CN116011830A - Decision method for assisting edge trusted intelligent service, electronic equipment and storage medium - Google Patents
Decision method for assisting edge trusted intelligent service, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN116011830A CN116011830A CN202211529576.8A CN202211529576A CN116011830A CN 116011830 A CN116011830 A CN 116011830A CN 202211529576 A CN202211529576 A CN 202211529576A CN 116011830 A CN116011830 A CN 116011830A
- Authority
- CN
- China
- Prior art keywords
- model
- vehicle
- target
- intelligent service
- parameters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Traffic Control Systems (AREA)
Abstract
The application relates to the technical field of artificial intelligence, and provides a decision method, electronic equipment and a storage medium for assisting an edge trusted intelligent service. The method comprises the steps of obtaining model parameters by carrying out weighted aggregation on parameters of an auxiliary edge trusted intelligent service decision model of a target vehicle, and obtaining target model parameters by carrying out weighted aggregation according to the model parameters of each target vehicle. The target model parameters are obtained through double aggregation, so that after the target model parameters are issued to each vehicle, the scheme obtained when each vehicle uses the target model obtained based on the update of the target model parameters to make an auxiliary edge credible intelligent service decision is more accurate. In addition, the vehicle does not need to upload data used for training the model, so that the data processing amount of the roadside units can be reduced, and the training speed of the model for making the auxiliary edge credible intelligent service decision can be improved. The training speed and the model precision of the model can be improved, so that the decision efficiency of the auxiliary edge trusted intelligent service can be improved.
Description
Technical Field
The application relates to the technical field of artificial intelligence, in particular to a decision method, electronic equipment and a storage medium for assisting an edge trusted intelligent service.
Background
In intelligent traffic, machine learning is often used for model training to provide intelligent and accurate traffic services. However, in conventional driving assistance schemes, the model is mostly trained on a centralized server, i.e.: all data is centralized for building a unified model. In addition, machine learning requires uploading data, processing by a centralized cloud server, and may cause congestion due to overload or excessive network traffic at the central server. Conventional distributed machine learning techniques require frequent exchanges of gradient and model parameters, which can result in excessive network traffic and high communication costs. Therefore, the training speed of the current model for making the decision of the auxiliary edge trusted intelligent service is low, and the decision efficiency of the auxiliary edge trusted intelligent service is low.
Disclosure of Invention
The embodiment of the application provides a decision-making method, electronic equipment and storage medium for an auxiliary edge trusted intelligent service, which are used for solving the problem that the decision-making efficiency of the auxiliary edge trusted intelligent service is low due to the fact that a traditional auxiliary driving scheme is used currently.
In a first aspect, an embodiment of the present application provides a decision method for an auxiliary edge trusted intelligent service, which is applied to a roadside unit, and the decision method for the auxiliary edge trusted intelligent service includes:
receiving model parameters sent by at least two target vehicles, and carrying out weighted aggregation based on the model parameters of each target vehicle to obtain target model parameters; the model parameters are obtained by weighting and aggregating based on the parameters of the auxiliary edge credible intelligent service decision model of each target vehicle;
and issuing the target model parameters to all initial vehicles in the area to which the roadside units belong, so that each initial vehicle can carry out auxiliary edge credible intelligent service decision by using a target model updated based on the target model parameters.
In one embodiment, before receiving the model parameters sent by the at least two target vehicles, the method includes:
determining the number of training nodes in the area to which the roadside units belong;
acquiring computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area to which the roadside units belong;
and determining the target vehicles with the same number as the training nodes from all the initial vehicles in the area of the roadside unit based on the computing capacity, the computing resources, the midway exit probability and the vehicle trust value of all the initial vehicles in the area of the roadside unit.
In one embodiment, the determining, from all the initial vehicles in the area to which the roadside unit belongs, the same number of target vehicles as required by the training nodes based on the computing power, the computing resources, the midway exit probability and the vehicle trust value of all the initial vehicles in the area to which the roadside unit belongs, includes:
inputting the computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area to which the roadside units belong to a trust value model between the vehicles and the roadside units, and obtaining the vehicle credibility of all the initial vehicles in the area to which the roadside units belong, which is output by the trust value model;
and determining the target vehicles with the same number as that required by the training nodes from all initial vehicles in the area of the roadside unit based on the reliability of each vehicle.
In one embodiment, the weighting and aggregation are performed based on the model parameters of each target vehicle to obtain target model parameters, including:
acquiring a weight value of each target vehicle;
and carrying out weighted aggregation according to the weight values and the model parameters of the target vehicles to obtain the target model parameters.
The embodiment of the application provides a decision method of an auxiliary edge trusted intelligent service, which is applied to a target vehicle, and comprises the following steps:
Training the initial model to obtain a first auxiliary edge trusted intelligent service decision model;
acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles;
the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of each second auxiliary edge trusted intelligent service decision model are weighted and aggregated to obtain model parameters, and the model parameters are sent to a roadside unit;
and receiving target model parameters obtained by the roadside units through weighted aggregation based on the model parameters, and updating the first auxiliary edge trusted intelligent service decision model based on the target model parameters to obtain a target model so as to carry out auxiliary edge trusted intelligent service decision based on the target model.
In one embodiment, the obtaining parameters of the second auxiliary edge trusted intelligent service decision model of the other target vehicle includes:
acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles based on the directed acyclic graph; the directed acyclic graph comprises parameters of an auxiliary edge credible intelligent service decision model of each vehicle and vehicle credibility of each vehicle.
In one embodiment, the vehicle confidence level is determined based on a vehicle trust value.
In one embodiment, the initial models for different types of vehicles include different loss functions.
In a second aspect, an embodiment of the present application provides a decision device for assisting an edge trusted intelligent service, including:
the receiving module is used for receiving model parameters sent by at least two target vehicles, and carrying out weighted aggregation based on the model parameters of each target vehicle to obtain target model parameters; the model parameters are obtained by weighting and aggregating based on the parameters of the auxiliary edge credible intelligent service decision model of each target vehicle;
and the decision module is used for issuing the target model parameters to all initial vehicles in the area to which the roadside units belong so that each initial vehicle can carry out auxiliary edge trusted intelligent service decision by using a target model updated based on the target model parameters.
