Time-varying channel estimation method based on Kalman filtering

Abstract

The invention belongs to the technical field of communication, particularly relates to a time-varying channel estimation method based on Kalman filtering, and relates to the field of wireless communication, in a wireless communication system, a quasi-static channel is converted into a dynamic time-varying channel along with the movement process of a user, and when an IRS deployed in the communication system is equipped with a large number of passive reflection elements, the time-varying channel can be estimated. Therefore, the channel estimation becomes complicated, and particularly in the channel estimation of a mobile user, the time-varying channel estimation becomes particularly complicated along with the continuous movement of the user and the continuous change of the state information of a signal. Therefore, on the basis of a state model of a time-varying channel, an improved algorithm based on Kalman filtering is introduced to predict information of a next moment according to state information of a mobile user at a previous moment, a time-varying coefficient is obtained, signal sparse features are calculated through a compressed sensing algorithm, subspace iteration tracking is adopted, and input signal sparse atoms are searched. And finally, reconstructing an initial signal by using the threshold optimization atomic characteristics to obtain a high-precision time-varying channel estimation result.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and particularly to a time-varying channel estimation method based on Kalman filtering. Background Art

[0004] The rapid development of fields such as information technology, sensor technology, Internet of Things, and artificial intelligence provides a technical basis for the realization of intelligent trash cans. The miniaturization and cost reduction of sensors make it possible for intelligent trash cans to monitor and manage garbage in real time. For example, information such as the filling level and type of garbage in the trash can can be detected through sensors.

[0005] City managers are faced with increasingly complex urban garbage management tasks and need more effective tools and methods to deal with them. Intelligent trash cans can help city managers better manage urban garbage, improve garbage disposal efficiency, and reduce management costs by providing functions such as real-time monitoring and data analysis.

[0006] With the development of society, people's awareness of environmental protection is constantly increasing, and the requirements for garbage disposal are also getting higher and higher. Intelligent trash cans can better promote garbage classification and resource utilization, meet the needs of contemporary society for environmental protection and sustainable development, and help popularize the culture of garbage classification.

[0007] Foreign research: Foreign research on intelligent sorting trash cans is relatively advanced. Demonstration projects are carried out in Europe to test the technical effects; American companies develop commercial solutions to achieve various functions; Canadian research institutions improve sorting accuracy and efficiency; Australian communities promote the development of garbage classification; intelligent trash cans in Asian regions such as Japan are equipped with voice prompt functions.

[0008] Domestic research: With the improvement of environmental protection awareness and policy support in China, the research and application of intelligent garbage sorting systems have gradually received attention and are expected to be comprehensively improved in aspects such as technological innovation, system integration, data-driven, application popularization, and environmental protection in the future.

[0009] There are a large number of intelligent sorting trash cans. Traditional communication technologies are easily interfered with, leading to data transmission problems. To ensure the efficient transmission of data by trash cans and achieve intelligent sorting, monitoring, and management, it is necessary to introduce more advanced wireless transmission anti-interference technologies to improve data transmission efficiency and reliability to meet the communication requirements of the intelligent trash can system.

[0010] In a wireless communication system, such as when intelligent reflecting surface assisted wireless communication generates cascaded channels, and the movement of users makes the quasi-static channel become a dynamic time-varying channel, channel estimation becomes complex. Existing technologies have problems such as low transmission rate and poor estimation accuracy in the time-varying channel estimation during the movement of users. Therefore, new methods are needed to improve the accuracy and performance of time-varying channel estimation.

[0011] Since cascaded channels are generated during intelligent reflecting surface assisted wireless communication, and there is a certain correlation between cascaded channels, the original channel estimation scheme is no longer applicable to the cascaded channel estimation in this system. Therefore, it is necessary to design a channel state information (CSI) acquisition algorithm with low pilot overhead and low complexity. By using the correlation between cascaded channels to transform the channel estimation problem into a sparse signal recovery problem, the pilot overhead is reduced and the estimation accuracy is improved, thereby further improving the estimation performance of cascaded channels. Therefore, we propose a method that transforms the channel estimation of cascaded channels into the recovery of sparse signals to solve the above problems.

