CN117692083A - Satellite communication channel modeling method and device - Google Patents
Satellite communication channel modeling method and device Download PDFInfo
- Publication number
- CN117692083A CN117692083A CN202311696480.5A CN202311696480A CN117692083A CN 117692083 A CN117692083 A CN 117692083A CN 202311696480 A CN202311696480 A CN 202311696480A CN 117692083 A CN117692083 A CN 117692083A
- Authority
- CN
- China
- Prior art keywords
- neural network
- communication channel
- satellite communication
- state
- symbol
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004891 communication Methods 0.000 title claims abstract description 126
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000013528 artificial neural network Methods 0.000 claims abstract description 154
- 238000012545 processing Methods 0.000 claims abstract description 44
- 230000014509 gene expression Effects 0.000 claims abstract description 32
- 230000009471 action Effects 0.000 claims description 53
- 230000006870 function Effects 0.000 claims description 53
- 238000011156 evaluation Methods 0.000 claims description 48
- 238000012549 training Methods 0.000 claims description 23
- 238000005070 sampling Methods 0.000 claims description 19
- 238000005457 optimization Methods 0.000 claims description 10
- 238000012163 sequencing technique Methods 0.000 claims description 8
- 238000013461 design Methods 0.000 claims description 5
- 230000002787 reinforcement Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 5
- 238000011161 development Methods 0.000 abstract description 3
- 230000001737 promoting effect Effects 0.000 abstract description 3
- 230000007246 mechanism Effects 0.000 description 6
- 238000013178 mathematical model Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000004590 computer program Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Radio Relay Systems (AREA)
Abstract
The invention discloses a satellite communication channel modeling method and device, wherein the method comprises the following steps: acquiring satellite communication channel data information; the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency; constructing a symbolic regression model based on a transducer; processing the satellite communication channel data information to obtain a transducer symbol regression system parameter; and processing the parameters of the transform symbol regression system and the satellite communication channel data information according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model, so as to realize satellite communication channel modeling. The method of the invention saves manpower and avoids the problem of unexplainability of the neural network, has the characteristics of simplicity and easiness in implementation, and is beneficial to promoting the development of a satellite communication channel modeling method.
Description
Technical Field
The present invention relates to the field of wireless communications technologies, and in particular, to a satellite communication channel modeling method and apparatus.
Background
Channel modeling has a significant role in satellite communication system design. Traditional channel modeling methods require significant time and labor costs to fit a mathematical model of the channel after the measured satellite communication channel data is obtained. To solve this problem, researchers have used neural networks directly as channel models. However, the method using the neural network directly as a channel model makes it difficult to reveal mathematical relationships and physical mechanisms between channel parameters because of the non-interpretability of the neural network.
Disclosure of Invention
The invention aims at solving the technical problems that a great deal of time and labor cost are required for satellite communication system channel modeling and the interpretability of using a neural network to directly serve as a channel model is insufficient, and provides a satellite communication channel modeling method and device.
In order to solve the technical problem, a first aspect of the embodiment of the present invention discloses a satellite communication channel modeling method, which includes:
s1, acquiring satellite communication channel data information;
the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency;
s2, constructing a symbolic regression model based on a transducer;
s3, processing the satellite communication channel data information to obtain a transducer symbol regression system parameter;
s4, processing the parameters of the transform symbol regression system and the satellite communication channel data information according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model, and realizing satellite communication channel modeling.
In a first aspect of the present embodiment, the constructing a symbolic regression model based on a transducer includes:
s21, constructing a neural network structure and a mathematical representation model of the neural network;
s22, establishing a symbol search space of a symbol regression system based on the neural network structure;
s23, establishing environment, state, action and rewarding function based on the neural network structure, the mathematical representation model of the neural network and the symbol search space to obtain a symbol regression initial model;
and S24, training the initial symbol regression model by utilizing a near-end strategy optimization method according to the satellite communication channel data information to obtain a symbol regression model based on a transducer.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the neural network structure includes an input linear layer, an encoder of a transducer neural network, and an output linear layer;
the symbol search space includes the satellite communication channel data information and operational symbols.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the establishing an environment, a state, an action, and a reward function includes:
s231, processing the mathematical representation model of the neural network to obtain an environment;
the environment is a binary tree, and the binary tree is composed of variables, coefficients, operation symbols, monocular operators and binocular operators; the variables and the coefficients are nodes with the degree of 0, the operation symbol is a node with the degree of not 0, the degree of the monocular operator is 1, and the degree of the binocular operator is 2;
s232, the current node, the previous node, the father node and the brother node of the binary tree form a state;
the state is:
{ current node, previous node, father node, sibling node }
S233, selecting one symbol in the symbol space as an action;
s234, establishing a reward function;
the reward function is:
wherein r is a prize value,σ y is the standard deviation of the target value, N is the number of sampling points, x i ,y i For the measured channel data points and labels, i=1, 2, …, N, f (·) is a function of the neural network fit.