In one embodiment, an embodiment of the present application provides a decision device for assisting an edge trusted intelligent service, further including:
the training module is used for training the initial model to obtain a first auxiliary edge credible intelligent service decision model;
The acquisition module is used for acquiring parameters of a second auxiliary edge trusted intelligent service decision model of the other target vehicles;
the aggregation module is used for carrying out weighted aggregation on the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of each second auxiliary edge trusted intelligent service decision model to obtain model parameters, and sending the model parameters to a roadside unit;
and the updating module is used for receiving target model parameters obtained by the roadside unit through weighted aggregation based on the model parameters, updating the first auxiliary edge trusted intelligent service decision model based on the target model parameters to obtain a target model, and carrying out auxiliary edge trusted intelligent service decision based on the target model.
In a third aspect, an embodiment of the present application provides an electronic device, including a processor and a memory storing a computer program, where the processor implements the steps of the method for determining an auxiliary edge trusted intelligent service according to the first aspect or the second aspect when the processor executes the program.
In a fourth aspect, an embodiment of the present application provides a storage medium, where the storage medium is a computer readable storage medium, including a computer program, where the computer program when executed by a processor implements the steps of the decision method of the auxiliary edge trusted intelligent service according to the first aspect or the second aspect.
According to the decision method, the electronic device and the storage medium for the auxiliary edge trusted intelligent service, the model parameters are obtained through weighting and aggregating the parameters of the auxiliary edge trusted intelligent service decision model of the target vehicle, and the target model parameters are obtained through weighting and aggregating the model parameters of each target vehicle. The target model parameters are obtained through double aggregation, so that after the target model parameters are issued to each vehicle, the scheme obtained when each vehicle uses the target model obtained based on the update of the target model parameters to make an auxiliary edge credible intelligent service decision is more accurate. In addition, the vehicle does not need to upload data used for training the model, so that the data processing amount of the roadside units can be reduced, and the training speed of the model for making the auxiliary edge credible intelligent service decision can be improved. The training speed and the model precision of the model can be improved, so that the decision efficiency of the auxiliary edge trusted intelligent service can be improved.
Drawings
For a clearer description of the present application or of the prior art, the drawings that are used in the description of the embodiments or of the prior art will be briefly described, it being apparent that the drawings in the description below are some embodiments of the present application, and that other drawings may be obtained from these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow diagram of a decision method for an assisted edge trusted intelligent service according to an embodiment of the present application;
FIG. 2 is a second flow chart of a decision method for an assisted edge trusted intelligent service according to an embodiment of the present application;
FIG. 3 is a third flow chart of a decision method for the auxiliary edge trusted intelligent service according to the embodiment of the present application;
FIG. 4 is a functional block diagram of an embodiment of a decision making device for the assisted edge trusted intelligent service of the present application;
fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions in the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
The decision method, the device, the electronic equipment and the storage medium for the auxiliary edge trusted intelligent service provided by the invention are described in detail below with reference to the embodiment.
Applicants have considered during the course of the inventive application the following aspects:
federal learning (Federated learning, FL) is an emerging artificial intelligence infrastructure aimed at efficient machine learning among multiple participants or computing nodes while protecting terminal data privacy. The FL system is typically composed of a parameter server and a large number of terminal devices. In each round of training, the training node trains the local model through the local data set, and then uploads the model parameters to the parameter server for aggregation.
Because the training node only uploads the local model, not the original data, the data transmission cost can be effectively reduced, and the privacy of a user is protected. The traditional FL model aggregation task is done by a centralized cloud server, however, the network connection between the cloud server and the vehicle may be inefficient and unreliable. Mobile Edge Computing (MEC) transfers cloud services to the Radio Access Network (RAN), and roadside units (RSUs) at the network edge provide communication, computing and storage resources for the vehicle, which may further reduce latency and ensure high bandwidth connections during model training.
There are two main types of FL. The first type is a sync FL, i.e.: after the parameter server receives the local models from all the training nodes, each local model is aggregated into a new global model. However, the FL training node may be a variety of devices having heterogeneous data quality and training capabilities. Devices with poor computing power take a long time to train locally. Meanwhile, the channel condition also affects the transmission time of the model, and a device under the weak wireless channel condition needs a long time to upload the model. Thus, the time of each training round depends on the slowest node.
The second type is an asynchronous FL, in which the parameter server performs global updates immediately after receiving the local model from any training node, and then issues the updated global model to each training node. This mechanism may result in a large consumption of communication resources, the global model is susceptible to outdated local models, and the accuracy fluctuates greatly. Furthermore, the device is not fully trusted and malicious nodes may upload wrong parameters, which will affect the performance of the global FL model.
Based on the above considerations, the applicant has proposed embodiments of the present application.
Fig. 1 is a schematic flow chart of a decision method of an auxiliary edge trusted intelligent service according to an embodiment of the present application. Referring to fig. 1, an embodiment of the present application provides a decision method applied to an auxiliary edge trusted intelligent service of a roadside unit, which may include:
step S100, receiving model parameters sent by at least two target vehicles, and carrying out weighted aggregation based on the model parameters of each target vehicle to obtain target model parameters;
it should be noted that, the implementation subject of the decision-making method of the auxiliary edge trusted intelligent service provided in the embodiments of the present application may be a computer device, such as a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (personal digital assistant, PDA), or the like. The computer device can be used as a Road Side Unit (RSU), wherein an edge server can be deployed on the Road Side Unit for data processing.
A plurality of roadside units may be included in the present application, and a plurality of vehicles may be included in an area to which each roadside unit belongs, wherein each vehicle is equipped with computing resources and communication resources.
The area to which the roadside unit belongs may be an area corresponding to the signal transmission/reception range of the roadside unit.
Further, the vehicles in the present application may be classified into a variety of different types, such as cars, buses, trucks, and the like.
The vehicle may collect data via an On-Board Unit (OBU).
The OBU and the RSU use orthogonal frequency spectrums to carry out data transmission, each device occupies one sub-channel, and interference among different devices is avoided. The RSU acts as an edge node in the edge layer.
