Intelligent scheduling and prediction method and system for star large model by combining low-altitude economy with intelligent calculation center

Abstract

The invention provides an intelligent scheduling and prediction method and system for a star large model by combining low-altitude economy with an intelligent calculation center, and the method comprises the steps: collecting radio wave data transmitted between a VLC transmitting end installed on each vehicle and a VLC receiving end installed at a traffic intersection, and forming a radio wave constellation; performing data classification on the radio wave constellation, and separating radio wave data received by each VLC receiving end according to a time sequence; acquiring traffic flow data distributed according to a time sequence by adopting a star large model according to the classified radio wave data; according to the traffic flow data of each traffic intersection, an optimal path algorithm is adopted to plan an optimal path for the target vehicle to arrive at the destination, and meanwhile, the congestion trend of arriving at the next traffic intersection is predicted; and after the planning result and the prediction result are integrated, report information is generated through the intelligent calculation center and transmitted to the target vehicle. According to the method, the star large model and the intelligent calculation center are used for intelligent scheduling and prediction of the vehicle network, and the effects of path planning and congestion prediction are improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent scheduling and prediction of vehicle networks, and specifically relates to a method and system for intelligent scheduling and prediction of a large star model using low-altitude economy combined with an intelligent computing center. Background Art

[0002] As more and more vehicles are connected to the Internet of Things (IoT), traditional VANETs are transforming into Internet of Vehicles (IoV). The communication quality of ground vehicles is the foundation of IoV, which is the data basis for intelligent scheduling and traffic prediction. The current application of visible light communication (VLC) based on car lights in IoV means that when communicating through car lights, multi-user access in IoV will cause mutual interference, and make subsequent intelligent scheduling and prediction operations based on VLC signals more complicated. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a method and system for intelligent scheduling and prediction of a large star model using low-altitude economy combined with an intelligent computing center.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides an intelligent scheduling and prediction method combining a large star model with an intelligent computing center, including:

[0006] Step 1: Collect radio wave data transmitted between the VLC transmitter installed on each vehicle and the VLC receiver installed at the traffic intersection to form a radio wave constellation;

[0007] Step 2: Use the K-Means algorithm to classify the radio wave constellation data and separate the radio wave data received by each VLC receiver according to the time series;

[0008] Step 3: Use the Xingchen big model to obtain the traffic flow data of each VLC receiving end corresponding to the traffic intersection distributed in time series according to the classified radio wave data;

[0009] Step 4: Based on the traffic flow data of each traffic intersection, the optimal path algorithm is used to plan the optimal path for the target vehicle to reach the destination, and the congestion trend at the next traffic intersection is predicted;

[0010] Step 5: After the planning results and prediction results are integrated, they are uploaded to the relay satellite layer through the Beidou chip deployed in the Internet of Vehicles, and then fed back to the ground intelligent computing center through IAB. The intelligent computing center generates report information and transmits it to the target vehicle.

[0011] Optionally, in step 1, the radio wave constellation adopts a 16QAM constellation.

[0012] Optionally, in step 2, the radio wave data includes vehicle IP, vehicle speed and destination IP.

[0013] Optionally, in step 2, the K-Means algorithm is used to classify the radio wave constellation data, specifically:

[0014] Step 2.1: Receive radio wave data R x , the signal sequence is represented as {x1, ..., x i , ..., x n}, where x i represents the I / Q data sample point, and n represents the number of sequence samples;

[0015] Step 2.2: Initialize the cluster centroids {c1, ..., c j , ..., c m}, the cluster centroid indicates the location of the VLC receiver;

[0016] Step 2.3: Calculate x i and each c j The shortest distance between And x i Divide into clusters corresponding to the shortest distance;

[0017] Step 2.4: Recalculate the cluster centroid according to the divided clusters, and determine whether the deviation between the two cluster centroids is less than the set threshold E. If so, output the cluster classification result; if not, return to step 2.3.

[0018] Optionally, in step 2.3, Euclidean distance, Manhattan distance or Mahalanobis distance is used to calculate x i and each c j The shortest distance between.

[0019] Optionally, in step 3, the traffic flow data includes vehicle flow, vehicle speed and lane traffic density.