[0012] For the cascaded channel estimation problem, existing technologies involve multiple fields such as binary optimization, parallel factor decomposition, and compressive sensing. However, during the movement of users, the estimation of time-varying channels becomes particularly complex. As the signal state information changes continuously, there are disadvantages such as low transmission rate and poor estimation accuracy when performing channel estimation. Therefore, it is proposed that before signal reconstruction, the state of the time-varying channel is predicted first, that is, the time-varying coefficients of the time-varying channel are predicted through a probability model framework to accurately obtain the correlation between channels and the prior information of mobile users. Finally, based on the predicted signal parameters obtained by compressive sensing technology, the sparse signal is recovered and reconstructed to complete the channel estimation of the time-varying channel. Summary of the Invention

[0013] (I) Technical problems to be solved

[0014] Aiming at the deficiencies of the existing technology, the present invention provides a time-varying channel estimation method based on Kalman filtering, which solves the problems raised in the above background technology.

[0015] (II) Technical solutions

[0016] The present invention specifically adopts the following technical solutions to achieve the above object:

[0017] A time-varying channel estimation method based on Kalman filtering includes the following steps,

[0018] Step 1: System modeling, obtaining the user state in the area, and constructing a channel model of the time-varying channel;

[0019] Step 2: Based on the time-varying channel model, establish a variational Bayesian Kalman filtering algorithm. Introduce a probability model in the time-varying channel to predict the state of the time-varying channel, and gradually approximate the true value through continuous iteration to obtain the time-varying coefficient;

[0020] Step 3: After the prediction in Step 2 is completed, use the compressive sensing algorithm to calculate the sparse characteristics of the signal from the parameters of the obtained time-varying channel and implement subspace iterative tracking, search for the most appropriate atoms to reconstruct the initial signal, and complete the channel estimation;

[0021] Step 4: After the channel estimation in Step 3 is completed, judge the effect of the algorithm based on the channel estimation performance results, by calculating the computational complexity, normalized mean square error, and pilot overhead.

[0022] Furthermore, the specific method of Step 1 is: Obtain the user status in the area and construct a channel model for the time-varying channel.

[0023] Furthermore, the specific implementation method of Step 2 is: When establishing the variational Bayesian Kalman filtering algorithm, introduce a probability model in the time-varying channel to predict the state distribution of the time-varying channel, including:

[0024] Determine the representation method of the time-varying channel, determine the state vector, and initialize the parameter distribution;

[0025] Through variational inference, calculate the evidence lower bound, construct an optimizable objective function, maximize the evidence lower bound, update the variational distribution parameters, continue to iterate the process of recalculating the evidence lower bound and variational distribution parameters, and set a threshold so that the output result meets the convergence condition.

[0026] Based on continuous iteration, make the variational distribution gradually approximate the true posterior distribution.

[0027] Furthermore, the specific implementation method of Step 2 is: When establishing the variational Bayesian Kalman filtering algorithm, calculate the time-varying coefficient of the channel through Kalman filtering after obtaining the probability distribution prediction, including:

[0028] Based on the probability model, perform state prediction and calculate the covariance matrix of the predicted state;

[0029] Combine the observed data, approximate the posterior probability distribution in the variational Bayesian framework and calculate the Kalman gain; use the observed data to correct the predicted state to obtain a new state estimate value, making it closer to the true state of the time-varying channel;

[0030] Through probability model prediction, continuously iterate and update the data to correct the state estimate value of the time-varying channel, and gradually approximate the true time-varying coefficient;

[0031] Further, the specific implementation of Step 3 is as follows: After the time-varying coefficients are predicted in Step 2, the sparse characteristics of the signal are calculated for the parameters of the obtained time-varying channel through the compressive sensing algorithm, and the most appropriate atoms are searched by subspace iterative pursuit to reconstruct the initial signal, including:

[0032] Performing subspace iterative pursuit through a step size factor;

[0033] Selecting atoms that meet the constraints and reconstructing the initial signal according to the atom characteristics;

[0034] Further, the specific implementation of Step 4 is as follows: According to the channel estimation performance results, the effectiveness of the algorithm is judged by the computational complexity, normalized mean square error, and pilot overhead.

[0035] (III) Beneficial Effects

[0036] Compared with the prior art, the present invention provides a time-varying channel estimation method based on Kalman filtering, which has the following beneficial effects:

[0037] Compared with the traditional channel estimation methods for time-varying channels, the time-varying channel estimation method based on Kalman filtering of the present invention includes: According to the characteristics of the time-varying channel of mobile users, before channel estimation, the prior knowledge and observation data are fused through a probability model to infer the relationship between the observation data and the unknown parameters. After multiple iterations, the data gradually approaches the true value, accurately predicting the time-varying coefficients and reducing the estimation error; Finally, the sparse characteristics of the signal are calculated through the compressive sensing algorithm, subspace iterative pursuit is adopted to search for the sparse atoms of the input signal, and finally the atom characteristics are optimized by a threshold to reconstruct the initial signal, obtaining a high-precision time-varying channel estimation result. It reduces the computational complexity, pilot overhead, and normalized mean square error of the channel estimation of mobile users, improves the reconstruction accuracy and efficiency, and improves the estimation accuracy of the time-varying channel. Brief Description of the Drawings

[0038] Figure 1 It is a schematic flowchart of the time-varying channel estimation method based on Kalman filtering of the present invention;

[0039] Figure 2 It is a schematic flowchart of the time-varying coefficient prediction based on variational Bayesian Kalman filtering of the present invention;

[0040] Figure 3 It is a schematic flowchart of the compressive sensing reconstruction algorithm of the present invention. Detailed Embodiments

[0041] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0042] Embodiment

[0043] As Figure 1 shown, a time-varying channel estimation method based on Kalman filtering proposed in an embodiment of the present invention is as follows:

[0044] Step 1, system modeling, obtaining the user state in the area, and constructing a channel model of the time-varying channel. The specific method for building the channel model in Step 1 is:

[0045] The user state considered in the present invention is the moving state, and a time-varying channel model is constructed. Assuming that the transmitted signal is , and the received signal after passing through the time-varying channel is , based on the channel parameter characteristics, multipath propagation effect, and time-varying factors, the time-varying channel model can be expressed as:

[0046]

[0047] where is the noise added to the received signal, used to simulate the inevitable interference factors in the actual environment.

[0048] Step 2, establishing a variational Bayesian Kalman filtering algorithm based on the above time-varying channel model, introducing a probability model in the time-varying channel to predict the state of the time-varying channel, and gradually approaching the true value through continuous iteration. Referring to Figure 2 , the specific implementation method of Step 2 is:

[0049] Determine the representation method of the time-varying channel, determine the state vector, and initialize the parameter distribution;

[0050] Through variational inference, calculate the evidence lower bound, construct an optimizable objective function, maximize the evidence lower bound, update the variational distribution parameters, continue to iterate the process of recalculating the evidence lower bound and variational distribution parameters, and set a threshold so that the output result meets the convergence condition.

[0051] Based on continuous iteration, the variational distribution gradually approaches the true posterior distribution.

[0052] Step 2, based on the established variational Bayesian Kalman filtering algorithm, calculate the time-varying coefficient of the channel through Kalman filtering after obtaining the probability distribution prediction. The specific implementation method of Step 2 is:

[0053] Based on a probability model, perform state prediction and calculate the covariance matrix of the predicted state;

[0054] Combine the observation data, approximate the posterior probability distribution under the variational Bayesian framework and calculate the Kalman gain; use the observation data to correct the predicted state to obtain a new state estimate, making it closer to the true time-varying channel state.

[0055] Make predictions through the probability model, continuously iterate and update the data to correct the state estimate of the time-varying channel, and gradually approach the true time-varying coefficient.

[0056] Based on the determined parameters of the variational distribution, update in a way that is more in line with the probability distribution. According to the probability distribution of the predicted state and the probability distribution of the new observation, recalculate the posterior distribution of the system state through variational inference, so as to obtain the updated state estimate and the corresponding probability distribution, more accurately reflecting the current actual state of the system. At the same time, according to the noise characteristics of the system and the state transition model, calculate the covariance of the predicted state, which reflects the uncertainty degree of the predicted state, facilitating subsequent accurate update in combination with new observations.

[0057] After obtaining a new measurement value in the measurement update stage, combine the predicted state and covariance with the new observation value, recalculate the posterior distribution of the system state through variational inference, and calculate the difference and covariance matrix between the measurement prediction value and the actual prediction value.

[0058] In the iterative loop stage, as new observation values are continuously obtained, the system state continuously changes. Based on the iterative operations of prediction update and measurement update, continuously calculate the Kalman gain, state estimate, and covariance matrix according to the new observation values, continuously optimize the estimation of the system state, accurately predict and track the dynamic changes of the system, so as to achieve accurate estimation of the channel state.

[0059] Step 3, after the prediction in Step 2 is completed, calculate the sparse features of the signal and implement subspace iterative tracking for the parameters of the obtained time-varying channel through the compressive sensing algorithm, search for the most appropriate atoms to reconstruct the initial signal, and complete the channel estimation. Refer to Figure 3 , the specific implementation method of Step 3 is:

[0060] Perform subspace iterative tracking through the step size factor;

[0061] Select atoms that meet the constraints and complete the reconstruction of the initial signal according to the atom characteristics;

[0062] Step 4, after the signal reconstruction in Step 3 is completed, the specific implementation method is: According to the channel estimation performance results, judge the effect of the algorithm by calculating the computational complexity, normalized mean square error, and pilot overhead.

[0063] In summary, aiming at the problems of inaccurate channel estimation and high computational complexity in time-varying channels, the present invention uses an improved Kalman filtering algorithm to reduce the error between the true value and the measured value through a probability model, complete the prediction of time-varying coefficients and parameter estimation problems in time-varying channels, calculate the sparse characteristics of the signal through a compressive sensing algorithm, adopt subspace iterative pursuit to search for the sparse atoms of the input signal, and finally use threshold optimization to reconstruct the initial signal of the atomic characteristics, so as to obtain a more accurate and faster high-precision time-varying channel estimation result to complete the channel estimation of the cascaded channel.

[0064] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A time-varying channel estimation method based on Kalman filtering, characterized in that: It includes the following steps: Step 1: System modeling, obtaining the user status in the area, and constructing a channel model for the time-varying channel; Step 2: Establishing a variational Bayesian Kalman filtering algorithm based on the time-varying channel model, introducing a probability model in the time-varying channel to predict the state of the time-varying channel, and gradually approaching the true value through continuous iteration to obtain the time-varying coefficient; Step 3: Calculating the sparse features of the signal through the compressed sensing algorithm for the parameters of the obtained time-varying channel and realizing band tracking, searching for appropriate atoms to reconstruct the initial signal, and completing channel estimation; Step 4: According to the channel estimation performance results, judging the effect of the algorithm by calculating the computational complexity, normalized mean square error, and pilot overhead.

2. The time-varying channel estimation method based on Kalman filtering according to claim 1, wherein: The user status is the moving state, and a time-varying channel model is constructed.

3. A time-varying channel estimation method based on Kalman filtering according to claim 1, characterized in that: The establishment of the variational Bayesian Kalman filtering algorithm, introducing a probability model in the time-varying channel to predict the state distribution of the time-varying channel, includes: Determining the representation method of the time-varying channel, determining the state vector, and initializing the parameter distribution; Through variational inference, calculating the evidence lower bound, constructing an optimizable objective function, maximizing the evidence lower bound, updating the variational distribution parameters, continuing to iterate the process of recalculating the evidence lower bound and variational distribution parameters, and setting a threshold so that the output result meets the convergence condition; Based on continuous iteration, the variational distribution gradually approaches the true posterior distribution.

4. A time-varying channel estimation method based on Kalman filtering according to claim 3, characterized in that: The establishment of the variational Bayesian Kalman filtering algorithm, calculating the time-varying coefficient of the channel through Kalman filtering after obtaining the probability distribution prediction: includes: Based on the probability model, performing state prediction and calculating the covariance matrix of the predicted state; Combining the observation data, approximately processing the posterior probability distribution in the variational Bayesian framework and calculating the Kalman gain; using the observation data to correct the predicted state to obtain a new state estimate value, making it closer to the true time-varying channel state, and obtaining a time-varying coefficient close to the true value; Performing prediction through the probability model, continuously iterating and updating the data to correct the state estimate value of the time-varying channel, and gradually approaching the true time-varying coefficient.

5. A time-varying channel estimation method based on Kalman filtering according to claim 1, characterized in that: After predicting the time-varying coefficient in Step 3, calculating the sparse features of the signal through the compressed sensing algorithm for the parameters of the obtained time-varying channel, and through subspace iterative pursuit, searching for appropriate atoms to reconstruct the initial signal: includes: Performing subspace iterative pursuit through the step size factor; Selecting the atoms that meet the constraints and completing the reconstruction of the initial signal according to the atom features.

Images