The target vehicle may be a roadside unit determined among all vehicles in the area to which it belongs, and used for a vehicle as a training node. The number of target vehicles is at least two, and may be, for example, two, three, ten, one hundred or more.
All vehicles in the area to which the roadside units belong are defined as initial vehicles, and vehicles selected as training nodes from the initial vehicles are defined as target vehicles.
The model parameters sent by the target vehicle can be sent after the target vehicle trains the local model of the target vehicle and after the local model converges, the model parameters are weighted and aggregated with the parameters of the local models of other target vehicles.
The local model in the vehicle may be a model for making an assisted-edge trusted intelligent service decision, which may include autopilot, internet of vehicles, mobile group images, mobile patrol, etc.
The weighted aggregation may be a process of performing weighted average according to parameters of different models and weight values corresponding to the models.
The target model parameters, namely parameters obtained after weighted aggregation of the roadside units according to the model parameters of each target vehicle, can be issued to all initial vehicles in the area to which the roadside units belong, so that each initial vehicle can update a local model through the target model parameters, and an auxiliary edge credible intelligent service decision is made through the updated local model.
Step S200, the target model parameters are issued to all initial vehicles in the area to which the roadside units belong, so that the initial vehicles can use the target model updated based on the target model parameters to make an auxiliary edge credible intelligent service decision.
According to the method and the system, the road side unit can send the target model parameters obtained through weighting aggregation to all initial vehicles located in the area range of the road side unit, so that after the target model parameters are received, each initial vehicle can update a local model according to the target model parameters, and an auxiliary edge credible intelligent service decision is made according to the updated local model.
It may be understood that, in this embodiment, the target model parameters of each roadside unit may be further weighted and aggregated, and the aggregated parameters are respectively issued to all the initial vehicles in the area to which the roadside unit belongs by each roadside unit, so that the parameters of the model received by each initial vehicle are more accurate. The scheme obtained by carrying out auxiliary edge trusted intelligent service decision on each subsequent initial vehicle through the local model updated based on the parameters is more accurate.
According to the decision method for the auxiliary edge trusted intelligent service, the model parameters are obtained by carrying out weighted aggregation on the parameters of the auxiliary edge trusted intelligent service decision model of the target vehicle, and the target model parameters are obtained by carrying out weighted aggregation according to the model parameters of each target vehicle. The target model parameters are obtained through double aggregation, so that after the target model parameters are issued to each vehicle, the scheme obtained when each vehicle uses the target model obtained based on the update of the target model parameters to make an auxiliary edge credible intelligent service decision is more accurate. In addition, the vehicle does not need to upload data used for training the model, so that the data processing amount of the roadside units can be reduced, and the training speed of the model for making the auxiliary edge credible intelligent service decision can be improved. The training speed and the model precision of the model can be improved, so that the decision efficiency of the auxiliary edge trusted intelligent service can be improved.
Fig. 2 is a second flowchart of a decision method of the auxiliary edge trusted intelligent service according to the embodiment of the present application. Referring to fig. 2, in one embodiment, prior to receiving model parameters transmitted by at least two target vehicles, the method includes:
step S1, determining the number of training nodes in an area to which a roadside unit belongs;
as can be appreciated, since one MBS (MoBile Station) requires a fixed number of vehicles as training nodes, the roadside units in this application need to determine the target vehicle as the training node from all the initial vehicles in the area to which the vehicle belongs.
Therefore, it is necessary to determine the number of training nodes required for the area to which the roadside unit belongs.
Specifically, the area selection ratio corresponding to the area to which the roadside unit belongs is determined, and the area selection ratio may be a ratio value set in advance according to actual requirements.
At the same time, the number of vehicles in the area to which the roadside unit belongs is determined.
And determining the number of training nodes in the area of the roadside unit according to the selection proportion of the number of vehicles in the area of the roadside unit and the area corresponding to the area of the roadside unit.
Specifically, the number of vehicles in the area to which the roadside unit belongs may be multiplied by an area selection ratio corresponding to the area to which the roadside unit belongs, and the calculation result may be determined as the number required by the training nodes of the area to which the roadside unit belongs.
To determine the appropriate C in each region r (t)·n r The influence of the vehicle which exits halfway on the round training is small, and meanwhile, the vehicle does not do invalid training. The goal may be formulated as:
E[X r (t)|C r * (t),n r ]＝C·n r
wherein E [ X ] r (t)|C r * (t),n r ]Is the expected value of the number of customers in region r that have not exited at round t; c is a definite proportion.
The number of training nodes in region r is:
wherein S is r (i) Is the set of vehicles selected in the ith round, q r (i) Is S r (i) At C r (i) Is a ratio of (a) to (b). After determining the number of nodes, it is necessary to select an appropriate vehicle to improve the accuracy of the model. Consider the number of vehicles in region r to be ve m Ve then m The number of training nodes of the vehicle is as follows:
by introduction of
An indication vector, a, representing a vehicle selection state i (t) =1 means that the vehicle is selected, otherwise a i (t)＝0。
Modeling the optimization problem as an MDP problem to maximize the accuracy of the global model. The present embodiment solves the above MDP problem using the A3C algorithm, and in a mobile environment, the probability of state transition and rewards is difficult to predict. The MDP problem is defined by tuples < S, A, P, r >, where S is the state set of the system, A is the action set of the system, P is the probability of a state transition, and r is the reward function.
Step S2, acquiring computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area to which the roadside units belong;
Further, the computing power of the vehicle, that is, the power of the vehicle when performing data calculation, may be directly obtained from the vehicle, or the vehicle may have uploaded its computing power to the roadside unit in advance.
The computing resources of the vehicle, that is, the resources that the vehicle has when it is used to perform data computation, may be directly obtained from the vehicle, or the vehicle may have uploaded its computing power to the roadside unit in advance.
The probability of vehicle exit halfway in this application follows a mathematical expectation μ, variance σ 2 Normal distribution of (c):
it will be appreciated that the initial vehicle trust value of each vehicle may be the same value, and may be specifically set according to actual requirements, for example, may be set to 1, 0.9, etc.
