A low-power vehicle-building collaborative communication method and system

Abstract

The present invention relates to the field of personnel communication technology, and in particular to a low-power vehicle-building collaborative communication method and system. A low-power vehicle-building collaborative communication system comprises: a vehicle communication status data set construction module and a transmission power adjustment module. The present invention establishes a communication link with vehicles within a certain range from the building node to achieve vehicle-building collaborative communication, and obtains the vehicle communication status data set fed back by the vehicle, and then adjusts the reflected power based on the vehicle communication status data set fed back by all vehicles, thereby reducing the overall transmission power of the building node while meeting the communication quality of the vehicle, thereby reducing the energy consumption of the vehicle-building collaborative communication; and in the process of predicting the vehicle communication status, the vehicle communication status prediction model will be adjusted based on the actual vehicle communication status information of the vehicles within a certain range, so that the short-term vehicle communication status prediction model obtained by training can fit the actual environmental conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of personnel communication, and in particular to a low-power vehicle-building collaborative communication method and system. Background Art

[0002] Vehicle-to-building collaborative communication is a crucial component of intelligent transportation systems (ITS) and smart cities. Through collaborative communication between vehicles and building nodes (such as roadside units and parking lot base stations), it supports vehicle navigation, parking guidance, and traffic management. However, existing vehicle-to-building collaborative communication technologies generally suffer from high energy consumption, limiting their widespread application in real-world scenarios.

[0003] In traditional vehicle-to-building communications, due to the complex communication environment (such as rapid switching between line-of-sight and non-line-of-sight, multipath effects, and obstructions), the system typically requires high transmit power to ensure communication stability and reliability. Furthermore, in densely populated environments with multiple vehicles, competition for communication links intensifies, further increasing the demand for transmit power, leading to a significant increase in overall energy consumption. Furthermore, existing power control strategies, which mostly rely on fixed power or simple feedback adjustments, fail to fully account for dynamic channel conditions, changes in vehicle position, and differences in communication needs, often resulting in excessively high transmit power. Summary of the Invention

[0004] The present invention establishes a communication link with vehicles within a certain range of the building node to achieve vehicle-building collaborative communication, obtains the vehicle communication status data set fed back by the vehicles, and then adjusts the reflected power based on the vehicle communication status data set fed back by all vehicles, thereby reducing the overall transmission power of the building node while meeting the communication quality of the vehicles, thereby reducing the energy consumption of vehicle-building collaborative communication; and in the process of predicting the vehicle communication status, the vehicle communication status prediction model will be adjusted based on the actual vehicle communication status information of the vehicles within a certain range, so that the short-term vehicle communication status prediction model obtained by training can fit the actual environmental conditions.

[0005] The present invention provides a low-power vehicle-building collaborative communication method, comprising:

[0006] Obtain the corresponding channel state data set for each vehicle, and obtain the corresponding location information of the vehicle, and then combine the corresponding channel state data set and location information of the vehicle to form a vehicle communication state data set;

[0007] At the transmit power adjustment time point, transmit power is allocated to all vehicles using the predicted transmit power allocation strategy corresponding to the transmit power adjustment time point;

[0008] The construction of the predictive transmit power allocation strategy includes the following steps:

[0009] At the building node, the following is performed for each vehicle with an established communication link: at the current transmission power adjustment time point, all vehicle communication status data sets corresponding to the vehicle are arranged in chronological order from front to back, and then N vehicle communication status data sets are selected from all the arranged vehicle communication status data sets from back to front to form a vehicle communication status timing analysis data set, and the N vehicle communication status data sets in the vehicle communication status timing analysis data set are in chronological order from front to back, and a vehicle communication status prediction model is obtained. The vehicle communication status prediction model is established based on the Transformer model, and the vehicle communication status prediction model is trained based on all vehicles within a preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model. The vehicle communication status timing analysis data set is then sent to the short-term vehicle communication status prediction model for processing, and the predicted vehicle communication status data set corresponding to the next transmission power adjustment time point is output; the predicted vehicle communication status data sets corresponding to all vehicles are arranged from large to small according to the distance from the building node to form a predicted vehicle power allocation state set, and the predicted vehicle power allocation state set is then sent to the vehicle communication power allocation model for processing, and the predicted transmission power allocation strategy is output; and at the transmission power adjustment time point, the vehicle communication power allocation model is trained in real time; and the vehicle communication status prediction model is also updated through real-time training at the model update time point, and several transmission power adjustment time points are included between two adjacent model update time points.

[0010] Preferably, the vehicle communication state prediction model is trained based on all vehicles within a preset distance range of the vehicle to obtain a short-term vehicle communication state prediction model, which specifically includes the following steps:

[0011] Vehicles within a preset distance range from the vehicle are recorded as target vehicles, and all target vehicles are grouped into a target vehicle set. The target vehicle set is traversed, and the following operations are performed for each target vehicle: obtain the predicted vehicle communication status dataset and vehicle communication status dataset corresponding to the target vehicle at the current transmission power adjustment time point, and obtain N vehicle communication status datasets of the target vehicle before the current transmission power adjustment time point, perform a differential operation on the predicted vehicle communication status dataset and the vehicle communication status dataset corresponding to the target vehicle at the current transmission power adjustment time point, and construct a difference vector, and combine the N vehicle communication status datasets of the target vehicle before the current transmission power adjustment time point, the vehicle communication status dataset corresponding to the current transmission power adjustment time point, and the difference vector to form a short-term training sample; sum and average the difference vectors corresponding to all target vehicles to construct a comprehensive difference vector, and combine the short-term training samples corresponding to all target vehicles into a short-term training set The vehicle communication status prediction model is trained using a short-time training set. During the training, N vehicle communication status data sets of the target vehicle before the current transmission power adjustment time point are used as the input of the vehicle communication status prediction model, and the vehicle communication status data set corresponding to the current transmission power adjustment time point is used as the target output of the vehicle communication status prediction model. The difference vector is used as the query vector when the vehicle communication status prediction model executes the self-attention mechanism. The short-time training loss value is calculated to determine whether the short-time training loss value is within the vehicle communication status prediction confidence range. If the short-time training loss value is within the vehicle communication status prediction confidence range, the short-time vehicle communication status prediction model is output, and the difference vector used when executing the self-attention mechanism in the short-time vehicle communication status prediction model is the comprehensive difference vector. If the short-time training loss value is not within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is continued to be trained using the short-time training set.