[0020] Optionally, in step 4, the Dijkstra algorithm or the A* algorithm is used to plan the optimal path for the target vehicle to reach the destination.

[0021] Optionally, in step 4, a Markov chain model is used to predict the congestion trend at the next traffic intersection, as follows:

[0022] Step 4.1: Set the state space of the Markov chain to S = {unblocked, lightly congested, moderately congested, heavily congested}, the state space size λ = 4, and each state is divided according to the vehicle speed and lane traffic density at the traffic intersection;

[0023] Step 4.2: Take the next traffic intersection as the traffic intersection to be tested, extract the traffic flow data of the traffic intersection to be tested, and count the traffic flow from state s in multiple time periods. p Transfer to q The number of times N pq ,s p ∈S,s q ∈S, calculate the transition probability And generate a transition probability matrix P according to the transition probability;

[0024] Step 4.3: Obtain the current state vector X(k) of the traffic intersection to be tested, calculate the state vector X(k+1) of the traffic intersection to be tested in the next time period according to the Markov formula X(k+1)=X(k)P, and take the state corresponding to the maximum element value in the state vector X(k+1) as the predicted congestion state.

[0025] Optionally, in step 5, the intelligent computing center uses a large language model to generate report information.

[0026] In a second aspect, the present invention provides an intelligent scheduling and prediction system combining a large star model with an intelligent computing center, including:

[0027] The VLC transmitter installed in each vehicle is used to transmit radio wave data;

[0028] The VLC receiving end installed at the traffic intersection is used to receive the radio wave data emitted by the VLC transmitting end within the signal transmission range;

[0029] A collection module, used for collecting radio wave data transmitted between the VLC transmitter and the VLC receiver to form a radio wave constellation;

[0030] The calculation module is used to classify the radio wave constellation data using the K-Means algorithm, and separate the radio wave data received by each VLC receiving end according to the time series; the Star Big Model is used to obtain the traffic flow data of each VLC receiving end corresponding to the traffic intersection distributed according to the time series based on the classified radio wave data; according to the traffic flow data of each traffic intersection, the optimal path algorithm is used to plan the optimal path for the target vehicle to reach the destination, and at the same time predict the congestion trend at the next traffic intersection;

[0031] The intelligent computing center generates report information based on the planning and prediction results and transmits it to the target vehicle.

[0032] The beneficial effects of the present invention are as follows: in order to overcome the problem of mutual interference caused by multi-user access in IoV, the present invention adopts the K-Means algorithm to classify the radio wave constellation data, and separates the radio wave signal according to the time series and the VLC receiving end, thereby ensuring the purity of the data used for subsequent operations; then, the present invention innovatively adopts the star big model to process the radio wave signal, fully utilizing the processing capability of the star big model for large-scale multi-source data, and effectively obtains the traffic flow data of each VLC receiving end corresponding to the traffic intersection distributed in the time series, and through these traffic flow data, the optimal path planning and traffic trend prediction can be quickly carried out without complicated feature extraction or parameter calculation; finally, the present invention generates report information through the intelligent computing center, and finally transmits the current report information to the target vehicle through the base station, thereby completing the full life cycle management of intelligent transportation travel and improving intelligence and automation. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of the intelligent scheduling and prediction method of the Star Big Model combined with the Intelligent Computing Center.

[0034] Figure 2 It is a schematic diagram of the transmission of VLC signals between vehicles and traffic intersections.

[0035] Figure 3 It is a flow chart for classifying data of radio wave constellations using K-Means algorithm. DETAILED DESCRIPTION

[0036] The present invention will now be described in further detail with reference to the accompanying drawings.

[0037] In one embodiment, the present invention proposes an intelligent scheduling and prediction method combining a large star model with an intelligent computing center. The overall process is as follows: Figure 1 As shown, the following steps are included:

[0038] Step 1: Collect the radio wave data transmitted between the VLC transmitter installed on each vehicle and the VLC receiver installed at the traffic intersection to form a radio wave constellation.