Further, the vehicle reliability can be calculated according to the vehicle trust value, and the calculation formula is as follows:
wherein x is i Is the vehicle trust value of vehicle i,
and (3) for the reliability of the verification result of the transaction sent to the DAG by the vehicle after the local model is constructed, wherein mu and gamma are preset parameters, and the lower limit and the change rate of the reliability of the verification result are respectively controlled.
The DAG (Directed acyclic graph ) can record the model parameters generated by local training and related data between vehicles as transactions, and transmit the transactions to surrounding vehicles for synchronization through the eight diagrams protocol.
The transaction in this embodiment may be defined as tr k Then
Representing a vehicle i versus a transaction tr k But not all the verification results in the same group have the same credibility, and the information sent by the vehicle with high trust value is more trustworthy.
If it is
It indicates that vehicle i has not validated the transaction.
Thus the roadside unit can obtain the transaction tr k Reliability set R of (2) k . Based on trusted set R k The aggregate confidence level of the verification result can be calculated:
where p (-) is the probability of occurrence of a certain situation.
Is tr k Is a weighted average of (2); n is n r The number of vehicles in the area to which the roadside units belong; c (C) r (t) selecting a proportion for the region corresponding to the region to which the roadside unit belongs; i is the i-th vehicle; the remaining symbols are as defined for the same symbols in the above formulas.
Wherein, if P (tr k R) exceeds a preset threshold, the roadside unit regards this verification result as true, and generates a positive rating (i.e., +1) for correctly reported result information. Otherwise, a negative rating (i.e., -1) will be generated. Vehicle i for transaction tr k The rating of the verification result is recorded as
Adding this verification result to the trust set Y of vehicle i i And calculates the trust value of vehicle i by:
it will be appreciated that the RSU may obtain conflicting ratings for certain results, and we obtain an offset to the trust value by using a weighted aggregation of these ratings. The offset is between-1 and +1, which is positively correlated to the positive rating ratio of this message. The area RSU j The trust value offset of (2) is calculated as follows:
where m and n are the number of positive ratings and the number of negative ratings, weighted by κ, respectively 1 And kappa (kappa) 2 。κ 1 And kappa (kappa) 2 Calculated using the following formula:
wherein F (-) controls sensitivity to a minority rating population. Finally, will
Put into RSU j Offset set O of (2) j The sum of the absolute values of their trust offsets can be calculated:
S max is S j Upper limit of (2).
In the method, the vehicle trust value can be calculated according to the initial vehicle trust value, and the vehicle trust value can be calculated according to the vehicle trust value in the next training process. Thereby forming a loop iterative process.
Therefore, the method and the device can acquire the computing power, the computing resources, the midway exit probability and the vehicle trust value of all the initial vehicles in the area to which the roadside units belong.
Step S3, determining target vehicles with the same number as the training nodes from all initial vehicles in the area of the roadside unit based on the computing power, the computing resources, the midway exit probability and the vehicle trust value of all the initial vehicles in the area of the roadside unit.
And obtaining the computing capacity, computing resources, midway exit probability and vehicle trust value of all the initial vehicles in the area of the roadside unit, and predicting the vehicle reliability by combining the pre-built and trained trust value model between the vehicles and the roadside unit through the computing capacity, computing resources, midway exit probability and vehicle trust value of all the initial vehicles in the area of the roadside unit.
The trust value model between the vehicle and the roadside unit is used for predicting the credibility of the vehicle according to the computing capability, computing resources, midway exit probability and the trust value of the vehicle.
After the vehicle credibility of each vehicle is obtained, determining the target vehicle used as the training node according to the vehicle credibility and the required quantity of the training nodes.
Further, determining the same number of target vehicles as required by the training nodes from all initial vehicles in the area to which the roadside units belong based on the calculation capacities, the calculation resources, the midway exit probability and the vehicle trust values of all the initial vehicles in the area to which the roadside units belong, including:
step S31, the computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area of the roadside unit are input into a trust value model between the vehicle and the roadside unit, and the vehicle credibility of all initial vehicles in the area of the roadside unit output by the trust value model is obtained;
the trust value model can be further set, wherein the trust value model can be obtained by training the model between the vehicle and the roadside units constructed according to the scene by taking the computing capacity, computing resources, midway exit probability and vehicle trust value of the vehicle as training data and taking the vehicle reliability corresponding to each vehicle as a training label, so that the vehicle reliability of model output can be obtained after the computing capacity, computing resources, midway exit probability and vehicle trust value of all initial vehicles in the area where the roadside units belong are input into the trust value model between the vehicle and the roadside units.
It should be noted that, the training process of the model is not specifically limited in the present application, and an existing training manner may be used for training.
Step S32, determining target vehicles with the same number as that required by the training nodes from all initial vehicles in the area of the roadside units based on the reliability of each vehicle.
After the vehicle credibility of each vehicle is obtained, ascending order sorting or descending order sorting can be performed on the vehicle credibility, the vehicle credibility which is arranged behind the ascending order sorting and has the same number as the number required by the training nodes is determined to be the target credibility, and the vehicle corresponding to the target credibility is determined to be the target vehicle.
Or determining the vehicle credibility which is arranged in the front after descending order and is the same as the number required by the training nodes as the target credibility, and determining the vehicle corresponding to the target credibility as the target vehicle.
According to the method and the device for determining the training node, the target vehicles serving as the training nodes in the area where the roadside units belong can be determined according to the computing capability, computing resources, midway exit probability and vehicle trust value of the vehicles, so that the weighting aggregation can be performed immediately after model parameters of all the target vehicles are received, aggregation is not needed when model parameters of all the vehicles are received, overlong waiting time caused by vehicle heterogeneity in synchronous FL and overlarge communication resource consumption caused by frequent communication in asynchronous FL are avoided, meanwhile, attack of malicious nodes is avoided, accuracy of the model is improved, and decision efficiency of auxiliary edge trusted intelligent service can be improved.