In a first aspect of the embodiment of the present invention, the training the initial symbolic regression model according to the satellite communication channel data information by using a near-end policy optimization method to obtain a symbolic regression model based on a transducer includes:
s241, based on the neural network structure, establishing a strategy neural network and an evaluation neural network;
s242, processing the state by using the strategy neural network to obtain probability distribution of elements in a symbol space;
s243, sampling the action according to the probability distribution to obtain an action sampling value;
s244, processing the current state and the current action in the action sampling value to obtain a next state, and updating the expression { current node, previous node, father node, brother node };
s245, processing the next state and judging whether the next state reaches a termination state or not; if the ending state is reached, processing the satellite communication channel data information by utilizing the rewarding function to obtain a rewarding value; if the state of termination is not reached, the current state and the current action are processed by using the evaluation neural network, and a reward value is obtained;
s246, storing the expression { current state, next state, current action, rewarding value } into an experience pool;
s247, judging whether a network needs to be updated, if so, updating the strategy neural network and the evaluation neural network to obtain an updated strategy neural network and an updated evaluation neural network;
s248, executing S242-S247 based on the updating strategy neural network and the updating evaluation neural network;
and S249, stopping training when the reward reaches 1 or the preset maximum training times, and obtaining a symbolic regression model based on the transducer.
In a first aspect of the embodiment of the present invention, updating the policy neural network and the evaluation neural network to obtain an updated policy neural network and an updated evaluation neural network includes:
s2471, when the training times reach a preset threshold value, sequencing the expressions in the experience pool according to the rewarding value of the termination node, and processing the expressions of which the number is 10% before sequencing to obtain the estimated network loss;
the evaluation network loss is:
critic_loss＝MSE(f c (state)-critic_rewards)
wherein critic_loss is an estimated network loss, f c (. Cndot.) is the estimated neural network, state is state, next_state is next state, r is the prize value, and critical_rewards=r+γ×f c (next_state), MSE (·) is a mean square error function, γ is a discount coefficient in deep reinforcement learning;
s2472, back-propagating the evaluation network loss to obtain an updated evaluation neural network;
s2473, back-propagating a preset strategy network loss function to obtain an updated strategy neural network;
the preset strategy network loss function is as follows:
wherein, actor_loss is a strategic network loss function, E # - []The table design finds mathematical expectations, min () represents the minimum,representing the current policy neural network f a (. About.) probability of getting action a in state,/>Representing a previous strategic neural network f a_old (. Cndot.) the probability of action a is obtained at state, A is the dominance function and clip (-) is the cut-off function.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the parameters of the transform symbol regression system include a learning rate, a batch_size, a discount coefficient, an iteration number m, and an exit condition;
the exit condition is that the iteration number is greater than m or the reward reaches 1.
The second aspect of the embodiment of the invention discloses a satellite communication channel modeling device, which comprises:
the data acquisition module is used for acquiring satellite communication channel data information;
the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency;
the symbolic regression model construction module is used for constructing a symbolic regression model based on a transducer;
the symbol regression system parameter acquisition module is used for processing the satellite communication channel data information to obtain a transducer symbol regression system parameter;
and the satellite communication channel modeling module is used for processing the parameters of the transform symbol regression system according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model and realize satellite communication channel modeling.
In a second aspect of the present embodiment, the constructing a symbolic regression model based on a transducer includes:
s21, constructing a neural network structure and a mathematical representation model of the neural network;
s22, establishing a symbol search space of a symbol regression system based on the neural network structure;
s23, establishing environment, state, action and rewarding function based on the neural network structure, the mathematical representation model of the neural network and the symbol search space to obtain a symbol regression initial model;
and S24, training the initial symbol regression model by utilizing a near-end strategy optimization method according to the satellite communication channel data information to obtain a symbol regression model based on a transducer.
As an alternative implementation manner, in the second aspect of the embodiment of the present invention, the neural network structure includes an input linear layer, an encoder of a transducer neural network, and an output linear layer;
the symbol search space includes the satellite communication channel data information and operational symbols.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the establishing an environment, a state, an action, and a reward function includes:
s231, processing the mathematical representation model of the neural network to obtain an environment;
the environment is a binary tree, and the binary tree is composed of variables, coefficients, operation symbols, monocular operators and binocular operators; the variables and the coefficients are nodes with the degree of 0, the operation symbol is a node with the degree of not 0, the degree of the monocular operator is 1, and the degree of the binocular operator is 2;
s232, the current node, the previous node, the father node and the brother node of the binary tree form a state;
the state is:
{ current node, previous node, father node, sibling node }
S233, selecting one symbol in the symbol space as an action;
s234, establishing a reward function;
the reward function is:
wherein r is a prize value,σ y is the standard deviation of the target value, N is the number of sampling points, x i ,y i For the measured channel data points and labels, i=1, 2, …, N, f (·) is a function of the neural network fit.
In a first aspect of the embodiment of the present invention, the training the initial symbolic regression model according to the satellite communication channel data information by using a near-end policy optimization method to obtain a symbolic regression model based on a transducer includes:
s241, based on the neural network structure, establishing a strategy neural network and an evaluation neural network;
s242, processing the state by using the strategy neural network to obtain probability distribution of elements in a symbol space;
s243, sampling the action according to the probability distribution to obtain an action sampling value;
s244, processing the current state and the current action in the action sampling value to obtain a next state, and updating the expression { current node, previous node, father node, brother node };
s245, processing the next state and judging whether the next state reaches a termination state or not; if the ending state is reached, processing the satellite communication channel data information by utilizing the rewarding function to obtain a rewarding value; if the state of termination is not reached, the current state and the current action are processed by using the evaluation neural network, and a reward value is obtained;
s246, storing the expression { current state, next state, current action, rewarding value } into an experience pool;
s247, judging whether a network needs to be updated, if so, updating the strategy neural network and the evaluation neural network to obtain an updated strategy neural network and an updated evaluation neural network;
s248, executing S242-S247 based on the updating strategy neural network and the updating evaluation neural network;
and S249, stopping training when the reward reaches 1 or the preset maximum training times, and obtaining a symbolic regression model based on the transducer.
In a second aspect of the embodiment of the present invention, updating the policy neural network and the evaluation neural network to obtain an updated policy neural network and an updated evaluation neural network includes:
s2471, when the training times reach a preset threshold value, sequencing the expressions in the experience pool according to the rewarding value of the termination node, and processing the expressions of which the number is 10% before sequencing to obtain the estimated network loss;
the evaluation network loss is:
critic_loss＝MSE(f c (state)-critic_rewards)
wherein critic_loss is an estimated network loss, f c (. Cndot.) is the estimated neural network, state is state, next_state is next state, r is the prize value, and critical_rewards=r+γ×f c (next_state), MSE (·) is a mean square error function, γ is a discount coefficient in deep reinforcement learning;
s2472, back-propagating the evaluation network loss to obtain an updated evaluation neural network;
s2473, back-propagating a preset strategy network loss function to obtain an updated strategy neural network;
the preset strategy network loss function is as follows:
wherein, actor_loss is a strategic network loss function, E # - []The table design finds mathematical expectations, min () represents the minimum,representing the current policy neural network f a (. About.) probability of getting action a in state,/>Representing a previous strategic neural network f a_old (. Cndot.) the probability of action a is obtained at state, A is the dominance function and clip (-) is the cut-off function.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the parameters of the transform symbol regression system include a learning rate, a batch_size, a discount coefficient, an iteration number m, and an exit condition;
the exit condition is that the iteration number is greater than m or the reward reaches 1.
Another satellite communication channel modeling apparatus is disclosed in a third aspect of the invention, the apparatus comprising:
a memory storing executable program code;
a processor coupled to the memory;
the processor invokes the executable program code stored in the memory to perform some or all of the steps in the satellite communication channel modeling method disclosed in the first aspect of the embodiment of the present invention.
A fourth aspect of the invention discloses a computer-readable medium storing computer instructions that, when invoked, are adapted to perform part or all of the steps of the satellite communication channel modeling method disclosed in the first aspect of the embodiments of the invention.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
the invention provides a satellite communication channel modeling method based on a transducer symbol regression, which is convenient for reasoning a mathematical expression model from satellite communication channel data obtained by a large amount of measurement, solves the problem of time and labor waste of the traditional channel modeling method, avoids the problem of a channel 'black box' caused by directly using a neural network as a channel model, is beneficial to efficiently establishing the channel model, reveals mathematical relations and physical mechanisms among channel parameters, and is beneficial to promoting the development of a wireless communication system.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a satellite communication channel modeling method according to an embodiment of the present invention;
FIG. 2 is a flow chart of another satellite communication channel modeling method disclosed in an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a satellite communication channel modeling apparatus according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of another satellite communication channel modeling apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the present invention better understood by those skilled in the art, the following description will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The terms first, second and the like in the description and in the claims and in the above-described figures are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order. Furthermore, the terms "comprise" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, apparatus, article, or device that comprises a list of steps or elements is not limited to the list of steps or elements but may, in the alternative, include other steps or elements not expressly listed or inherent to such process, method, article, or device.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
The invention discloses a satellite communication channel modeling method and device, wherein the method comprises the following steps: acquiring satellite communication channel data information; the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency; constructing a symbolic regression model based on a transducer; processing the satellite communication channel data information to obtain a transducer symbol regression system parameter; and processing the parameters of the transform symbol regression system and the satellite communication channel data information according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model, so as to realize satellite communication channel modeling. The method of the invention saves manpower and avoids the problem of unexplainability of the neural network, has the characteristics of simplicity and easiness in implementation, and is beneficial to promoting the development of a satellite communication channel modeling method. The following will describe in detail.
Example 1
Referring to fig. 1, fig. 1 is a flowchart of a satellite communication channel modeling method according to an embodiment of the invention. The satellite communication channel modeling method described in fig. 1 is applied to the technical field of wireless communication, a channel mathematical model is directly fitted from channel measurement data by using a transducer neural network as a realizing tool of symbol regression, manpower is saved, the problem of unexplainability of the neural network is avoided, the mathematical relationship and physical mechanism between channel parameters can be revealed, and the embodiment of the invention is not limited. As shown in fig. 1, the satellite communication channel modeling method may include the following operations:
s1, acquiring satellite communication channel data information;
the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency;
s2, constructing a symbolic regression model based on a transducer;
s3, processing the satellite communication channel data information to obtain a transducer symbol regression system parameter;
s4, processing the parameters of the transform symbol regression system and the satellite communication channel data information according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model, and realizing satellite communication channel modeling.
The method comprises the following steps: the satellite communication channel data information is directly input into a symbolic regression model based on a transducer, and then the transducer outputs a formula, and the formula is a mathematical expression model of the satellite communication channel.
Optionally, the constructing a symbolic regression model based on the transducer includes:
s21, constructing a neural network structure and a mathematical representation model of the neural network;
s22, establishing a symbol search space of a symbol regression system based on the neural network structure;
s23, establishing environment, state, action and rewarding function based on the neural network structure, the mathematical representation model of the neural network and the symbol search space to obtain a symbol regression initial model;
and S24, training the initial symbol regression model by utilizing a near-end strategy optimization method according to the satellite communication channel data information to obtain a symbol regression model based on a transducer.
Optionally, the neural network structure includes an input linear layer, an encoder (encoder) of a transducer neural network, and an output linear layer;
the transducer neural network has certain memory, can memorize context information when processing sequence data, captures the interrelation between the sequence data, and has certain advantages when processing sequence data such as mathematical expression.
The symbol search space includes the satellite communication channel data information and operational symbols. The satellite communication channel data information is the variable in the measured satellite communication channel data, and various mathematical operations are performed:
symbol space = { channel variable, operation symbol }
The variables required for the channel model in different situations are different, taking the free space path loss model as an example: the variables are the signal frequency f, the distance d between the transmitter and the receiver, and the path loss magnitude pl, and the invention is not limited and can be determined according to practice.
Optionally, the establishing the environment, the state, the action and the reward function includes:
s231, processing the mathematical representation model of the neural network to obtain an environment;
the environment is a binary tree, and the binary tree is composed of variables, coefficients, operation symbols, monocular operators and binocular operators; the variables and the coefficients are nodes with the degree of 0, the operation symbol is a node with the degree of not 0, the degree of the monocular operator is 1, and the degree of the binocular operator is 2;
s232, the current node, the previous node, the father node and the brother node of the binary tree form a state;
the state is:
{ current node, previous node, father node, sibling node }
S233, selecting one symbol in the symbol space as an action;
s234, establishing a reward function;
the reward function is:
wherein r is a prize value,σ y is the standard deviation of the target value, N is the number of sampling points, x i ,y i For the measured channel data points and labels, i=1, 2, …, N, f (·) is a function of the neural network fit.
Optionally, the training the initial symbolic regression model according to the satellite communication channel data information by using a near-end policy optimization method to obtain a symbolic regression model based on a transducer includes:
s241, based on the neural network structure, establishing a strategy neural network and an evaluation neural network;
s242, processing the state by using the strategy neural network to obtain probability distribution of elements in a symbol space;
s243, sampling the action according to the probability distribution to obtain an action sampling value;
s244, processing the current state and the current action in the action sampling value to obtain a next state, and updating the expression { current node, previous node, father node, brother node };
s245, processing the next state and judging whether the next state reaches a termination state or not; if the ending state is reached, processing the satellite communication channel data information by utilizing the rewarding function to obtain a rewarding value; if the state of termination is not reached, the current state and the current action are processed by using the evaluation neural network, and a reward value is obtained;
s246, storing the expression { current state, next state, current action, rewarding value } into an experience pool;
s247, judging whether a network needs to be updated, if so, updating the strategy neural network and the evaluation neural network to obtain an updated strategy neural network and an updated evaluation neural network;
s248, based on the updating strategy neural network and the updating evaluation neural network, updating the strategy neural network in S242 into an updating strategy neural network, updating the evaluation neural network in S245 into an updating evaluation neural network, and executing S242-S247;
and S249, stopping training when the reward reaches 1 or the preset maximum training times, and obtaining a symbolic regression model based on the transducer.
Optionally, the updating the policy neural network and the evaluation neural network to obtain an updated policy neural network and an updated evaluation neural network includes:
s2471, when the training times reach a preset threshold value, sequencing the expressions in the experience pool according to the rewarding value of the termination node, and processing the expressions of which the number is 10% before sequencing to obtain the estimated network loss;
the evaluation network loss is:
critic_loss＝MSE(f c (state)-critic_rewards)
wherein critic_loss is an estimated network loss, f c (. Cndot.) is the estimated neural network, state is state, next_state is next state, r is the prize value, and critical_rewards=r+γ×f c (next_state), MSE (·) is a mean square error function, γ is a discount coefficient in deep reinforcement learning;
s2472, back-propagating the evaluation network loss to obtain an updated evaluation neural network;
s2473, back-propagating a preset strategy network loss function to obtain an updated strategy neural network;
the preset strategy network loss function is as follows:
wherein, actor_loss is a strategic network loss function, E # - []Mathematical expectation for table design, min () represents minimumThe value of the sum of the values,representing the current policy neural network f a (. About.) probability of getting action a in state,/>Representing a previous strategic neural network f a_old (. Cndot.) the probability of getting action a at state, A is the dominance function, clip (-) is the truncate function, ε is a constant set, typically 0.2.
Optionally, the parameters of the transform symbolic regression system include a learning rate, batch_size (batch size), a discount coefficient, the iteration number m and an exit condition;
the exit condition is that the iteration number is greater than m or the reward reaches 1.
Example two
Referring to fig. 2, fig. 2 is a flowchart illustrating another satellite communication channel modeling method according to an embodiment of the invention. The satellite communication channel modeling method described in fig. 2 is applied to the technical field of wireless communication, a channel mathematical model is directly fitted from channel measurement data by using a transducer neural network as a realizing tool of symbol regression, manpower is saved, the problem of unexplainability of the neural network is avoided, the mathematical relationship and physical mechanism between channel parameters can be revealed, and the embodiment of the invention is not limited. As shown in fig. 2, the satellite communication channel modeling method may include the following operations:
step 1, establishing a symbolic regression system based on a transducer;
and 11, building a neural network architecture, wherein the neural network architecture comprises an input linear layer, an encoder part of a transducer neural network and an output linear layer. The transducer neural network has certain memory, can memorize context information when processing sequence data, captures the interrelation between the sequence data, and has certain advantages when processing sequence data such as mathematical expression;
step 12, establishing a symbol search space, wherein the space comprises:
variables in the measured satellite communication channel data, and various mathematical operations:
symbol space = { channel variable, operation symbol };
step 13, establishing environment, state, action and rewards:
step 131, the environment is: constructing a mathematical expression as a binary tree, wherein variables and coefficients are nodes with the degree of 0, an operation symbol is a node with the degree of not 0, the monocular operator degree is 1, and the binocular operator degree is 2;
step 132, the state is: the state consists of a current node, a previous node, a father node and a brother node in the binary tree;
step 133, the action is to select a symbol in the symbol space;
step 133, rewarding is:
wherein r is a prize, sigma y Is the standard deviation of the target value, N is the number of sampling points, x i ,y i For the measured channel data points and labels, f (·) is a function of the neural network fit;
step 14, training the symbol regression system by using the measured satellite communication channel data through a near-end policy optimization (Proximal Policy Optimization, PPO) algorithm:
step 141, establishing a policy neural network f according to the neural network architecture a (. Cndot.) and evaluation of neural network f c (·)；
Step 142, inputting the state into the policy neural network f a In (-), the strategic neural network outputs the probability distribution of the elements in the symbol space;
step 143, sampling action according to probability distribution;
step 144, according to the current state and action, the next state is obtained, i.e. the { current node, previous node, father node, brother node };
step 145, determine whether next_state is in a termination state: if all the nodes can not be added with child nodes, the final complete mathematical expression is obtained, and the termination state is reached;
step 146, calculating rewards: if the state is the termination state, calculating rewards according to a rewards formula, and if the state is not the termination state, inputting state states and action actions into the evaluation neural network f c In (-), outputting the assessed prize r;
step 147, storing { state, next_state, action, r } in the experience pool;
step 148, determining whether an update to the network is required: if training reaches a certain number of times, updating the strategy neural network f a (. Cndot.) and evaluation of neural network f c (-), sorting the expressions in the experience pool by the prize value of the termination node, sampling a set of data from the first 10% experience, calculating the loss of the evaluation network:
critic_rewards＝r+γ*f c (next_state) (0-4)
critic_loss＝MSE(f c (state)-critic_rewards) (0-5)
wherein MSE (·) is a mean square error function, gamma is a discount coefficient in deep reinforcement learning, and the critic_loss is counter-propagated to update and evaluate the neural network f c (·)；
Calculation strategy neural network f a Loss function of (-):
wherein,representing the current policy neural network f a (. Cndot.) the probability of getting action a at state,representing a previous strategic neural network f a_old (. Cndot.) the probability of action a is obtained at state, A is the dominance function and clip (-) is the cut-off function. The obtained actor_loss back propagation update strategy neural network f a (·)；
In step 149, training is stopped when the prize reaches 1 or the maximum number of exercises is reached.
Step 2, inputting the measured satellite communication channel data into the transducer symbol regression system;
step 3, setting a transducer symbol regression system parameter according to satellite communication channel data;
the method of setting the parameters of the transform symbolic regression system is prior art in the art, and the invention is not limited.
Step 31, determining measured satellite communication channel data, including { d, p, f }, where d is a distance between receiving ends, p is signal power, and f is signal frequency, where the channel data is only an example;
and step 32, setting the super parameters of the symbolic regression system, including the learning rate, the batch_size, the discount coefficient and the iteration number m, wherein the exit condition is that the iteration number is larger than m or the reward reaches 1.
And 4, outputting a satellite communication channel mathematical expression model by the transducer symbol regression system. Thus completing modeling of the channel.
Example III
Referring to fig. 3, fig. 3 is a schematic structural diagram of a satellite communication channel modeling apparatus according to an embodiment of the invention. The satellite communication channel modeling device described in fig. 3 is applied to the technical field of wireless communication, a channel mathematical model is directly fitted from channel measurement data by using a transducer neural network as a realizing tool of symbol regression, manpower is saved, the problem of unexplainability of the neural network is avoided, the mathematical relationship and physical mechanism between channel parameters can be revealed, and the embodiment of the invention is not limited. As shown in fig. 3, the satellite communication channel modeling apparatus may include the following operations:
s301, a data acquisition module is used for acquiring satellite communication channel data information;
the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency;
s302, a symbolic regression model building module is used for building a symbolic regression model based on a transducer;
s303, a symbol regression system parameter acquisition module is used for processing the satellite communication channel data information to obtain a transducer symbol regression system parameter;
s304, a satellite communication channel modeling module is used for processing the parameters of the transform symbol regression system according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model so as to realize satellite communication channel modeling.
Example IV
Referring to fig. 4, fig. 4 is a schematic structural diagram of another satellite communication channel modeling apparatus according to an embodiment of the present invention. The satellite communication channel modeling device described in fig. 4 is applied to the technical field of wireless communication, a channel mathematical model is directly fitted from channel measurement data by using a transducer neural network as a realizing tool of symbol regression, manpower is saved, the problem of unexplainability of the neural network is avoided, the mathematical relationship and physical mechanism between channel parameters can be revealed, and the embodiment of the invention is not limited. As shown in fig. 4, the satellite communication channel modeling apparatus may include the following operations:
a memory 401 storing executable program codes;
a processor 402 coupled with the memory 401;
the processor 402 invokes executable program codes stored in the memory 401 for performing the steps in the satellite communication channel modeling method described in the first and second embodiments.
Example five
The embodiment of the invention discloses a computer readable storage medium storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute the steps in the satellite communication channel modeling method described in the first and second embodiments.
The apparatus embodiments described above are merely illustrative, in which the modules illustrated as separate components may or may not be physically separate, and the components shown as modules may or may not be physical, i.e., may be located in one place, or may be distributed over multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above detailed description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course by means of hardware. Based on such understanding, the foregoing technical solutions may be embodied essentially or in part in the form of a software product that may be stored in a computer-readable storage medium including Read-Only Memory (ROM), random-access Memory (Random Access Memory, RAM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM), one-time programmable Read-Only Memory (OTPROM), electrically erasable programmable Read-Only Memory (EEPROM), compact disc Read-Only Memory (Compact Disc Read-Only Memory, CD-ROM) or other optical disc Memory, magnetic disc Memory, tape Memory, or any other medium that can be used for computer-readable carrying or storing data.
Finally, it should be noted that: the embodiment of the invention discloses a satellite communication channel modeling method and device, which are disclosed as preferred embodiments of the invention, and are only used for illustrating the technical scheme of the invention, but not limiting the technical scheme; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that; the technical scheme recorded in the various embodiments can be modified or part of technical features in the technical scheme can be replaced equivalently; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (10)
1. A method of modeling a satellite communications channel, the method comprising:
s1, acquiring satellite communication channel data information;
the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency;
s2, constructing a symbolic regression model based on a transducer;
s3, processing the satellite communication channel data information to obtain a transducer symbol regression system parameter;
s4, processing the parameters of the transform symbol regression system and the satellite communication channel data information according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model, and realizing satellite communication channel modeling.
2. The method of claim 1, wherein the constructing a transform-based symbolic regression model comprises:
s21, constructing a neural network structure and a mathematical representation model of the neural network;
s22, establishing a symbol search space of a symbol regression system based on the neural network structure;
s23, establishing environment, state, action and rewarding function based on the neural network structure, the mathematical representation model of the neural network and the symbol search space to obtain a symbol regression initial model;
and S24, training the initial symbol regression model by utilizing a near-end strategy optimization method according to the satellite communication channel data information to obtain a symbol regression model based on a transducer.
3. The satellite communication channel modeling method of claim 2, wherein the neural network structure comprises an input linear layer, an encoder of a transducer neural network, and an output linear layer;
the symbol search space includes the satellite communication channel data information and operational symbols.
4. The method of modeling a satellite communications channel according to claim 2, wherein said establishing an environment, state, action and reward function comprises:
s231, processing the mathematical representation model of the neural network to obtain an environment;
the environment is a binary tree, and the binary tree is composed of variables, coefficients, operation symbols, monocular operators and binocular operators; the variables and the coefficients are nodes with the degree of 0, the operation symbol is a node with the degree of not 0, the degree of the monocular operator is 1, and the degree of the binocular operator is 2;
s232, the current node, the previous node, the father node and the brother node of the binary tree form a state;
the state is:
{ current node, previous node, father node, sibling node }
S233, selecting one symbol in the symbol space as an action;
s234, establishing a reward function;
the reward function is:
wherein r is a prize value,σ y is the standard deviation of the target value, N is the number of sampling points, x i ,y i For the measured channel data points and labels, i=1, 2, …, N, f (·) is a function of the neural network fit.
5. The method for modeling a satellite communication channel according to claim 2, wherein training the initial symbolic regression model according to the satellite communication channel data information by using the near-end policy optimization method to obtain a symbolic regression model based on a transducer comprises:
s241, based on the neural network structure, establishing a strategy neural network and an evaluation neural network;
s242, processing the state by using the strategy neural network to obtain probability distribution of elements in a symbol space;
s243, sampling the action according to the probability distribution to obtain an action sampling value;
s244, processing the current state and the current action in the action sampling value to obtain a next state, and updating the expression { current node, previous node, father node, brother node };
s245, processing the next state and judging whether the next state reaches a termination state or not; if the ending state is reached, processing the satellite communication channel data information by utilizing the rewarding function to obtain a rewarding value; if the state of termination is not reached, the current state and the current action are processed by using the evaluation neural network, and a reward value is obtained;
s246, storing the expression { current state, next state, current action, rewarding value } into an experience pool;
s247, judging whether a network needs to be updated, if so, updating the strategy neural network and the evaluation neural network to obtain an updated strategy neural network and an updated evaluation neural network;
s248, executing S242-S247 based on the updating strategy neural network and the updating evaluation neural network;
and S249, stopping training when the reward reaches 1 or the preset maximum training times, and obtaining a symbolic regression model based on the transducer.
6. The method for modeling a satellite communication channel according to claim 5, wherein updating the policy neural network and the evaluation neural network to obtain an updated policy neural network and an updated evaluation neural network comprises:
s2471, when the training times reach a preset threshold value, sequencing the expressions in the experience pool according to the rewarding value of the termination node, and processing the expressions of which the number is 10% before sequencing to obtain the estimated network loss;
the evaluation network loss is:
critic_loss＝MSE(f c (state)-critic_rewards)
wherein, critic_loss is the evaluation network loss, fc (·) is the evaluation neural network, state is the state, next_state is the next state, r is the prize value, critic_rewards=r+γ×f c (next_state), MSE () is a mean square error function, and γ is a discount coefficient in deep reinforcement learning;
s2472, back-propagating the evaluation network loss to obtain an updated evaluation neural network;
s2473, back-propagating a preset strategy network loss function to obtain an updated strategy neural network;
the preset strategy network loss function is as follows:
wherein, actor_loss is a strategic network loss function, E # - []The table design finds mathematical expectations, min () represents the minimum,representing the current policy neural network f a (. About.) probability of getting action a in state,/>Representing a previous strategic neural network f a_old () Obtaining the probability of action a under the state, wherein A is an advantage function, and clip (·) is a cut-off function.
7. The method of claim 1, wherein the transform symbol regression system parameters include a learning rate, a batch_size, a discount coefficient, a number of iterations m, and an exit condition;
the exit condition is that the iteration number is greater than m or the reward reaches 1.
8. A satellite communications channel modeling apparatus, the apparatus comprising:
the data acquisition module is used for acquiring satellite communication channel data information;
the satellite communication channel data information comprises the distance between receiving ends, signal power and signal frequency;
the symbolic regression model construction module is used for constructing a symbolic regression model based on a transducer;
the symbol regression system parameter acquisition module is used for processing the satellite communication channel data information to obtain a transducer symbol regression system parameter;
and the satellite communication channel modeling module is used for processing the parameters of the transform symbol regression system according to the transform-based symbol regression model to obtain a satellite communication channel mathematical expression model and realize satellite communication channel modeling.
9. A satellite communications channel modeling apparatus, the apparatus comprising:
a memory storing executable program code;
a processor coupled to the memory;
the processor invokes the executable program code stored in the memory to perform the satellite communication channel modeling method of any of claims 1-7.
10. A computer-storable medium storing computer instructions that, when invoked, are adapted to perform the satellite communication channel modeling method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311696480.5A CN117692083A (en) | 2023-12-11 | 2023-12-11 | Satellite communication channel modeling method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311696480.5A CN117692083A (en) | 2023-12-11 | 2023-12-11 | Satellite communication channel modeling method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117692083A true CN117692083A (en) | 2024-03-12 |
Family
ID=90127904
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311696480.5A Pending CN117692083A (en) | 2023-12-11 | 2023-12-11 | Satellite communication channel modeling method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117692083A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10756809B1 (en) * | 2018-11-21 | 2020-08-25 | Beijing Yuritan Technology Co.Ltd | Emergency communication satellite terminal management system |
| CN115642970A (en) * | 2022-09-16 | 2023-01-24 | 北京交通大学 | Self-learning channel modeling method and system |
| CN115882985A (en) * | 2022-11-28 | 2023-03-31 | 西安交通大学 | Low-orbit satellite channel prediction method and system based on Gaussian process regression |
| CN116318371A (en) * | 2023-04-10 | 2023-06-23 | 哈尔滨工业大学（深圳） | Communication resource allocation method and device for satellite Internet and readable storage medium |
-
2023
- 2023-12-11 CN CN202311696480.5A patent/CN117692083A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10756809B1 (en) * | 2018-11-21 | 2020-08-25 | Beijing Yuritan Technology Co.Ltd | Emergency communication satellite terminal management system |
| CN115642970A (en) * | 2022-09-16 | 2023-01-24 | 北京交通大学 | Self-learning channel modeling method and system |
| CN115882985A (en) * | 2022-11-28 | 2023-03-31 | 西安交通大学 | Low-orbit satellite channel prediction method and system based on Gaussian process regression |
| CN116318371A (en) * | 2023-04-10 | 2023-06-23 | 哈尔滨工业大学（深圳） | Communication resource allocation method and device for satellite Internet and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210133536A1 (en) | Load prediction method and apparatus based on neural network | |
| CN111027732B (en) | Method and system for generating multi-wind power plant output scene | |
| CN111737640B (en) | Water level prediction method, device and computer readable storage medium | |
| CN111461463A (en) | Short-term load prediction method, system and equipment based on TCN-BP | |
| CN112131888A (en) | Method, device and equipment for analyzing semantic emotion and storage medium | |
| CN114694379B (en) | Traffic flow prediction method and system based on self-adaptive dynamic graph convolution | |
| CN113360848A (en) | Time sequence data prediction method and device | |
| Rodríguez-López | Periodic boundary value problems for impulsive fuzzy differential equations | |
| CN116311880A (en) | Traffic flow prediction method and equipment based on local-global space-time feature fusion | |
| CN113194493B (en) | Wireless network data missing attribute recovery method and device based on graph neural network | |
| CN103607219B (en) | A kind of noise prediction method of electric line communication system | |
| CN117692083A (en) | Satellite communication channel modeling method and device | |
| US20230401363A1 (en) | GaN Distributed RF Power Amplifier Automation Design with Deep Reinforcement Learning | |
| CN115630566B (en) | Data assimilation method and system based on deep learning and dynamic constraint | |
| CN115619449A (en) | Deep learning model-based fertilizer price prediction method, device and storage medium | |
| CN115829717A (en) | Wind control decision rule optimization method, system, terminal and storage medium | |
| CN111882124A (en) | Homogeneous platform development effect prediction method based on generation confrontation simulation learning | |
| CN113705614B (en) | GAN-based complex industrial process operation index correction method | |
| CN117750436B (en) | Security service migration method and system in mobile edge computing scene | |
| CN114298445B (en) | Site quantitative precipitation forecast method and system based on graph convolution neural network | |
| CN112884129B (en) | Multi-step rule extraction method, device and storage medium based on teaching data | |
| Armenio et al. | Scenario optimization for optimal training of Echo State Networks | |
| Wu et al. | Efficient solutions of interactive dynamic influence diagrams using model identification | |
| CN118036825A (en) | Power load prediction method and system based on multi-scale feature fusion | |
| CN117933402A (en) | Multi-hop reasoning method and system for power grid knowledge graph based on GNN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |