The prediction module is used to continuously try to predict non-standard communication situations based on the driving profile of the target vehicle in the Internet of Vehicles, including: when an accident occurs on the road, information is released to inform vehicles that are about to pass the accident location, and when the road may be under construction, information needs to be released to inform vehicles that are about to pass the construction location; A decision module, for deciding a resource allocation strategy for a non-standard communication situation based on a pre-trained federated learning model when attempting to predict; wherein the resource allocation strategy is a strategy for indicating how to allocate communication resources to a target vehicle in a non-standard communication situation; An allocation module is used to allocate corresponding communication resources from a communication resource reserve pool equipped by the Internet of Vehicles to a target vehicle based on a resource allocation strategy; wherein the communication resource reserve pool contains communication resources to be allocated; Also includes: Optimization modules for: Obtaining a resource allocation history of the communication resource reserve pool in a first time period; wherein the end time of the first time period is the time when the remaining unallocated resource amount of the communication resource reserve pool does not exceed the first resource amount threshold; Performing feature graph representation on the resource allocation history to obtain a feature graph; Identify a trigger event on the feature graph; wherein the event type of the trigger event exists in a standard type library, and the correlation between the trigger event and the non-standard communication situation does not exceed a correlation threshold; A policy generation template corresponding to the event type of the triggering event is used to generate a resource allocation optimization policy according to the triggering event; Based on the resource allocation optimization strategy, resource allocation optimization is performed on the communication resource reserve pool in the second time period; wherein the second time period is after the first time period and adjacent to the first time period; there is a positive correlation between the duration of the second time period and the impact degree corresponding to the event type of the triggering event, and the impact degree represents the degree of influence of the triggering event on the communication resource reserve pool.

7. The vehicle network resource allocation system based on federated learning as claimed in claim 6, characterized in that: The optimization module represents the resource allocation history in a characteristic graph to obtain the characteristic graph, including: Parsing multiple history items in the resource allocation history and the corresponding generation time; Based on the time of generation, each historical item is set on the timeline; The timeline after all history items are set is used as the history timeline; Determine multiple content clusters on the historical time axis; wherein each content cluster contains more than a threshold number of historical items, and the maximum on-axis spacing between two adjacent historical items in each content cluster does not exceed a spacing threshold, and historical items in each content cluster that exceed a logarithmic threshold have the same standard content relationship; Traverse each content cluster in turn; Each time the traversal is performed, a feature slice of the traversed content cluster is created; wherein all historical items in the traversed content cluster are associated and displayed on the feature slice; Map feature slices onto a multimodal map of the connected vehicle network; After traversing each content cluster, the multimodal map after all feature slices are set is used as the feature map.

8. The vehicle network resource allocation system based on federated learning as claimed in claim 6, characterized in that: Also includes: Boot module for: When the remaining unallocated resources in the communication resource reserve pool exceed the second resource amount threshold for a period of time exceeding the duration threshold, non-standard communication guidance is performed on the Internet of Vehicles; Among them, non-standard communication guidance for the Internet of Vehicles includes: Conduct event inquiries on the Internet of Vehicles based on event inquiry templates; When the IoV receives the query and replies with the target event set, the target event set is pre-processed; Generate a non-standard communication guidance template based on the preprocessed target event set; Based on the non-standard communication guidance template, non-standard communication guidance is carried out for the Internet of Vehicles.

Images