[0039] The VLC signal is transmitted between the vehicle and the traffic intersection. Figure 2 As shown in the figure, the VLC transmitter is usually installed at the position of the car light, and the VLC receiver is installed at the traffic light at the intersection. Since many vehicles cross at the intersection, the mutual interference of VLC signals can easily cause unstable signal transmission. DFT-S OFDM has high spectrum efficiency and can reduce PAPR, so it is very effective in anti-nonlinearity, so the geometric shaping 16QAM constellation can be used.

[0040] Radio wave data usually includes vehicle IP, vehicle speed and destination IP. By combining these dynamic parameters with some fixed parameters stored in the database (such as road-related parameters), data for subsequent optimal path planning and congestion probability prediction can be quickly calculated.

[0041] Step 2: Use the K-Means algorithm to classify the radio wave constellation data and separate the radio wave data received by each VLC receiver according to the time series.

[0042] When various cross-radio wave constellations occur, in order to separate the radio wave signals received by each VLC receiver well, the K-Means algorithm is used to classify the radio wave constellations, such as Figure 3 As shown, specifically:

[0043] Step 2.1: Receive radio wave data R x , the signal sequence is represented as {x1, ..., x i , ..., x n}, where x i represents the I / Q data sample point, and n represents the number of sequence samples;

[0044] Step 2.2: Initialize the cluster centroids {c1, ..., c j , ..., c m}, the cluster centroid indicates the location of the VLC receiver;

[0045] Step 2.3: Calculate x i and each c j The shortest distance between And x i Divide into clusters corresponding to the shortest distance; in this embodiment, the Euclidean distance, Manhattan distance or Mahalanobis distance can be used to calculate x i and each c j The shortest distance between.

[0046] Step 2.4: Recalculate the cluster centroid according to the divided clusters, and determine whether the deviation between the two cluster centroids is less than the set threshold E. If so, output the cluster classification result; if not, return to step 2.3.

[0047] Step 3: Use the Xingchen big model to obtain the traffic flow data of each VLC receiving end corresponding to the traffic intersection distributed in time series according to the classified radio wave data.

[0048] The Star Big Model is a product called the Star Big Model Software Factory released by China Telecom Artificial Intelligence Technology Co., Ltd. and China Telecom (601728) Artificial Intelligence Research Institute (TeleAI). It is the first breakthrough AI development tool in China. The Star Big Model has high computing efficiency and excellent semantic processing capabilities, and can be used for the calculation and processing of various data. In this embodiment, the Star Big Model is used for multi-user radio wave data processing, with the purpose of calculating traffic flow data, including vehicle flow, vehicle speed, and lane traffic density.

[0049] Step 4: Based on the traffic flow data of each traffic intersection, the optimal path algorithm is used to plan the optimal path for the target vehicle to reach the destination, and the congestion trend at the next traffic intersection is predicted.

[0050] Since the traffic flow data calculated in the previous article can quickly perform optimal path planning and traffic trend prediction without complicated feature extraction or parameter calculation, some existing optimal path algorithms, such as Dijkstra algorithm or A* algorithm, can be used here to plan the optimal path for the target vehicle to reach the destination.

[0051] The application of large models also needs to consider issues such as interpretability and robustness. Due to the huge scale of large models, their internal working mechanisms are often difficult to explain, which may lead to some unpredictable behaviors of the models in practical applications. Therefore, this embodiment does not use large models for prediction, but uses the Markov prediction model to predict congestion based on the data obtained in the result set of China Telecom Xingchen artificial intelligence large model. The details are as follows:

[0052] Step 4.1: Set the state space of the Markov chain to S = {unblocked, lightly congested, moderately congested, heavily congested}, the state space size λ = 4, and each state is divided according to the vehicle speed and lane traffic density at the traffic intersection;

[0053] Step 4.2: Take the next traffic intersection as the traffic intersection to be tested, extract the traffic flow data of the traffic intersection to be tested, and count the traffic flow from state s in multiple time periods. p Transfer to q The number of times N pq ,s p ∈S,s q ∈S, calculate the transition probability And generate a transition probability matrix P according to the transition probability;

[0054] Step 4.3: Obtain the current state vector X(k) of the traffic intersection to be tested, calculate the state vector X(k+1) of the traffic intersection to be tested in the next time period according to the Markov formula X(k+1)=X(k)P, and take the state corresponding to the maximum element value in the state vector X(k+1) as the predicted congestion state.

[0055] Step 5: After the planning results and prediction results are integrated, they are uploaded to the relay satellite layer through the Beidou chip deployed in the Internet of Vehicles, and then fed back to the ground intelligent computing center through Integrated Access and Backhaul (IAB). The intelligent computing center generates report information and transmits it to the target vehicle.

[0056] After the BeiDou-3 network was established, the capacity of BeiDou regional short messages was expanded to more than 8 times the original capacity, from a maximum of 120 Chinese characters to a maximum of 1,000 Chinese characters, which greatly broadened the development path of BeiDou short messages at the application level. In this embodiment, BeiDou chips can also be deployed in mobile terminals, etc. After being uploaded to the relay satellite layer of the constellation through the Internet of Vehicles or BeiDou chips deployed in mobile terminals, Rdss short messages are sent to BeiDou navigation to assist navigation, and at the same time, they are fed back to the ground intelligent computing center through IAB, and the current report information is finally delivered to the target vehicle through the base station to complete the full life cycle management of intelligent transportation travel.

[0057] The intelligent computing center can use the language large model to generate report information, describing the information obtained in the previous step and the possible safety issues and traffic rules as T_question. If the VLC receiver deployed at the current intersection has different policies for the safety issues and traffic rules of different ground vehicles, then all relevant policies can be divided into text blocks according to Chinese delimiters (such as "\n" line breaks, periods, etc.), input lengths, and overlaps. Then, the text is vectorized through the embedding model bge-large-v1.5-zh, and then uploaded to the vector database. T_question is embedded in the model vector The top-k traffic rules that are most relevant to the safety issue mentioned in T_question are retrieved using a similarity search method. The retrieved top-k policies are input into the language model M_language together with T_question as prompt words. At the same time, a template for generating an uploaded report is added to the prompt word P_language, that is, what content needs to be included in the uploaded report, such as existing problems, rectification measures, regulatory units, etc. Different templates can be customized according to the requirements of different government departments, and finally a report A=M_language(P_language T_question) corresponding to the intersection is generated. The report output is a string type, which includes the content required by the policy keywords in the vector database.

[0058] In another embodiment, the present invention proposes an intelligent scheduling and prediction system combining a large star model with an intelligent computing center, which corresponds to an intelligent scheduling and prediction method combining a large star model with an intelligent computing center in the aforementioned embodiment, including:

[0059] The VLC transmitter installed in each vehicle is used to transmit radio wave data;

[0060] The VLC receiving end installed at the traffic intersection is used to receive the radio wave data emitted by the VLC transmitting end within the signal transmission range;

[0061] A collection module, used for collecting radio wave data transmitted between the VLC transmitter and the VLC receiver to form a radio wave constellation;

[0062] The calculation module is used to classify the radio wave constellation data using the K-Means algorithm, and separate the radio wave data received by each VLC receiving end according to the time series; the Star Big Model is used to obtain the traffic flow data of each VLC receiving end corresponding to the traffic intersection distributed according to the time series based on the classified radio wave data; according to the traffic flow data of each traffic intersection, the optimal path algorithm is used to plan the optimal path for the target vehicle to reach the destination, and at the same time predict the congestion trend at the next traffic intersection;

[0063] The intelligent computing center generates report information based on the planning and prediction results and transmits it to the target vehicle.

[0064] The specific functions and implementation processes of each component in the system are the same as the steps of the intelligent scheduling and prediction method in the aforementioned embodiment, so they will not be repeated here.