In one embodiment, the weighting aggregation is performed based on model parameters of each target vehicle to obtain target model parameters, including:
step S101, obtaining weight values of all target vehicles;
after the roadside unit receives the model parameters sent by each target vehicle, the global model can be respectively updated through the model parameters of each target vehicle, the accuracy of the global model is calculated, the accuracy of the model updated based on the model parameters of each target vehicle is used as the contribution degree of each target vehicle to the model in the training of the present wheel, and the contribution degree of the target vehicle to the model is determined as the weight value of the target vehicle in the global weighted aggregation process.
Step S102, weighting and aggregation are carried out according to the weight values and the model parameters of the target vehicles, and the target model parameters are obtained.
After the model parameters of each target vehicle and the weight values of each target vehicle are obtained, the model parameters of each target vehicle can be combined with the corresponding weight values to carry out weighted aggregation, and the obtained parameters are determined as the target model parameters after the calculation is completed. The specific calculation process can be realized by the following formula:
wherein C is i The contribution degree of the model in the present round iteration is the weight value of the vehicle; omega i,k Model parameters of the vehicle i in the kth iteration; z is the number of target vehicles.
According to the method and the device, after the model parameters sent by each target vehicle are obtained, the model parameters are weighted and aggregated by combining the contribution values of each target vehicle to the model to obtain the target model parameters, aggregation is not needed when the model parameters of all vehicles are received, the problems of overlong waiting time caused by vehicle heterogeneity in the synchronous FL and overlarge communication resource consumption caused by frequent communication in the asynchronous FL are avoided, and the decision efficiency of the auxiliary edge trusted intelligent service can be improved.
Fig. 3 is a third flow chart of a decision method of the auxiliary edge trusted intelligent service according to the embodiment of the present application. Referring to fig. 3, the embodiment of the application further provides a decision method applied to the auxiliary edge trusted intelligent service of the vehicle, where the decision method of the auxiliary edge trusted intelligent service includes:
step S300, training an initial model to obtain a first auxiliary edge credible intelligent service decision model;
the decision method of the auxiliary edge trusted intelligent service in the embodiment can be applied to a target vehicle and can also be applied to other vehicles except the target vehicle. The target vehicle is a vehicle which is determined by the roadside unit from all initial vehicles in the area where the roadside unit belongs and is used as a training node.
Wherein each vehicle is equipped with computing and communication resources. The vehicle collects data through the On-Board Unit. The OBU and the RSU use orthogonal frequency spectrums to carry out data transmission, each device occupies one sub-channel, and interference among different devices is avoided.
An initial model for making an auxiliary edge trusted intelligent service decision, previously constructed from the actual scenario, may be included in the target vehicle. The structure of the initial model is not particularly limited in this application.
Therefore, the target vehicle can acquire training data required for training the initial model, and iteratively train the initial model through the acquired training data, and obtain an auxiliary edge trusted intelligent service decision model of the vehicle after training is completed.
It should be noted that, the initial model constructed in the different types of vehicles of the present application may include different loss functions to provide personalized driving assistance services.
After the first auxiliary edge trusted intelligent service decision model is obtained, the target vehicle can broadcast the parameters trained in the first auxiliary edge trusted intelligent service decision model to the DAG, so that the DAG can record the model parameters generated by local training and related data between vehicles as transactions, and the transactions are transmitted to surrounding vehicles for synchronization through the eight diagrams protocol.
The constituent elements of the DAG are transactions, not blocks, so that the block packing time need not be considered. The consistency of the DAG is ensured by the verification of the previous transaction by the subsequent transaction. Namely: to add a transaction, the vehicle must verify a certain number of previous transactions according to certain rules. Once the vehicle adds a new transaction to the DAG, and becomes a reliable transaction after two successful verifications. The verification method allows the DAG to asynchronously and concurrently write a plurality of transactions, and finally forms a topological tree structure, thereby greatly improving the expandability.
The weight of each transaction is proportional to the computational resources consumed by the vehicle and the accuracy of the model, and is expressed as:
wherein d is i Is the data size of the vehicle local training that submitted the transaction,
is the cumulative data size s of the vehicle submitting the transaction for local aggregation i Is the duration of the present training, acc (tr i ) Is the model accuracy.
Further calculating an accumulated weight for the transaction:
wherein ΔAcc (tr) i (l))＝Acc(tr i (l))-W(tr i (l))；
L is commit transaction tr i The number of other transactions previously submitted by the vehicle of (t).
The DAG can record the model parameters generated by local training and related data between vehicles as transactions, and transmit the transactions to surrounding vehicles through the eight diagrams protocol for synchronization; the related data between vehicles may include vehicle trustworthiness.
The process of adding transactions to a DAG includes three steps. In a first step, the vehicle selects some prompts according to an algorithm or randomly. In a second step, the vehicle verifies the accuracy of the selected prompt. In the last step, a new tip will be built and released on the DAG, including the parameters of the local model aggregated by the vehicle. Through these three steps, each new tip requires two unverified transactions to verify its legitimacy. The vehicle verifies the tips by calculating the accuracy of the model corresponding to the tips on its own dataset. The vehicle then appends the hashes of the two validated hints to the new hint, adds the new hint to the DAG, and broadcasts it to the nodes of the DAG. Since there are typically more than two unverified tips, we use the Markov chain Monte Carlo method to model the probability of each step to one unverified tip. Transactions with high probability of confirmation can be selected by a weighting step using a "high cumulative weight" strategy. The simulation is based on: a unidirectional connection is formed between transactions.
Since DAG is a graph structure, a model that has not been approved by other models for a long period of time (called an obsolete model) will always be considered a tip. If not handled in time, the obsolete model will persist in the DAG for a long period of time and the probability of nodes selecting the obsolete model for aggregation will increase. We therefore set a waiting time called "freshness time". Each tip will experience one of three conditions throughout the life cycle, and the tip in the third condition will be judged to be out of date and not selected by the node. Most tips that are determined to be out of date are less accurate models and are removed from the DAG.
Transactions issued by malicious nodes are typically less accurate on the test set than transactions issued by other nodes. Thus, the probability that an anomalous transaction is approved by a subsequently issued transaction is much less than a normal transaction. During the FL process, abnormal transactions are isolated, the impact of which is minimized. In addition, nodes with excessive outdated models can be detected as abnormal nodes and then react to the abnormal nodes.
Local training and local aggregation are performed on the vehicle. The loss function of vehicle i is defined as it is in the sample dataset d m,i The difference between the predicted value and the actual value is expressed asThe method comprises the following steps:
where ω is a parameter vector and f () is a user-specified loss function, such as linear regression, logistic regression, support Vector Machine (SVM).
In this embodiment, the vehicle may be a different type of vehicle, for example, ve 1 ，ve 2 ，ve 3 ，...，ve m Etc.
Thus, for the vehicle type ve m The loss function of the auxiliary driving model FL task on all data sets can be defined as:
wherein the FL target is to find the optimal parameter vector ω * To minimize F (ω), i.e.:
ω * ＝argmin ω F m (ω)
vehicle i based on local data D m,i After training the model, the update of the model parameters is expressed as:
wherein η is a fixed learning rate.
Step S400, acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles;
further, each target vehicle determined by the roadside unit in addition to the current target vehicle
Other target vehicles in the vehicle also need to train the local model, and the auxiliary edge trusted intelligent service decision model obtained by training the other target vehicles can be determined as a second model
An auxiliary edge trusted intelligent service decision model.
After the other target vehicles complete the local model training, the parameters of the second auxiliary edge trusted intelligent service decision model are broadcast into the DAG.
Thus, the target vehicle may obtain parameters of the second auxiliary edge 0 trusted intelligent service decision model of the other target vehicle from the DAG.
Step S500, carrying out weighted aggregation on the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of each second auxiliary edge trusted intelligent service decision model to obtain model parameters, and sending the model parameters to a roadside unit;
after obtaining the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of the second auxiliary 5 auxiliary edge trusted intelligent service decision models, each target can be obtained from the DAG
Vehicle credibility of the vehicle.
Further, local weighted aggregation is performed according to the parameters of each target vehicle and the corresponding vehicle credibility, and the aggregated parameters are determined to be model parameters.
Further, the model parameters are sent to the roadside unit for the roadside unit to perform according to the model parameters sent by the current 0 target vehicle and the model parameters sent by other target vehicles
And carrying out global weighted aggregation to obtain target model parameters and issuing the target model parameters to all initial vehicles in the affiliated area.
Step S600, receiving target model parameters obtained by weighting and aggregating roadside units based on the model parameters, and updating a first auxiliary edge trusted intelligent service decision based on the target model parameters
The model is obtained to obtain a target model, and the auxiliary edge trusted intelligent service block 5 is conducted based on the target model.
Further, the target vehicle can receive target model parameters obtained by weighting and aggregating the roadside units based on the model parameters, update the local first auxiliary edge trusted intelligent service decision model through the target model parameters, and determine the model after parameter update as the target model.
After the target model is obtained, the target vehicle or other vehicles can input data required by the auxiliary edge credible intelligent service decision into the target model, and after the target model completes prediction through the input data, an auxiliary driving scheme output by the target model is obtained. Such as an autopilot scenario, an optimal navigation route, etc.
According to the decision method for the auxiliary edge trusted intelligent service, the model parameters are obtained by carrying out weighted aggregation on the parameters of the auxiliary edge trusted intelligent service decision model of the target vehicle, and the target model parameters are obtained by carrying out weighted aggregation according to the model parameters of each target vehicle. The target model parameters are obtained through double aggregation, so that after the target model parameters are issued to each vehicle, the scheme obtained when each vehicle uses the target model obtained based on the update of the target model parameters to make an auxiliary edge credible intelligent service decision is more accurate. In addition, the vehicle does not need to upload data used for training the model, so that the data processing amount of the roadside units can be reduced, and the training speed of the model for making the auxiliary edge credible intelligent service decision can be improved. The training speed and the model precision of the model can be improved, so that the decision efficiency of the auxiliary edge trusted intelligent service can be improved.
Further, obtaining parameters of a second auxiliary edge trusted intelligent service decision model of the other target vehicle includes:
step S401, acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles based on the directed acyclic graph; the directed acyclic graph comprises parameters of the auxiliary edge credible intelligent service decision model of each vehicle and the vehicle credibility of each vehicle.
When the parameters of the second auxiliary edge trusted intelligent service decision model of other target vehicles are acquired, the DAG of the directed acyclic graph is included in the application, and the parameters broadcast by each target vehicle after the local model training and the vehicle credibility of each target vehicle are included in the DAG.
Thus, the current target vehicle may directly obtain parameters of the second auxiliary edge trusted intelligent service decision model of the other target vehicle from the DAG. Meanwhile, the vehicle credibility of other target vehicles and the current target vehicle can be acquired.
In this application, the vehicle reliability may be determined by the vehicle trust value, and the specific determining process may refer to the calculating process of the vehicle trust value and the vehicle reliability, which is not described herein.
According to the embodiment, the parameters after the local training in each target vehicle can be obtained under the DAG structure, so that local aggregation is carried out according to the parameters, the problems of overlong waiting time caused by vehicle heterogeneity in the synchronous FL and overlarge communication resource consumption caused by frequent communication in the asynchronous FL can be avoided, model training speed and model accuracy can be improved, and therefore decision efficiency of the auxiliary edge trusted intelligent service can be improved.
Further, the application also provides a decision device for assisting the edge trusted intelligent service.
Referring to fig. 4, fig. 4 is a schematic functional block diagram of an embodiment of a decision device for assisting an edge trusted intelligent service according to the present application.
The decision device of the auxiliary edge trusted intelligent service comprises:
the receiving module 410 is configured to receive model parameters sent by at least two target vehicles, and perform weighted aggregation based on the model parameters of each target vehicle to obtain target model parameters; the model parameters are obtained by weighting and aggregating based on the parameters of the auxiliary edge credible intelligent service decision model of each target vehicle;
the decision module 420 is configured to issue the target model parameters to all initial vehicles in the area to which the roadside unit belongs, so that each initial vehicle can use a target model updated based on the target model parameters to make an auxiliary edge trusted intelligent service decision.
The decision device of the auxiliary edge trusted intelligent service provided by the embodiment of the application obtains model parameters by carrying out weighted aggregation on the parameters of the auxiliary edge trusted intelligent service decision model of the target vehicle, and obtains the target model parameters by carrying out weighted aggregation according to the model parameters of each target vehicle. The target model parameters are obtained through double aggregation, so that after the target model parameters are issued to each vehicle, the scheme obtained when each vehicle uses the target model obtained based on the update of the target model parameters to make an auxiliary edge credible intelligent service decision is more accurate. In addition, the vehicle does not need to upload data used for training the model, so that the data processing amount of the roadside units can be reduced, and the training speed of the model for making the auxiliary edge credible intelligent service decision can be improved. The training speed and the model precision of the model can be improved, so that the decision efficiency of the auxiliary edge trusted intelligent service can be improved.
In one embodiment, the receiving module 410 is specifically configured to:
acquiring a weight value of each target vehicle;
and carrying out weighted aggregation according to the weight values and the model parameters of the target vehicles to obtain the target model parameters.
In one embodiment, the receiving module 410 further comprises a first determining unit for:
determining the number of training nodes in the area to which the roadside units belong;
acquiring computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area to which the roadside units belong;
and determining the target vehicles with the same number as the training nodes from all the initial vehicles in the area of the roadside unit based on the computing capacity, the computing resources, the midway exit probability and the vehicle trust value of all the initial vehicles in the area of the roadside unit.
In an embodiment, the first determining unit further comprises a second determining unit for:
inputting the computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area to which the roadside units belong to a trust value model between the vehicles and the roadside units, and obtaining the vehicle credibility of all the initial vehicles in the area to which the roadside units belong, which is output by the trust value model;
And determining the target vehicles with the same number as that required by the training nodes from all initial vehicles in the area of the roadside unit based on the reliability of each vehicle.
The decision device of the auxiliary edge trusted intelligent service comprises:
the training module is used for training the initial model to obtain a first auxiliary edge credible intelligent service decision model;
the acquisition module is used for acquiring parameters of a second auxiliary edge trusted intelligent service decision model of the other target vehicles;
the aggregation module is used for carrying out weighted aggregation on the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of each second auxiliary edge trusted intelligent service decision model to obtain model parameters, and sending the model parameters to a roadside unit;
and the updating module is used for receiving target model parameters obtained by the roadside unit through weighted aggregation based on the model parameters, updating the first auxiliary edge trusted intelligent service decision model based on the target model parameters to obtain a target model, and carrying out auxiliary edge trusted intelligent service decision based on the target model.
In one embodiment, the acquisition module further comprises an acquisition unit for:
Acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles based on the directed acyclic graph; the directed acyclic graph comprises parameters of an auxiliary edge credible intelligent service decision model of each vehicle and vehicle credibility of each vehicle.
Fig. 5 illustrates a physical schematic diagram of an electronic device, as shown in fig. 5, which may include: processor 510, communication interface (Communication Interface) 520, memory 530, and communication bus 540, wherein processor 510, communication interface 520, memory 530 complete communication with each other through communication bus 540. Processor 510 may invoke a computer program in memory 530 to perform the steps of a decision method for assisting an edge trusted intelligent service, including, for example:
receiving model parameters sent by at least two target vehicles, and carrying out weighted aggregation based on the model parameters of each target vehicle to obtain target model parameters; the model parameters are obtained by weighting and aggregating based on the parameters of the auxiliary edge credible intelligent service decision model of each target vehicle;
and issuing the target model parameters to all initial vehicles in the area to which the roadside units belong, so that each initial vehicle can carry out auxiliary edge credible intelligent service decision by using a target model updated based on the target model parameters.
Or, include:
training the initial model to obtain a first auxiliary edge trusted intelligent service decision model;
acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles;
the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of each second auxiliary edge trusted intelligent service decision model are weighted and aggregated to obtain model parameters, and the model parameters are sent to a roadside unit;
and receiving target model parameters obtained by the roadside units through weighted aggregation based on the model parameters, and updating the first auxiliary edge trusted intelligent service decision model based on the target model parameters to obtain a target model so as to carry out auxiliary edge trusted intelligent service decision based on the target model.
Further, the logic instructions in the memory 530 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, embodiments of the present application further provide a storage medium, where the storage medium is a computer readable storage medium, where the computer readable storage medium stores a computer program, where the computer program is configured to cause a processor to execute the steps of the method provided in the foregoing embodiments, where the method includes:
receiving model parameters sent by at least two target vehicles, and carrying out weighted aggregation based on the model parameters of each target vehicle to obtain target model parameters; the model parameters are obtained by weighting and aggregating based on the parameters of the auxiliary edge credible intelligent service decision model of each target vehicle;
and issuing the target model parameters to all initial vehicles in the area to which the roadside units belong, so that each initial vehicle can carry out auxiliary edge credible intelligent service decision by using a target model updated based on the target model parameters.
Or, include:
training the initial model to obtain a first auxiliary edge trusted intelligent service decision model;
acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles;
the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of each second auxiliary edge trusted intelligent service decision model are weighted and aggregated to obtain model parameters, and the model parameters are sent to a roadside unit;
And receiving target model parameters obtained by the roadside units through weighted aggregation based on the model parameters, and updating the first auxiliary edge trusted intelligent service decision model based on the target model parameters to obtain a target model so as to carry out auxiliary edge trusted intelligent service decision based on the target model.
The computer readable storage medium may be any available medium or data storage device that can be accessed by a processor including, but not limited to, magnetic memory (e.g., floppy disks, hard disks, magnetic tape, magneto-optical disks (MOs), etc.), optical memory (e.g., CD, DVD, BD, HVD, etc.), and semiconductor memory (e.g., ROM, EPROM, EEPROM, nonvolatile memory (NAND FLASH), solid State Disk (SSD)), etc.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and are not limiting thereof; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (10)
1. The decision method of the auxiliary edge trusted intelligent service is characterized by being applied to roadside units, and comprises the following steps of:
receiving model parameters sent by at least two target vehicles, and carrying out weighted aggregation based on the model parameters of each target vehicle to obtain target model parameters; the model parameters are obtained by weighting and aggregating based on the parameters of the auxiliary edge credible intelligent service decision model of each target vehicle;
and issuing the target model parameters to all initial vehicles in the area to which the roadside units belong, so that each initial vehicle can carry out auxiliary edge credible intelligent service decision by using a target model updated based on the target model parameters.