[0012] Preferably, real-time training is performed on the vehicle communication power allocation model, specifically including the following steps:

[0013] The predicted vehicle power allocation state set corresponding to the previous transmission power adjustment time point, the predicted transmission power allocation strategy output at the previous transmission power adjustment time point, and the vehicle power allocation state set corresponding to the current transmission power adjustment time point are combined into a strategy evaluation set, and the vehicle power allocation state set is composed of vehicle communication state data sets corresponding to all vehicles arranged from large to small according to the distance from the building node, and the strategy evaluation value is calculated for the predicted transmission power allocation strategy in the strategy evaluation set and the vehicle power allocation state set corresponding to the current transmission power adjustment time point; the predicted vehicle power allocation state set corresponding to the previous transmission power adjustment time point and the predicted transmission power allocation strategy output at the previous transmission power adjustment time point are combined into vehicle communication power allocation real-time training samples, and the vehicle communication power allocation real-time training samples are labeled by the strategy evaluation value; all labeled vehicle communication power allocation real-time training samples are combined into a vehicle communication power allocation real-time training set;

[0014] The vehicle communication power allocation strategy evaluation model is trained in real time using the vehicle communication power allocation real-time training set. During the real-time training, the vehicle communication power allocation real-time training sample is used as input, and the strategy evaluation value is used as the target output. The vehicle communication power allocation real-time training loss value is calculated to determine whether the vehicle communication power allocation real-time training loss value is within the vehicle communication power allocation evaluation confidence range. If the vehicle communication power allocation real-time training loss value is within the vehicle communication power allocation evaluation confidence range, the trained vehicle communication power allocation strategy evaluation model is output; otherwise, the vehicle communication power allocation strategy evaluation model is continued to be trained in real time using the vehicle communication power allocation real-time training set.

[0015] The predicted vehicle power allocation state set corresponding to the current transmission power adjustment time point is sent to the target vehicle communication power allocation model for processing, and the target predicted transmission power allocation strategy is output. The predicted vehicle power allocation state set corresponding to the current transmission power adjustment time point and the target predicted transmission power allocation strategy are combined and sent to the trained vehicle communication power allocation strategy evaluation model, and the target strategy evaluation value is output. Based on the target strategy evaluation value, the parameters of the vehicle communication power allocation model are adjusted by the gradient ascent method to realize real-time training of the vehicle communication power allocation model; and it also includes updating the target vehicle communication power allocation model at the model update time point.

[0016] Preferably, the target vehicle communication power distribution model is updated at the target model update time point, specifically including the following steps: directly replacing the target vehicle communication power distribution model with the vehicle communication power distribution model at the target model update time point, and in the initial state, the vehicle communication power distribution model and the target vehicle communication power distribution model are consistent.

[0017] Preferably, at the model update time point, the vehicle communication status prediction model is updated through real-time training, which specifically includes the following steps:

[0018] At a model update time point, a vehicle communication status prediction real-time training sample between the current model update time point and the previous model update time point is obtained. The vehicle communication status prediction real-time training sample is N+1 vehicle communication status data sets corresponding to any vehicle arranged in chronological order. All vehicle communication status prediction real-time training samples are combined into a vehicle communication status prediction real-time training set. The vehicle communication status prediction model is trained in real time using the vehicle communication status prediction real-time training set. During the real-time training, the first N vehicle communication status data sets in the vehicle communication status prediction real-time training set are used as inputs of the vehicle communication status prediction model, and the last vehicle communication status data set in the vehicle communication status prediction real-time training set is used as the target output of the vehicle communication status prediction model. A vehicle communication status prediction real-time training loss value is calculated, and it is determined whether the vehicle communication status prediction real-time training loss value is within a vehicle communication status prediction confidence range. If the vehicle communication status prediction real-time training loss value is within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is updated. If the vehicle communication status prediction real-time training loss value is not within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is continued to be trained using the vehicle communication status prediction real-time training set.

[0019] Preferably, it also includes pre-training of the vehicle communication status prediction model, and the specific steps are as follows:

[0020] A number of vehicle communication status prediction pre-training samples are obtained. The vehicle communication status prediction pre-training samples include N+1 vehicle communication status data sets corresponding to any vehicle and arranged in chronological order. All vehicle communication status prediction pre-training samples are combined into a vehicle communication status prediction pre-training set, and a vehicle communication status prediction model is pre-trained using the vehicle communication status prediction pre-training set. The pre-training method is consistent with the real-time training method of the vehicle communication status prediction model.

[0021] Preferably, the training of the vehicle communication power allocation model and the vehicle communication power allocation strategy evaluation model specifically includes the following steps:

[0022] Obtain several vehicle communication power allocation training samples, the vehicle communication power allocation training samples include a vehicle power allocation state set, the vehicle power allocation state set includes a vehicle communication state data set of all vehicles at the same time point, the vehicle communication power allocation training samples are labeled according to the target transmission power allocation strategy, all labeled vehicle communication power allocation training samples are combined into a vehicle communication power allocation training set, a vehicle communication power allocation model is trained using the vehicle communication power allocation training set, a vehicle communication power allocation loss value is calculated, and it is determined whether the vehicle communication power allocation loss value is within the vehicle communication power allocation confidence range. If the vehicle communication power allocation loss value is within the vehicle communication power allocation confidence range, the trained vehicle communication power allocation model is output; otherwise, the vehicle communication power allocation model is continued to be trained using the vehicle communication power allocation training set;

[0023] Acquire several communication power allocation strategy training samples, which include a vehicle power allocation state set and the transmission power allocation strategy it adopts, and label the communication power allocation strategy training samples with a strategy evaluation value, and form a communication power allocation strategy training set with all labeled communication power allocation strategy training samples. Train a vehicle communication power allocation strategy evaluation model with the communication power allocation strategy training set, calculate a vehicle communication power allocation strategy evaluation loss value, and determine whether the vehicle communication power allocation strategy evaluation loss value is within a vehicle communication power allocation evaluation confidence range. If the vehicle communication power allocation strategy evaluation loss value is within the vehicle communication power allocation evaluation confidence range, output the trained vehicle communication power allocation strategy evaluation model; otherwise, continue to train the vehicle communication power allocation strategy evaluation model in real time with the communication power allocation strategy training set.

[0024] The present invention also provides a low-power vehicle-building collaborative communication system, comprising:

[0025] The vehicle communication status data set construction module is used to obtain the corresponding channel status data set for each vehicle and the corresponding location information of the vehicle, and then combine the corresponding channel status data set and the location information of the vehicle to form a vehicle communication status data set;

[0026] A transmission power adjustment module is used to allocate transmission power to all vehicles at a transmission power adjustment time point according to a predicted transmission power allocation strategy corresponding to the transmission power adjustment time point;

[0027] The construction of the predictive transmit power allocation strategy includes the following steps:

[0028] At the building node, the following is performed for each vehicle with an established communication link: at the current transmission power adjustment time point, all vehicle communication status data sets corresponding to the vehicle are arranged in chronological order from front to back, and then N vehicle communication status data sets are selected from all the arranged vehicle communication status data sets from back to front to form a vehicle communication status timing analysis data set, and the N vehicle communication status data sets in the vehicle communication status timing analysis data set are in chronological order from front to back, and a vehicle communication status prediction model is obtained. The vehicle communication status prediction model is established based on the Transformer model, and the vehicle communication status prediction model is trained based on all vehicles within a preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model. The vehicle communication status timing analysis data set is then sent to the short-term vehicle communication status prediction model for processing, and the predicted vehicle communication status data set corresponding to the next transmission power adjustment time point is output; the predicted vehicle communication status data sets corresponding to all vehicles are arranged from large to small according to the distance from the building node to form a predicted vehicle power allocation state set, and the predicted vehicle power allocation state set is then sent to the vehicle communication power allocation model for processing, and the predicted transmission power allocation strategy is output; and at the transmission power adjustment time point, the vehicle communication power allocation model is trained in real time; and the vehicle communication status prediction model is also updated through real-time training at the model update time point, and several transmission power adjustment time points are included between two adjacent model update time points.

[0029] The present invention has the following advantages:

[0030] The present invention establishes a communication link with vehicles within a certain range of the building node to achieve vehicle-building collaborative communication, obtains the vehicle communication status data set fed back by the vehicles, and then adjusts the reflected power based on the vehicle communication status data set fed back by all vehicles, thereby reducing the overall transmission power of the building node while meeting the communication quality of the vehicles, thereby reducing the energy consumption of vehicle-building collaborative communication; and in the process of predicting the vehicle communication status, the vehicle communication status prediction model will be adjusted based on the actual vehicle communication status information of the vehicles within a certain range, so that the short-term vehicle communication status prediction model obtained by training can fit the actual environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a flow chart of the low-power vehicle-building collaborative communication method adopted in an embodiment of the present invention.

[0032] Figure 2 This is a structural diagram of the low-power vehicle-building collaborative communication system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0034] Example 1, a low-power vehicle-building collaborative communication method, such as Figure 1 Shown, including:

[0035] For each vehicle, the corresponding channel state data set is obtained. The channel state data set includes data describing the channel state such as the signal-to-interference-noise ratio and bit error rate obtained by the vehicle based on the pilot signal detection, and the corresponding position information of the vehicle is obtained. Then, the channel state data set corresponding to the vehicle and the position information are combined to form a vehicle communication state data set. It should be noted that in the process of guiding the vehicle to park through vehicle-building collaborative communication, the building node will periodically send a pilot signal to a certain range. The pilot signal (Pilot signal) A pilot signal is a predefined, known signal sequence used for link establishment, channel estimation, and synchronization in wireless communication systems. When a vehicle receives a pilot signal, it establishes a communication link based on the owner's parking intention and conducts real-time communication based on the communication link to provide parking guidance. Parking guidance specifically includes the following: dynamically planning parking paths, updating the availability of parking spaces, and providing accurate parking navigation (such as remaining distance and turn guidance) for vehicles. It also measures the received signal strength, signal-to-noise ratio, and channel capacity of the pilot signal, and other data describing the channel state to form a channel state data set. The channel state data set and location information are combined to form a vehicle communication state data set and fed back to the building node. The building node will perform subsequent transmission power adjustments based on the vehicle's vehicle communication state data set. In addition, between the building node and the vehicle that have established a communication link, the vehicle will also regularly feed back the vehicle communication state data set to the building node during real-time communication. It should be noted that the vehicle's location information will affect the power loss during the communication process.

[0036] At the transmit power adjustment time point, transmit power is allocated to all vehicles using the predicted transmit power allocation strategy corresponding to the transmit power adjustment time point. It should be noted that building nodes can set different transmit powers for different vehicles through a multi-antenna system, and the predicted transmit power allocation strategy can reduce the overall transmit power of building nodes while ensuring satisfactory vehicle communication quality, thereby reducing the energy consumption of vehicle-building collaborative communication.

[0037] The construction of the predictive transmit power allocation strategy includes the following steps:

[0038] At the building node, the following is performed for each vehicle with an established communication link: at the current transmission power adjustment time point, all vehicle communication status data sets corresponding to the vehicle are arranged in chronological order from front to back, and then N vehicle communication status data sets are selected from all the arranged vehicle communication status data sets from back to front to form a vehicle communication status timing analysis data set, and the N vehicle communication status data sets in the vehicle communication status timing analysis data set are in chronological order from front to back. It should be noted that the building node will continuously receive the vehicle communication status data sets uploaded by the vehicle on a regular basis, but when the total amount of the vehicle communication status data sets corresponding to the vehicle exceeds the upper limit, the vehicle communication status data sets with an older time will be cleared, and when the vehicle is in contact with the building node, When the communication link between the points is disconnected, all vehicle communication status data sets corresponding to the vehicle will be cleared to obtain the vehicle communication status prediction model. The vehicle communication status prediction model is established based on the Transformer model and is trained based on all vehicles within the preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model. The preset range here is determined by the developer, generally 20m. It should be noted that at the building node, the vehicle communication status prediction model will be regularly updated through real-time training, and this vehicle communication status prediction model is for all vehicles, but because the actual environment faced by the vehicle will change, such as vehicle obstruction or obstacle obstruction, etc., each vehicle will be based on The vehicle communication status prediction model is adjusted based on the actual vehicle communication status information of all vehicles within the preset range, so that the trained short-term vehicle communication status prediction model can fit the actual environmental conditions, thereby improving the accuracy of subsequent predictions of the vehicle communication status data set, and then the vehicle communication status time series analysis data set is sent to the short-term vehicle communication status prediction model for processing, and the predicted vehicle communication status data set corresponding to the next transmission power adjustment time point is output; and the predicted vehicle communication status data sets corresponding to all vehicles are arranged from large to small according to the distance from the building node to form a predicted vehicle power allocation state set, which can reflect the channel status and position of all vehicles at the next transmission power adjustment time point. The predicted vehicle power allocation state set is then sent to the vehicle communication power allocation model for processing, and a predicted transmission power allocation strategy is output. The vehicle communication power allocation model is established based on the BP neural network. The predicted transmission power allocation strategy includes the transmission power corresponding to all vehicles that have established communication links. At the transmission power adjustment time point, the vehicle communication power allocation model is trained in real time, so that the predicted transmission power allocation strategy is more in line with the constraint of reducing the overall transmission power of the building node while meeting the communication quality of the vehicle. The vehicle communication state prediction model is also updated through real-time training at the model update time point, and several transmission power adjustment time points are included between two adjacent model update time points.

[0039] This application establishes a communication link with vehicles within a certain range of the building node to achieve vehicle-building collaborative communication, obtains the vehicle communication status data set fed back by the vehicles, and then adjusts the reflected power based on the vehicle communication status data set fed back by all vehicles, thereby reducing the overall transmission power of the building node while meeting the communication quality of the vehicle, thereby reducing the energy consumption of vehicle-building collaborative communication; and in the process of predicting the vehicle communication status, the vehicle communication status prediction model will be adjusted based on the actual vehicle communication status information of vehicles within a certain range, so that the trained short-term vehicle communication status prediction model can fit the actual environmental conditions.

[0040] The vehicle communication status prediction model is trained based on all vehicles within a preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model, which specifically includes the following steps:

[0041] The vehicles within the preset distance range from the vehicle are recorded as target vehicles, and all target vehicles are grouped into a target vehicle set. The target vehicle set is traversed, and the following operations are performed for each target vehicle: obtain the predicted vehicle communication status dataset and vehicle communication status dataset corresponding to the target vehicle at the current transmission power adjustment time point, and obtain N copies of vehicle communication status datasets of the target vehicle before the current transmission power adjustment time point, perform a differential operation on the predicted vehicle communication status dataset and vehicle communication status dataset corresponding to the target vehicle at the current transmission power adjustment time point, and construct a difference vector. It should be noted that the difference vector here can reflect the predicted vehicle communication status dataset under the influence of the environment. The difference between the communication status dataset and the actual vehicle communication status dataset can be used for subsequent training of the vehicle communication status prediction model. The N vehicle communication status datasets of the target vehicle before the current transmission power adjustment time point, the vehicle communication status dataset corresponding to the current transmission power adjustment time point, and the difference vector are combined into a short-term training sample; the difference vectors corresponding to all target vehicles are added and averaged to construct a comprehensive difference vector, and the short-term training samples corresponding to all target vehicles are combined into a short-term training set, and the vehicle communication status prediction model is trained through the short-term training set. During the training, the N vehicle communication status datasets of the target vehicle before the current transmission power adjustment time point are used as the input of the vehicle communication status prediction model, the vehicle communication status dataset corresponding to the current transmission power adjustment time point is used as the target output of the vehicle communication status prediction model, and the difference vector is used as the query vector when the vehicle communication status prediction model executes the self-attention mechanism. It should be noted that the vehicle communication status prediction model is established based on the Transformer model, which generally includes six encoders and six decoders, and both the encoder and the decoder include a self-attention mechanism. Generally, the value vector, key vector and query vector are constructed with the input vector input to the encoder or decoder, and in this application, the vehicle communication status prediction model is performed. During training, the difference vector in the short-time training sample is used as the query vector. The difference information caused by the environment can be integrated to calculate the short-time training loss value and determine whether the short-time training loss value is within the vehicle communication status prediction confidence range. The vehicle communication status prediction confidence range is set manually. If the short-time training loss value is within the vehicle communication status prediction confidence range, the short-time vehicle communication status prediction model is output, and the difference vector used when executing the self-attention mechanism in the short-time vehicle communication status prediction model is the comprehensive difference vector; if the short-time training loss value is not within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is continued to be trained through the short-time training set.

[0042] Real-time training of the vehicle communication power allocation model includes the following steps:

[0043] The predicted vehicle power allocation state set corresponding to the last transmission power adjustment time point, the predicted transmission power allocation strategy output at the last transmission power adjustment time point, and the vehicle power allocation state set corresponding to the current transmission power adjustment time point are composed of a strategy evaluation set, and the vehicle power allocation state set is composed of the vehicle communication state data sets corresponding to all vehicles arranged from large to small according to the distance from the building node, and the strategy evaluation value is calculated for the predicted transmission power allocation strategy in the strategy evaluation set and the vehicle power allocation state set corresponding to the current transmission power adjustment time point. The strategy evaluation value is composed of the satisfaction of the transmission power and the communication quality. The specific contents are as follows: traverse all vehicles, and for each vehicle, perform the following operations: obtain the transmission power P allocated to the vehicle by the predicted transmission power allocation strategy j , calculate the minimum required transmission power p corresponding to the vehicle by the following formula min , p min =(P noise +P interference )SINR target L / G, where P noise is the background noise power, through P noise =kTB, where k is the Boltzmann constant, T (in Kelvin) is the system temperature, B is the communication bandwidth, and P interference is the interference power at the receiving end, which is monitored by the vehicle’s spectrum analyzer, and P noise and P interference The building node queries the vehicle and obtains the SINR. target is the target signal-to-interference-noise ratio, which is determined by the communication protocol between the vehicle and the building node. It is generally 6dB under QPSK modulation. L is the path loss, specifically L=10 (H / 10) , H is the logarithmic unit of path loss, usually calculated by the free space propagation model, H = 20log 10 (d)+20log 10 (P j )-147.55, where d is the distance between the vehicle and the building node, G is the channel gain, which is determined by the antenna specifications corresponding to the building node. The communication quality Q of the vehicle communication under the vehicle power allocation state set corresponding to the current transmission power adjustment time point is calculated. Q=[γ1ln(1+SINR)-γ2BER], where SINR is the signal to interference and noise ratio corresponding to the vehicle power allocation state set corresponding to the current transmission power adjustment time point, BER is the bit error rate corresponding to the vehicle power allocation state set corresponding to the current transmission power adjustment time point, γ1 and γ2 are the weights corresponding to the signal to interference and noise ratio and the bit error rate, respectively, which are generally determined by the developer. Then the local strategy evaluation value E corresponding to the vehicle is E=Q / (P j -p min), which can reflect the premise of meeting the communication quality of the vehicle, reduce the overall transmission power of the building node, and then perform weighted averaging of the local strategy evaluation values ​​E corresponding to all vehicles to obtain a strategy evaluation value; the predicted vehicle power allocation state set corresponding to the last transmission power adjustment time point and the predicted transmission power allocation strategy output at the last transmission power adjustment time point constitute the vehicle communication power allocation real-time training sample, and the vehicle communication power allocation real-time training sample is labeled by the strategy evaluation value; all labeled vehicle communication power allocation real-time training samples are composed into a vehicle communication power allocation real-time training set;

[0044] The vehicle communication power allocation strategy evaluation model is trained in real time using the vehicle communication power allocation real-time training set. During the real-time training, the vehicle communication power allocation real-time training sample is used as input, and the strategy evaluation value is used as the target output. The vehicle communication power allocation real-time training loss value is calculated to determine whether the vehicle communication power allocation real-time training loss value is within the vehicle communication power allocation evaluation confidence range. The vehicle communication power allocation evaluation confidence range is set manually. If the vehicle communication power allocation real-time training loss value is within the vehicle communication power allocation evaluation confidence range, the trained vehicle communication power allocation strategy evaluation model is output; otherwise, the vehicle communication power allocation strategy evaluation model continues to be trained in real time using the vehicle communication power allocation real-time training set.

[0045] The predicted vehicle power allocation state set corresponding to the current transmission power adjustment time point is sent to the target vehicle communication power allocation model for processing, and the target predicted transmission power allocation strategy is output. The predicted vehicle power allocation state set corresponding to the current transmission power adjustment time point and the target predicted transmission power allocation strategy are combined and sent to the trained vehicle communication power allocation strategy evaluation model, and the target strategy evaluation value is output. Based on the target strategy evaluation value, the parameters of the vehicle communication power allocation model are adjusted by the gradient ascent method to realize real-time training of the vehicle communication power allocation model; and it also includes updating the target vehicle communication power allocation model at the model update time point.

[0046] The target vehicle communication power allocation model is updated at the target model update time point, specifically including the following steps: directly replacing the target vehicle communication power allocation model with the vehicle communication power allocation model at the target model update time point, and in the initial state, the vehicle communication power allocation model and the target vehicle communication power allocation model are consistent.

[0047] At the model update time point, the vehicle communication status prediction model is updated through real-time training, which specifically includes the following steps:

[0048] At a model update time point, a vehicle communication status prediction real-time training sample between the current model update time point and the previous model update time point is obtained. The vehicle communication status prediction real-time training sample is N+1 vehicle communication status data sets corresponding to any vehicle arranged in chronological order. All vehicle communication status prediction real-time training samples are combined into a vehicle communication status prediction real-time training set. The vehicle communication status prediction model is trained in real time using the vehicle communication status prediction real-time training set. During the real-time training, the first N vehicle communication status data sets in the vehicle communication status prediction real-time training set are used as inputs of the vehicle communication status prediction model, and the last vehicle communication status data set in the vehicle communication status prediction real-time training set is used as the target output of the vehicle communication status prediction model. A vehicle communication status prediction real-time training loss value is calculated, and it is determined whether the vehicle communication status prediction real-time training loss value is within a vehicle communication status prediction confidence range. If the vehicle communication status prediction real-time training loss value is within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is updated. If the vehicle communication status prediction real-time training loss value is not within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is continued to be trained using the vehicle communication status prediction real-time training set.

[0049] It should be noted that before the vehicle communication status prediction model is actually used, it is necessary to pre-train the vehicle communication status prediction model with existing historical data so that it can be put into practical use. Therefore, pre-training of the vehicle communication status prediction model is also included. The specific steps are as follows:

[0050] Obtain several vehicle communication status prediction pre-training samples. The vehicle communication status prediction pre-training samples include N+1 vehicle communication status data sets corresponding to any vehicle arranged in chronological order. It should be noted that the vehicle communication status prediction pre-training samples here are obtained by developers based on actual experiments in vehicle-building collaborative communication. All vehicle communication status prediction pre-training samples are combined into a vehicle communication status prediction pre-training set, and the vehicle communication status prediction model is pre-trained using the vehicle communication status prediction pre-training set. The pre-training method is consistent with the real-time training method of the vehicle communication status prediction model.

[0051] The training of the vehicle communication power allocation model and the vehicle communication power allocation strategy evaluation model specifically includes the following steps:

[0052] Obtain several vehicle communication power allocation training samples, which include a vehicle power allocation state set. The vehicle power allocation state set includes a vehicle communication state data set of all vehicles at the same time point, and the vehicle communication power allocation training samples are also obtained in the ongoing vehicle-building collaborative communication actual experiment. The vehicle communication power allocation training samples are labeled using the target transmission power allocation strategy. It should be noted that the target transmission power allocation strategy is consistent with the predicted transmission power allocation strategy and the target predicted transmission power allocation strategy, and the target transmission power allocation strategy is obtained by experts adjusting according to the vehicle communication state corresponding to the vehicle communication power allocation training samples. All labeled vehicle communication power allocation training samples are composed into a vehicle communication power allocation training set. The vehicle communication power allocation model is trained using the vehicle communication power allocation training set, and the vehicle communication power allocation loss value is calculated to determine whether the vehicle communication power allocation loss value is within the vehicle communication power allocation confidence range. The vehicle communication power allocation confidence range is manually set. If the vehicle communication power allocation loss value is within the vehicle communication power allocation confidence range, the trained vehicle communication power allocation model is output; otherwise, the vehicle communication power allocation model continues to be trained using the vehicle communication power allocation training set.

[0053] Obtain several communication power allocation strategy training samples, which include a vehicle power allocation state set and the transmission power allocation strategy it adopts. The vehicle power allocation state set and the transmission power allocation strategy it adopts here are consistent with the acquisition method of the above-mentioned vehicle communication power allocation training samples, and the communication power allocation strategy training samples are labeled by the strategy evaluation value. It should be noted that in the communication power allocation strategy training samples, the vehicle state after the vehicle power allocation state set adopts the transmission power allocation strategy is monitored, and the corresponding strategy evaluation value is calculated based on the vehicle state after the transmission power allocation strategy is adopted. All labeled communication power allocation strategy training samples are combined into a communication power allocation strategy training set. The vehicle communication power allocation strategy evaluation model is trained using the communication power allocation strategy training set, and the vehicle communication power allocation strategy evaluation loss value is calculated to determine whether the vehicle communication power allocation strategy evaluation loss value is within the vehicle communication power allocation evaluation confidence range. If the vehicle communication power allocation strategy evaluation loss value is within the vehicle communication power allocation evaluation confidence range, the trained vehicle communication power allocation strategy evaluation model is output; otherwise, the vehicle communication power allocation strategy evaluation model continues to be trained in real time using the communication power allocation strategy training set.

[0054] Example 2, see Figure 2 , a low-power vehicle-building collaborative communication system, comprising:

[0055] The vehicle communication status data set construction module is used to obtain the corresponding channel status data set for each vehicle. The channel status data set includes data describing the channel status such as the signal-to-interference-noise ratio and bit error rate obtained by the vehicle based on the pilot signal detection, and obtains the corresponding position information of the vehicle. Then, the channel status data set corresponding to the vehicle and the position information are combined to form the vehicle communication status data set. It should be noted that in the process of guiding the vehicle to park through the vehicle-building collaborative communication, the building node will periodically send a pilot signal to a certain range. The pilot signal (Pilot signal) A pilot signal is a predefined, known signal sequence used for link establishment, channel estimation, and synchronization in wireless communication systems. When a vehicle receives a pilot signal, it establishes a communication link based on the owner's parking intention and conducts real-time communication based on the communication link to provide parking guidance. Parking guidance specifically includes the following: dynamically planning parking paths, updating the availability of parking spaces, and providing accurate parking navigation (such as remaining distance and turn guidance) for vehicles. It also measures the received signal strength, signal-to-noise ratio, and channel capacity of the pilot signal, and other data describing the channel state to form a channel state data set. The channel state data set and location information are combined to form a vehicle communication state data set and fed back to the building node. The building node will perform subsequent transmission power adjustments based on the vehicle's vehicle communication state data set. In addition, between the building node and the vehicle that have established a communication link, the vehicle will also regularly feed back the vehicle communication state data set to the building node during real-time communication. It should be noted that the vehicle's location information will affect the power loss during the communication process.

[0056] The transmit power adjustment module is used to allocate transmit power to all vehicles at the transmit power adjustment time point according to the predicted transmit power allocation strategy corresponding to the transmit power adjustment time point. It should be noted that the building node can set different transmit powers for different vehicles through the multi-antenna system, and the predicted transmit power allocation strategy can reduce the overall transmit power of the building node while ensuring the communication quality of the vehicle, thereby reducing the energy consumption of vehicle-building collaborative communication;

[0057] The construction of the predictive transmit power allocation strategy includes the following steps:

[0058] At the building node, the following is performed for each vehicle with an established communication link: at the current transmission power adjustment time point, all vehicle communication status data sets corresponding to the vehicle are arranged in chronological order from front to back, and then N vehicle communication status data sets are selected from all the arranged vehicle communication status data sets from back to front to form a vehicle communication status timing analysis data set, and the N vehicle communication status data sets in the vehicle communication status timing analysis data set are in chronological order from front to back. It should be noted that the building node will continuously receive the vehicle communication status data sets uploaded by the vehicle on a regular basis, but when the total amount of the vehicle communication status data sets corresponding to the vehicle exceeds the upper limit, the vehicle communication status data sets with an older time will be cleared, and when the vehicle is in contact with the building node, When the communication link between the points is disconnected, all vehicle communication status data sets corresponding to the vehicle will be cleared to obtain the vehicle communication status prediction model. The vehicle communication status prediction model is established based on the Transformer model and is trained based on all vehicles within the preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model. The preset range here is determined by the developer, generally 20m. It should be noted that at the building node, the vehicle communication status prediction model will be regularly updated through real-time training, and this vehicle communication status prediction model is for all vehicles, but because the actual environment faced by the vehicle will change, such as vehicle obstruction or obstacle obstruction, etc., each vehicle will be based on The vehicle communication status prediction model is adjusted based on the actual vehicle communication status information of all vehicles within the preset range, so that the trained short-term vehicle communication status prediction model can fit the actual environmental conditions, thereby improving the accuracy of subsequent predictions of the vehicle communication status data set, and then the vehicle communication status time series analysis data set is sent to the short-term vehicle communication status prediction model for processing, and the predicted vehicle communication status data set corresponding to the next transmission power adjustment time point is output; and the predicted vehicle communication status data sets corresponding to all vehicles are arranged from large to small according to the distance from the building node to form a predicted vehicle power allocation state set, which can reflect the channel status and position of all vehicles at the next transmission power adjustment time point. The predicted vehicle power allocation state set is then sent to the vehicle communication power allocation model for processing, and a predicted transmission power allocation strategy is output. The vehicle communication power allocation model is established based on the BP neural network. The predicted transmission power allocation strategy includes the transmission power corresponding to all vehicles that have established communication links. At the transmission power adjustment time point, the vehicle communication power allocation model is trained in real time, so that the predicted transmission power allocation strategy is more in line with the constraint of reducing the overall transmission power of the building node while meeting the communication quality of the vehicle. The vehicle communication state prediction model is also updated through real-time training at the model update time point, and several transmission power adjustment time points are included between two adjacent model update time points.

[0059] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail herein is prior art known to those skilled in the art.

Claims

1. A low-power vehicle-building collaborative communication method, characterized in that: include: Obtain the corresponding channel state data set for each vehicle, and obtain the corresponding location information of the vehicle, and then combine the corresponding channel state data set and location information of the vehicle to form a vehicle communication state data set; At the transmit power adjustment time point, transmit power is allocated to all vehicles using the predicted transmit power allocation strategy corresponding to the transmit power adjustment time point; The construction of the predictive transmit power allocation strategy includes the following steps: At the building node, the following is performed for each vehicle with an established communication link: at the current transmission power adjustment time point, all vehicle communication status data sets corresponding to the vehicle are arranged in chronological order from front to back, and then N vehicle communication status data sets are selected from all the arranged vehicle communication status data sets from back to front to form a vehicle communication status timing analysis data set, and the N vehicle communication status data sets in the vehicle communication status timing analysis data set are in chronological order from front to back, and a vehicle communication status prediction model is obtained. The vehicle communication status prediction model is established based on the Transformer model, and the vehicle communication status prediction model is trained based on all vehicles within a preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model. The vehicle communication status timing analysis data set is then sent to the short-term vehicle communication status prediction model for processing, and the predicted vehicle communication status data set corresponding to the next transmission power adjustment time point is output; the predicted vehicle communication status data sets corresponding to all vehicles are arranged from large to small according to the distance from the building node to form a predicted vehicle power allocation state set, and the predicted vehicle power allocation state set is then sent to the vehicle communication power allocation model for processing, and the predicted transmission power allocation strategy is output; and at the transmission power adjustment time point, the vehicle communication power allocation model is trained in real time; and the vehicle communication status prediction model is also updated through real-time training at the model update time point, and several transmission power adjustment time points are included between two adjacent model update time points.

2. A low-power vehicle-building collaborative communication method according to claim 1, characterized in that: The vehicle communication status prediction model is trained based on all vehicles within a preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model, which specifically includes the following steps: Vehicles within a preset distance range from the vehicle are recorded as target vehicles, and all target vehicles are grouped into a target vehicle set. The target vehicle set is traversed, and the following operations are performed for each target vehicle: obtain the predicted vehicle communication status dataset and vehicle communication status dataset corresponding to the target vehicle at the current transmission power adjustment time point, and obtain N vehicle communication status datasets of the target vehicle before the current transmission power adjustment time point, perform a differential operation on the predicted vehicle communication status dataset and the vehicle communication status dataset corresponding to the target vehicle at the current transmission power adjustment time point, and construct a difference vector, and combine the N vehicle communication status datasets of the target vehicle before the current transmission power adjustment time point, the vehicle communication status dataset corresponding to the current transmission power adjustment time point, and the difference vector to form a short-term training sample; sum and average the difference vectors corresponding to all target vehicles to construct a comprehensive difference vector, and combine the short-term training samples corresponding to all target vehicles into a short-term training set The vehicle communication status prediction model is trained using a short-time training set. During the training, N vehicle communication status data sets of the target vehicle before the current transmission power adjustment time point are used as the input of the vehicle communication status prediction model, and the vehicle communication status data set corresponding to the current transmission power adjustment time point is used as the target output of the vehicle communication status prediction model. The difference vector is used as the query vector when the vehicle communication status prediction model executes the self-attention mechanism. The short-time training loss value is calculated to determine whether the short-time training loss value is within the vehicle communication status prediction confidence range. If the short-time training loss value is within the vehicle communication status prediction confidence range, the short-time vehicle communication status prediction model is output, and the difference vector used when executing the self-attention mechanism in the short-time vehicle communication status prediction model is the comprehensive difference vector. If the short-time training loss value is not within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is continued to be trained using the short-time training set.

3. A low-power vehicle-building collaborative communication method according to claim 2, characterized in that: Real-time training of the vehicle communication power allocation model includes the following steps: The predicted vehicle power allocation state set corresponding to the previous transmission power adjustment time point, the predicted transmission power allocation strategy output at the previous transmission power adjustment time point, and the vehicle power allocation state set corresponding to the current transmission power adjustment time point are composed of a strategy evaluation set, and the vehicle power allocation state set is composed of vehicle communication state data sets corresponding to all vehicles arranged from large to small according to the distance from the building node, and the strategy evaluation value is calculated for the predicted transmission power allocation strategy in the strategy evaluation set and the vehicle power allocation state set corresponding to the current transmission power adjustment time point; the predicted vehicle power allocation state set corresponding to the previous transmission power adjustment time point and the predicted transmission power allocation strategy output at the previous transmission power adjustment time point are composed of vehicle communication power allocation real-time training samples, and the vehicle communication power allocation real-time training samples are evaluated by the strategy evaluation value. The training samples are labeled; all labeled vehicle communication power allocation real-time training samples are converted into a vehicle communication power allocation real-time training set; the vehicle communication power allocation strategy evaluation model is trained in real time through the vehicle communication power allocation real-time training set. During the real-time training, the vehicle communication power allocation real-time training samples are used as input, and the strategy evaluation value is used as the target output. The vehicle communication power allocation real-time training loss value is calculated to determine whether the vehicle communication power allocation real-time training loss value is within the vehicle communication power allocation evaluation confidence range. If the vehicle communication power allocation real-time training loss value is within the vehicle communication power allocation evaluation confidence range, the trained vehicle communication power allocation strategy evaluation model is output; otherwise, the vehicle communication power allocation strategy evaluation model is continued to be trained in real time through the vehicle communication power allocation real-time training set; The predicted vehicle power allocation state set corresponding to the current transmission power adjustment time point is sent to the target vehicle communication power allocation model for processing, and the target predicted transmission power allocation strategy is output. The predicted vehicle power allocation state set corresponding to the current transmission power adjustment time point and the target predicted transmission power allocation strategy are combined and sent to the trained vehicle communication power allocation strategy evaluation model, and the target strategy evaluation value is output. Based on the target strategy evaluation value, the parameters of the vehicle communication power allocation model are adjusted by the gradient ascent method to realize real-time training of the vehicle communication power allocation model; and it also includes updating the target vehicle communication power allocation model at the model update time point.

4. A low-power vehicle-building collaborative communication method according to claim 3, characterized in that: The target vehicle communication power allocation model is updated at the target model update time point, specifically including the following steps: directly replacing the target vehicle communication power allocation model with the vehicle communication power allocation model at the target model update time point, and in the initial state, the vehicle communication power allocation model and the target vehicle communication power allocation model are consistent.

5. A low-power vehicle-building collaborative communication method according to claim 4, characterized in that: At the model update time point, the vehicle communication status prediction model is updated through real-time training, which specifically includes the following steps: At a model update time point, a vehicle communication status prediction real-time training sample between the current model update time point and the previous model update time point is obtained. The vehicle communication status prediction real-time training sample is N+1 vehicle communication status data sets corresponding to any vehicle arranged in chronological order. All vehicle communication status prediction real-time training samples are combined into a vehicle communication status prediction real-time training set. The vehicle communication status prediction model is trained in real time using the vehicle communication status prediction real-time training set. During the real-time training, the first N vehicle communication status data sets in the vehicle communication status prediction real-time training set are used as inputs of the vehicle communication status prediction model, and the last vehicle communication status data set in the vehicle communication status prediction real-time training set is used as the target output of the vehicle communication status prediction model. A vehicle communication status prediction real-time training loss value is calculated, and it is determined whether the vehicle communication status prediction real-time training loss value is within a vehicle communication status prediction confidence range. If the vehicle communication status prediction real-time training loss value is within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is updated. If the vehicle communication status prediction real-time training loss value is not within the vehicle communication status prediction confidence range, the vehicle communication status prediction model is continued to be trained using the vehicle communication status prediction real-time training set.

6. A low-power vehicle-building collaborative communication method according to claim 5, characterized in that: It also includes pre-training of the vehicle communication status prediction model. The specific steps are as follows: A number of vehicle communication status prediction pre-training samples are obtained. The vehicle communication status prediction pre-training samples include N+1 vehicle communication status data sets corresponding to any vehicle and arranged in chronological order. All vehicle communication status prediction pre-training samples are combined into a vehicle communication status prediction pre-training set, and a vehicle communication status prediction model is pre-trained using the vehicle communication status prediction pre-training set. The pre-training method is consistent with the real-time training method of the vehicle communication status prediction model.

7. A low-power vehicle-building collaborative communication method according to claim 6, characterized in that: The training of the vehicle communication power allocation model and the vehicle communication power allocation strategy evaluation model specifically includes the following steps: Obtain several vehicle communication power allocation training samples, the vehicle communication power allocation training samples include a vehicle power allocation state set, the vehicle power allocation state set includes a vehicle communication state data set of all vehicles at the same time point, the vehicle communication power allocation training samples are labeled according to the target transmission power allocation strategy, all labeled vehicle communication power allocation training samples are combined into a vehicle communication power allocation training set, a vehicle communication power allocation model is trained using the vehicle communication power allocation training set, a vehicle communication power allocation loss value is calculated, and it is determined whether the vehicle communication power allocation loss value is within the vehicle communication power allocation confidence range. If the vehicle communication power allocation loss value is within the vehicle communication power allocation confidence range, the trained vehicle communication power allocation model is output; otherwise, the vehicle communication power allocation model is continued to be trained using the vehicle communication power allocation training set; Acquire several communication power allocation strategy training samples, which include a vehicle power allocation state set and the transmission power allocation strategy it adopts, and label the communication power allocation strategy training samples with a strategy evaluation value, and form a communication power allocation strategy training set with all labeled communication power allocation strategy training samples. Train a vehicle communication power allocation strategy evaluation model with the communication power allocation strategy training set, calculate a vehicle communication power allocation strategy evaluation loss value, and determine whether the vehicle communication power allocation strategy evaluation loss value is within a vehicle communication power allocation evaluation confidence range. If the vehicle communication power allocation strategy evaluation loss value is within the vehicle communication power allocation evaluation confidence range, output the trained vehicle communication power allocation strategy evaluation model; otherwise, continue to train the vehicle communication power allocation strategy evaluation model in real time with the communication power allocation strategy training set.

8. A low-power vehicle-building collaborative communication system, characterized in that: The system applies the low-power vehicle-building collaborative communication method according to any one of claims 1 to 7, including: The vehicle communication status data set construction module is used to obtain the corresponding channel status data set for each vehicle and the corresponding location information of the vehicle, and then combine the corresponding channel status data set and the location information of the vehicle to form a vehicle communication status data set; A transmission power adjustment module is used to allocate transmission power to all vehicles at a transmission power adjustment time point according to a predicted transmission power allocation strategy corresponding to the transmission power adjustment time point; The construction of the predictive transmit power allocation strategy includes the following steps: At the building node, the following is performed for each vehicle with an established communication link: at the current transmission power adjustment time point, all vehicle communication status data sets corresponding to the vehicle are arranged in chronological order from front to back, and then N vehicle communication status data sets are selected from all the arranged vehicle communication status data sets from back to front to form a vehicle communication status timing analysis data set, and the N vehicle communication status data sets in the vehicle communication status timing analysis data set are in chronological order from front to back, and a vehicle communication status prediction model is obtained. The vehicle communication status prediction model is established based on the Transformer model, and the vehicle communication status prediction model is trained based on all vehicles within a preset distance range of the vehicle to obtain a short-term vehicle communication status prediction model. The vehicle communication status timing analysis data set is then sent to the short-term vehicle communication status prediction model for processing, and the predicted vehicle communication status data set corresponding to the next transmission power adjustment time point is output; the predicted vehicle communication status data sets corresponding to all vehicles are arranged from large to small according to the distance from the building node to form a predicted vehicle power allocation state set, and the predicted vehicle power allocation state set is then sent to the vehicle communication power allocation model for processing, and the predicted transmission power allocation strategy is output; and at the transmission power adjustment time point, the vehicle communication power allocation model is trained in real time; and the vehicle communication status prediction model is also updated through real-time training at the model update time point, and several transmission power adjustment time points are included between two adjacent model update time points.

Images