[0065] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. An intelligent scheduling and prediction method combining a large star model with an intelligent computing center, characterized in that: include: Step 1: Collect radio wave data transmitted between the VLC transmitter installed on each vehicle and the VLC receiver installed at the traffic intersection to form a radio wave constellation; Step 2: Use the K-Means algorithm to classify the radio wave constellation data and separate the radio wave data received by each VLC receiver according to the time series; Step 3: Use the Xingchen big model to obtain the traffic flow data of each VLC receiving end corresponding to the traffic intersection distributed in time series according to the classified radio wave data; Step 4: Based on the traffic flow data of each traffic intersection, the optimal path algorithm is used to plan the optimal path for the target vehicle to reach the destination, and the congestion trend at the next traffic intersection is predicted; Step 5: After the planning results and prediction results are integrated, they are uploaded to the relay satellite layer through the Beidou chip deployed in the Internet of Vehicles, and then fed back to the ground intelligent computing center through IAB. The intelligent computing center generates report information and transmits it to the target vehicle.

2. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 1 is characterized by: In step 1, the radio wave constellation adopts a 16QAM constellation.

3. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 1 is characterized by: In step 2, the radio wave data includes vehicle IP, vehicle speed and destination IP.

4. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 1 is characterized by: In step 2, the K-Means algorithm is used to classify the radio wave constellation data, specifically: Step 2.1: Receive radio wave data R x , the signal sequence is represented as {x1, ..., x i , ..., x n }, where x i represents the I / Q data sample point, and n represents the number of sequence samples; Step 2.2: Initialize the cluster centroids {c1, ..., c j , ..., c m }, the cluster centroid indicates the location of the VLC receiver; Step 2.3: Calculate x i and each c j The shortest distance between And x i Divide into clusters corresponding to the shortest distance; Step 2.4: Recalculate the cluster centroid according to the divided clusters, and determine whether the deviation between the two cluster centroids is less than the set threshold E. If so, output the cluster classification result; if not, return to step 2.

3.

5. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 4 is characterized by: In step 2.3, Euclidean distance, Manhattan distance or Mahalanobis distance is used to calculate x i and each c j The shortest distance between.

6. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 1 is characterized by: In step 3, the traffic flow data includes vehicle flow, vehicle speed and lane traffic density.

7. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 1 is characterized by: In step 4, the Dijkstra algorithm or the A* algorithm is used to plan the optimal path for the target vehicle to reach the destination.

8. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 6 is characterized by: In step 4, the Markov chain model is used to predict the congestion trend at the next traffic intersection, as follows: Step 4.1: Set the state space of the Markov chain to S = {unblocked, lightly congested, moderately congested, heavily congested}, the state space size λ = 4, and each state is divided according to the vehicle speed and lane traffic density at the traffic intersection; Step 4.2: Take the next traffic intersection as the traffic intersection to be tested, extract the traffic flow data of the traffic intersection to be tested, and count the traffic flow from state s in multiple time periods. p Transfer to q The number of times N pq ,s p ∈S,s q ∈S, calculate the transition probability And generate a transition probability matrix P according to the transition probability; Step 4.3: Obtain the current state vector X(k) of the traffic intersection to be tested, calculate the state vector X(k+1) of the traffic intersection to be tested in the next time period according to the Markov formula X(k+1)=X(k)P, and take the state corresponding to the maximum element value in the state vector X(k+1) as the predicted congestion state.

9. The intelligent scheduling and prediction method of the star big model combined with the intelligent computing center as claimed in claim 1 is characterized by: In step 5, the intelligent computing center uses a large language model to generate report information.

10. An intelligent scheduling and prediction system combining a large star model with an intelligent computing center, characterized in that: include: The VLC transmitter installed in each vehicle is used to transmit radio wave data; The VLC receiving end installed at the traffic intersection is used to receive the radio wave data emitted by the VLC transmitting end within the signal transmission range; A collection module, used for collecting radio wave data transmitted between the VLC transmitter and the VLC receiver to form a radio wave constellation; The calculation module is used to classify the radio wave constellation data using the K-Means algorithm, and separate the radio wave data received by each VLC receiving end according to the time series; the Star Big Model is used to obtain the traffic flow data of each VLC receiving end corresponding to the traffic intersection distributed according to the time series based on the classified radio wave data; according to the traffic flow data of each traffic intersection, the optimal path algorithm is used to plan the optimal path for the target vehicle to reach the destination, and at the same time predict the congestion trend at the next traffic intersection; The intelligent computing center generates report information based on the planning and prediction results and transmits it to the target vehicle.

Images