2. The method for decision making for a secondary edge trusted intelligent service according to claim 1, wherein prior to receiving model parameters sent by at least two target vehicles, comprising:
determining the number of training nodes in the area to which the roadside units belong;
acquiring computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area to which the roadside units belong;
And determining the target vehicles with the same number as the training nodes from all the initial vehicles in the area of the roadside unit based on the computing capacity, the computing resources, the midway exit probability and the vehicle trust value of all the initial vehicles in the area of the roadside unit.
3. The method for decision-making of an assisted border trusted intelligent service according to claim 2, wherein the determining, based on the computing power, computing resources, probability of halfway exit, and vehicle trust values of all initial vehicles in the area to which the roadside unit belongs, the same number of target vehicles as required by training nodes from all initial vehicles in the area to which the roadside unit belongs includes:
inputting the computing capacity, computing resources, midway exit probability and vehicle trust values of all initial vehicles in the area to which the roadside units belong to a trust value model between the vehicles and the roadside units, and obtaining the vehicle credibility of all the initial vehicles in the area to which the roadside units belong, which is output by the trust value model;
and determining the target vehicles with the same number as that required by the training nodes from all initial vehicles in the area of the roadside unit based on the reliability of each vehicle.
4. The method for decision-making for an assisted edge trusted intelligent service according to claim 1, wherein said weighting and aggregating based on model parameters of each of said target vehicles to obtain target model parameters comprises:
acquiring a weight value of each target vehicle;
and carrying out weighted aggregation according to the weight values and the model parameters of the target vehicles to obtain the target model parameters.
5. A decision method for an auxiliary edge trusted intelligent service, which is applied to a target vehicle, the decision method for the auxiliary edge trusted intelligent service comprising:
training the initial model to obtain a first auxiliary edge trusted intelligent service decision model;
acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles;
the parameters of the first auxiliary edge trusted intelligent service decision model and the parameters of each second auxiliary edge trusted intelligent service decision model are weighted and aggregated to obtain model parameters, and the model parameters are sent to a roadside unit;
and receiving target model parameters obtained by the roadside units through weighted aggregation based on the model parameters, and updating the first auxiliary edge trusted intelligent service decision model based on the target model parameters to obtain a target model so as to carry out auxiliary edge trusted intelligent service decision based on the target model.
6. The method for determining the auxiliary edge trusted intelligent service according to claim 5, wherein the obtaining parameters of the second auxiliary edge trusted intelligent service determining model of the other target vehicle comprises:
acquiring parameters of a second auxiliary edge trusted intelligent service decision model of other target vehicles based on the directed acyclic graph; the directed acyclic graph comprises parameters of an auxiliary edge credible intelligent service decision model of each vehicle and vehicle credibility of each vehicle.
7. The method of claim 6, wherein the vehicle confidence level is determined based on a vehicle confidence value.
8. The method of claim 5, wherein the initial models for different types of vehicles include different loss functions.
9. An electronic device comprising a processor and a memory storing a computer program, characterized in that the processor implements the steps of the decision method of the auxiliary edge trusted intelligent service of any one of claims 1 to 8 when the computer program is executed.
10. A storage medium, which is a computer readable storage medium comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the decision method of the auxiliary edge trusted intelligent service of any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211529576.8A CN116011830A (en) | 2022-11-30 | 2022-11-30 | Decision method for assisting edge trusted intelligent service, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211529576.8A CN116011830A (en) | 2022-11-30 | 2022-11-30 | Decision method for assisting edge trusted intelligent service, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116011830A true CN116011830A (en) | 2023-04-25 |
Family
ID=86028755
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211529576.8A Pending CN116011830A (en) | 2022-11-30 | 2022-11-30 | Decision method for assisting edge trusted intelligent service, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116011830A (en) |
-
2022
- 2022-11-30 CN CN202211529576.8A patent/CN116011830A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20180372502A1 (en) | Road traffic management | |
| CN112037930B (en) | Infectious disease prediction equipment, method, device and storage medium | |
| CN111294812B (en) | Resource capacity-expansion planning method and system | |
| CN109788489A (en) | A kind of base station planning method and device | |
| CN110445788B (en) | Content-oriented trust evaluation system and method under vehicle-mounted ad hoc network environment | |
| JP2023538141A (en) | Traffic jam detection method, device, electronic device and storage medium | |
| CN104346925A (en) | Method and system for predicting running time | |
| CN115708343A (en) | Method for collecting data from a set of vehicles | |
| CN116541106A (en) | Computing task unloading method, computing device and storage medium | |
| CN116669111A (en) | Mobile edge computing task unloading method based on blockchain | |
| EP4184361A1 (en) | Federated learning method, apparatus and system, electronic device and storage medium | |
| CN111465057B (en) | Edge caching method and device based on reinforcement learning and electronic equipment | |
| CN111327473B (en) | Network regulation and control method, device, network regulation and control system and electronic equipment | |
| CN116011830A (en) | Decision method for assisting edge trusted intelligent service, electronic equipment and storage medium | |
| CN116663675A (en) | Block chain enabling federal learning system suitable for edge car networking | |
| CN113852933B (en) | Relay node selection method, device and equipment of Internet of vehicles network and storage medium | |
| CN115640852B (en) | Federal learning participation node selection optimization method, federal learning method and federal learning system | |
| CN115941332A (en) | Vehicle credibility measuring method based on block chain and recommendation trust | |
| CN110942178B (en) | Charging pile recommendation method based on link prediction method of resource allocation index | |
| CN114566048A (en) | Traffic control method based on multi-view self-adaptive space-time diagram network | |
| CN113342474A (en) | Method, device and storage medium for forecasting customer flow and training model | |
| CN116744319B (en) | Road side unit deployment method, device, equipment and readable storage medium | |
| CN113923605B (en) | Distributed edge learning system and method for industrial internet | |
| CN115544870B (en) | Road network approach detection method, device and storage medium | |
| CN112506673B (en) | Intelligent edge calculation-oriented collaborative model training task configuration method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |