A method and system for joint estimation of semantics and channel in low-altitude MIMO systems

Abstract

The application discloses a kind of joint estimation method and system of semantics and channel in low-altitude MIMO system, belong to low-altitude communication guarantee field, including: obtaining the initial channel state information of each unmanned aerial vehicle and ground control station between uplink;The semantic information matrix uploaded by each unmanned aerial vehicle is decoded, and the semantic information is obtained, and it is mapped as the "space-state" semantic prior of each unmanned aerial vehicle;The received pilot signal is compensated for Doppler phase;Construct semantic constraint index set, carry out channel sparse reconstruction to pilot signal after Doppler compensation, update the channel matrix estimation value between each unmanned aerial vehicle and ground control station between uplink;Judge whether the iteration termination condition is satisfied, otherwise, with the updated channel matrix estimation value return to continue iteration.The application significantly improves the precision of channel estimation and the reliability of semantic decoding under complex low-altitude environment, enhances the robustness of algorithm in high dynamic scene.

Description

Technical Field

[0001] This invention belongs to the field of low-altitude communication support, specifically relating to a joint estimation method and system for semantics and channel in a low-altitude MIMO system, which is particularly suitable for the joint iterative enhancement of semantic decoding and channel state information in low-altitude unmanned aerial vehicle (UAV) communication scenarios. Background Technology

[0002] With the booming development of the low-altitude economy, drones are increasingly being used in logistics delivery, power line inspection, environmental monitoring, emergency communications, and disaster relief. Meanwhile, massive MIMO technology, with its superior spatial diversity and multiplexing capabilities, has become a key enabling technology for high-reliability communication of low-altitude drones. However, the high-speed maneuvering of drones leads to significant Doppler shift, and the complex propagation environment at low altitudes (such as multipath reflections caused by urban buildings and high-voltage power towers) results in fast fading and sparse channel characteristics. These non-ideal channel factors severely restrict the reliability of information transmission under traditional communication systems. Semantic communication, as an emerging communication architecture, transmits the "meaning" of information rather than raw bits, which can greatly reduce bandwidth overhead and improve transmission efficiency. In drone inspection and emergency monitoring tasks, data collected by sensors such as optical cameras, infrared thermal imagers, and lidar on the drone are semantically encoded to generate semantic feature information for transmission, offering a significant bandwidth advantage compared to traditional raw data transmission methods. However, semantic feature information is highly susceptible to non-ideal channel factors such as Doppler shift and multipath fading during wireless channel transmission, leading to distortion or even failure in decoding at ground control stations. Existing channel estimation methods (such as least squares estimation and minimum mean square error estimation) typically rely on significant pilot overhead, making it difficult to balance estimation accuracy and resource efficiency in the high-dynamic scenarios of low-altitude UAVs. While compressed sensing methods can leverage channel sparsity to reduce pilot overhead, their search space is enormous, resulting in high computational complexity. Furthermore, most existing semantic decoding methods assume known and ideal channel state information, lacking robustness to channel estimation errors and failing to fully utilize the motion state and spatial location information acquired by the UAV's own sensing capabilities to assist in the joint optimization of channel estimation and semantic decoding. Therefore, the industry urgently needs a method that can use the sensing information of low-altitude UAVs as prior knowledge, integrating semantic decoding and channel estimation for joint iterative enhancement, to achieve low-overhead, high-precision, and high-reliability semantic information transmission in complex low-altitude link environments. This would effectively reduce the communication costs of large-scale low-altitude economic applications, improve the timeliness and reliability of critical low-altitude infrastructure applications, and support the high-quality development of the low-altitude economy. Summary of the Invention

[0003] To address the aforementioned problems, this invention proposes a joint estimation method and system for semantics and channel in low-altitude MIMO systems.

[0004] The technical solution of this invention is:

[0005] A joint estimation method for semantics and channel in a low-altitude MIMO system includes the following steps:

[0006] S1. Based on the pilot signals uploaded by each UAV received by the ground control station, perform initial channel estimation on the channel matrix of the uplink between each UAV and the ground control station to obtain initial channel state information.

[0007] S2. Based on the initial channel state information, decode the semantic information matrix uploaded by each UAV received by the ground control station to obtain the semantic information of each UAV; perform an inverse mapping operation on the semantic information to extract the flight speed, horizontal angle of arrival range and pitch angle of arrival range of each UAV to form a "space-state" semantic prior.

[0008] S3. Based on the flight speed in the "space-state" semantic prior, calculate the maximum Doppler frequency offset of the uplink between each UAV and the ground control station, construct the Doppler compensation matrix, and perform Doppler phase compensation on the received pilot signal.

[0009] S4. Based on the horizontal angle of arrival range and pitch angle of arrival range in the "space-state" semantic prior, determine the horizontal angle of arrival index and pitch angle of arrival index respectively, and construct a semantic constraint index set; within the search space defined by the semantic constraint index set, use the matching pursuit algorithm to perform channel sparse reconstruction on the Doppler-compensated pilot signal, and update the estimated value of the channel matrix of the uplink between each UAV and the ground control station.

[0010] S5. Determine whether the iteration termination condition is met. If it is met, output the final decoded semantic information and the final channel matrix estimate; otherwise, return to step S2 with the updated channel matrix estimate to continue the iteration.

[0011] Furthermore, the "space-state" semantic prior described in step S2 includes the first The drone obtained in the next iteration flight speed drones Horizontal angle of arrival range of the resolvable path between the ground control station and the ground control station and horizontal central angle Pitch reach range and pitch center angle .

[0012] Furthermore, the maximum Doppler frequency offset mentioned in step S3 according to Calculate, where, The carrier wavelength is used; a frequency domain Doppler compensation matrix is ​​constructed based on the maximum Doppler frequency offset. For the received pilot matrix Compensation is performed, and the compensated pilot matrix is ​​as follows: The frequency domain Doppler compensation matrix Diagonal matrix of time-domain Doppler deflection The time-domain Doppler deflection diagonal matrix, obtained through discrete Fourier transform, is as follows: ;in, Indicates the subcarrier spacing. For the number of subcarriers, For subcarrier index, For the first The maximum Doppler frequency shift obtained in the next iteration This represents the diagonalization operation that forms a diagonal matrix.

[0013] Furthermore, the semantic constraint index set mentioned in step S4 Build it as follows: ;in, Indicates the corresponding horizontal angle of arrival and pitch angle of arrival Index in the angle field dictionary The range of horizontal arrival angles. Range of pitch angles;

[0014] The search space for each iteration of the matching pursuit algorithm is limited to... Within, the optimal index value Select as follows: ;in, For the first The observation residual vector of the second matching search For the first angle field dictionary List, Represents finding vectors and dot product operation, Expressing the request The index at which the inner function reaches its maximum value. This indicates the absolute value operation.

[0015] Furthermore, the iteration termination condition in step S5 is that the iteration terminates when any of the following conditions are met:

[0016] S5.1 Normalized difference between the channel matrix estimates of the current iteration and the previous iteration Less than the preset iteration termination threshold ;

[0017] The normalized differences are as follows Calculate, where, Represents finding a matrix The Frobenius norm, and They represent the first Second and third The estimated channel matrix obtained in the next iteration;

[0018] S5.2 The current iteration count has reached the preset maximum iteration count. .

[0019] The iteration termination threshold according to Settings, in which The preset constant false alarm rate, Represents the Gaussian error function. For noise variance, This refers to the number of antennas at the ground station. For the first The estimated channel matrix obtained in the next iteration;

[0020] The maximum number of iterations Maximum processing latency allowed by the system Constraints, satisfaction ;

[0021] in To initialize processing delay, This represents the processing delay for a single iteration.

[0022] Furthermore, the inner loop stopping criterion of the matching pursuit algorithm is: the number of currently estimated paths reaches the preset maximum number of paths for low-altitude UAV communication links. Or the energy of the current observed residual is less than the preset stop decision threshold. ;

[0023] The stop judgment threshold according to set up, As a preset penalty factor, This represents the noise variance.

[0024] This invention also provides a joint semantic and channel estimation system for low-altitude MIMO systems, comprising:

[0025] The initial channel estimation module is used to perform initial channel estimation on the uplink channel matrix between each UAV and the ground control station based on the pilot signals uploaded by each UAV received by the ground control station, and to obtain initial channel state information.

[0026] The semantic decoding and prior extraction module is used to decode the semantic information matrix uploaded by each UAV received by the ground control station based on the initial channel state information or the updated channel matrix estimate, to obtain the semantic information of each UAV, and to perform an inverse mapping operation on the semantic information to extract the flight speed, horizontal angle of arrival range and pitch angle of arrival range of each UAV, forming a "space-state" semantic prior.

[0027] The Doppler compensation module is used to calculate the maximum Doppler frequency offset and construct the Doppler compensation matrix based on the flight speed in the "space-state" semantic prior, and to perform Doppler phase compensation on the received pilot signal.

[0028] The semantic constraint channel reconstruction module is used to construct a semantic constraint index set based on the horizontal angle of arrival range and the pitch angle of arrival range in the "space-state" semantic prior, and to perform sparse channel reconstruction using the matching pursuit algorithm within the search space defined by the semantic constraint index set, and update the channel matrix estimate.

[0029] The iteration control module is used to determine whether the iteration termination condition is met. If it is not met, the semantic decoding and prior extraction module, Doppler compensation module and semantic constraint channel reconstruction module are controlled to continue iterating.

[0030] This invention also provides an emergency communication support system: employing the joint semantic and channel estimation method in a low-altitude MIMO system, or the aforementioned system; comprising: a ground control station deployed at an emergency command center or mobile command platform, and unmanned aerial vehicles (UAVs) for performing emergency missions, such as aerial patrol UAVs or communication relay UAVs; in the event of a disruption or unavailability of conventional ground communication links, semantic information is transmitted through the uplink between each UAV and the ground control station, and reliable communication between the emergency command center and the UAVs is enhanced by joint iteration of semantic decoding and channel estimation.

[0031] Furthermore, the semantic information includes semantic feature information generated by semantically encoding disaster site environmental perception data collected by optical cameras, infrared thermal imagers, or lidar sensors mounted on the UAV; the flight speed, horizontal angle of arrival range, and pitch angle of arrival range in the "space-state" semantic prior are obtained by fusing the UAV's onboard global positioning system, inertial measurement unit, and visual odometry and then embedded into the semantic information and uploaded.

[0032] The beneficial effects of this invention are:

[0033] (1) This invention extracts the perception prior information such as flight speed, horizontal angle of arrival range, and pitch angle of arrival range embedded in the semantic information of the UAV into "space-state" semantic prior, and drives Doppler compensation and channel sparse reconstruction, realizing deep coupling and joint iterative enhancement of semantic decoding and channel estimation, which significantly improves the accuracy of channel estimation and the reliability of semantic decoding in complex low-altitude environments.

[0034] (2) This invention utilizes the horizontal angle of arrival range and the pitch angle of arrival range in the semantic prior to construct a semantic constraint index set, which greatly reduces the search space of the matching pursuit algorithm from the global angle domain dictionary to the local range of the semantic prior constraints. While ensuring the channel reconstruction accuracy, it significantly reduces the computational complexity and is suitable for low-altitude UAV communication scenarios with limited resources.

[0035] (3) This invention uses the flight speed information in the semantic prior to calculate the maximum Doppler frequency offset and construct the Doppler compensation matrix, which effectively eliminates the interference of Doppler phase rotation caused by the high-speed maneuver of the UAV on the channel estimation and enhances the robustness of the algorithm in high dynamic scenarios.

[0036] (4) This invention establishes clear iteration termination conditions (normalized difference threshold and maximum number of iterations as dual criteria) and inner loop stopping criteria, which take into account the balance between estimation accuracy and processing delay. The number of iterations can be adaptively adjusted according to the maximum processing delay allowed by the system, which has good engineering practicality.

[0037] (5) This invention is particularly suitable for emergency communication support scenarios. When the conventional ground communication link is interrupted or unavailable, semantic information is transmitted through the UAV uplink and a joint iterative enhancement mechanism is used to effectively ensure reliable communication between the emergency command center and the UAV, providing strong support for disaster relief, emergency patrol and other tasks. Attached Figure Description

[0038] Figure 1 This is a flowchart of a joint estimation method for semantics and channel in a low-altitude MIMO system.

[0039] Figure 2 This is a schematic diagram of a low-altitude MIMO system. Detailed Implementation

[0040] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0041] like Figure 1 As shown, this invention provides a joint estimation method for semantics and channel in a low-altitude MIMO system, comprising the following steps:

[0042] S1. Based on the pilot signals uploaded by each UAV received by the ground control station, perform initial channel estimation on the uplink channel matrix between each UAV and the ground control station to obtain initial channel state information.

[0043] In this step, the ground control station is equipped with a massive MIMO antenna array. Each UAV first transmits a known pilot sequence before transmitting data. The ground control station receiver performs correlation operations between the received pilot signals and the locally stored pilot sequences. Classical channel estimation algorithms such as least squares (LS) or minimum mean square error (MMSE) can be used to obtain the initial channel matrix estimate of the uplink between each UAV and the ground control station. This initial estimate serves as the initial value for subsequent joint iterations.

[0044] S2. Based on the initial channel state information, decode the semantic information matrix uploaded by each UAV received by the ground control station to obtain the semantic information of each UAV; perform an inverse mapping operation on the semantic information to extract the flight speed, horizontal angle of arrival range and pitch angle of arrival range of each UAV to form a "space-state" semantic prior.

[0045] In this step, the ground control station uses the initial channel matrix estimate obtained in step S1 to perform channel equalization and decoding on the semantic information matrices uploaded by each UAV, recovering the semantic information of each UAV. The semantic information is generated by the UAV end through semantic feature extraction and compression encoding of data collected by local sensors (such as optical cameras, infrared thermal imagers, and lidar). More importantly, during the semantic encoding process, the UAV embeds perception information such as flight speed, horizontal angle of arrival range, and pitch angle of arrival range obtained by fusing its onboard Global Positioning System (GPS), Inertial Measurement Unit (IMU), and Visual Odometry into the semantic information and uploads it together.

[0046] The ground control station performs an inverse mapping operation on the decoded semantic information to extract the first... The drone obtained in the next iteration flight speed drones Horizontal angle of arrival range of the resolvable path between the ground control station and the ground control station and horizontal central angle Pitch reach range and pitch center angle This forms a "space-state" semantic prior. This prior information integrates the UAV's motion state (velocity) and spatial orientation (angle range), providing crucial physical layer auxiliary information for subsequent Doppler compensation and channel sparse reconstruction.

[0047] S3. Based on the flight speed in the "space-state" semantic prior, calculate the maximum Doppler frequency offset of the uplink between each UAV and the ground control station, construct the Doppler compensation matrix, and perform Doppler phase compensation on the received pilot signal.

[0048] The high-speed maneuvers of UAVs cause significant carrier frequency shifts (Doppler shifts), severely impacting the accuracy of channel estimation. In this step, the flight speed is determined based on the "space-state" semantic prior. ,according to Calculate the maximum Doppler frequency offset ,in The carrier wavelength.

[0049] Based on the maximum Doppler frequency offset, a frequency domain Doppler compensation matrix is ​​constructed. Specifically, the frequency domain Doppler compensation matrix Diagonal matrix of time-domain Doppler deflection The time-domain Doppler deflection diagonal matrix, obtained through discrete Fourier transform, is as follows: ,in, Indicates the subcarrier spacing. For the number of subcarriers, For subcarrier index, For the first The maximum Doppler frequency shift obtained in the next iteration This represents the diagonalization operation that forms a diagonal matrix.

[0050] Using the constructed frequency domain Doppler compensation matrix For the received pilot matrix Compensation is performed, and the compensated pilot matrix is ​​as follows: The Doppler compensation operation described above effectively eliminates the phase rotation introduced by the high-speed movement of the UAV, providing high-quality observation input for subsequent channel sparse reconstruction.

[0051] S4. Based on the horizontal angle of arrival range and pitch angle of arrival range in the "space-state" semantic prior, determine the horizontal angle of arrival index and pitch angle of arrival index respectively, and construct a semantic constraint index set; within the search space defined by the semantic constraint index set, use the matching pursuit algorithm to perform channel sparse reconstruction on the Doppler-compensated pilot signal, and update the estimated value of the channel matrix of the uplink between each UAV and the ground control station.

[0052] The communication channels of low-altitude unmanned aerial vehicles (UAVs) exhibit significant sparsity in the angle domain, meaning that channel energy is concentrated on a limited number of resolvable paths. Traditional compressed sensing methods perform a global search across the entire angle domain dictionary, resulting in high computational complexity. This invention utilizes angle range information from the "space-state" semantic prior to significantly reduce the search space.

[0053] Specifically, semantic constraint index set Build it as follows: ;in, Indicates the corresponding horizontal angle of arrival and pitch angle of arrival Index in the angle field dictionary The range of horizontal arrival angles. Range of pitch angles.

[0054] Within the search space defined by the aforementioned semantic constraint index set, the matching pursuit algorithm is used to perform sparse channel reconstruction of the Doppler-compensated pilot signal. The search space for each iteration of the matching pursuit algorithm is limited to... Within, the optimal index value Select as follows: ;in, For the first The observation residual vector of the second matching search For the first angle field dictionary List, Represents finding vectors and dot product operation, Expressing the request The index at which the inner function reaches its maximum value. This indicates the absolute value operation.

[0055] The inner loop stopping criterion of the matching pursuit algorithm is: the number of currently estimated paths reaches the preset maximum number of paths for low-altitude UAV communication links. Or the energy of the current observed residual is less than the preset stop decision threshold. The aforementioned stop decision threshold according to set up, As a preset penalty factor, Let be the noise variance. Using the matching pursuit algorithm with the aforementioned semantic constraints, the updated channel matrix estimate is obtained.

[0056] S5. Determine whether the iteration termination condition is met. If it is met, output the final decoded semantic information and the final channel matrix estimate; otherwise, return to step S2 with the updated channel matrix estimate to continue the iteration.

[0057] The iteration terminates when any of the following conditions are met:

[0058] Condition 1: The normalized difference between the channel matrix estimates of the current iteration and the previous iteration. Less than the preset iteration termination threshold The normalized differences are as follows: Calculate, where, Represents finding a matrix The Frobenius norm, and They represent the first Second and third The estimated channel matrix obtained in the next iteration. The iteration termination threshold. according to Settings, in which The preset constant false alarm rate, Represents the Gaussian error function. For noise variance, This refers to the number of antennas at the ground station. For the first The estimated channel matrix obtained in the next iteration.

[0059] Condition 2: The current iteration count has reached the preset maximum iteration count. The maximum number of iterations. Maximum processing latency allowed by the system Constraints, satisfaction ;

[0060] in To initialize processing delay, This represents the processing delay for a single iteration.

[0061] Through the above iterative process, semantic decoding and channel estimation reinforce each other: more accurate channel estimation helps to decode semantic information more reliably, while the decoded semantic priors, in turn, constrain and optimize channel estimation, forming a positive feedback iterative enhancement mechanism.

[0062] like Figure 2 The diagram shows the structure of the joint semantic and channel estimation system in the low-altitude MIMO system of this invention. The system mainly consists of a ground control station and several low-altitude UAVs. The ground control station is equipped with a large-scale MIMO antenna array and a high-performance computing unit to execute the joint iterative enhancement algorithm. The UAVs have built-in local sensors (including GPS, IMU, visual sensors, etc.) and a semantic extraction and encoding unit, which encodes the perceived data and flight status information into semantic information before uploading it.

[0063] The system includes the following functional modules:

[0064] The initial channel estimation module is used to perform initial channel estimation on the uplink channel matrix between each UAV and the ground control station based on the pilot signals uploaded by each UAV received by the ground control station, and to obtain initial channel state information.

[0065] The semantic decoding and prior extraction module is used to decode the semantic information matrix uploaded by each UAV received by the ground control station based on the initial channel state information or the updated channel matrix estimate, to obtain the semantic information of each UAV, and to perform an inverse mapping operation on the semantic information to extract the flight speed, horizontal angle of arrival range and pitch angle of arrival range of each UAV, forming a "space-state" semantic prior.

[0066] The Doppler compensation module is used to calculate the maximum Doppler frequency offset and construct the Doppler compensation matrix based on the flight speed in the "space-state" semantic prior, and to perform Doppler phase compensation on the received pilot signal.

[0067] The semantic constraint channel reconstruction module is used to construct a semantic constraint index set based on the horizontal angle of arrival range and the pitch angle of arrival range in the "space-state" semantic prior, and to perform sparse channel reconstruction using the matching pursuit algorithm within the search space defined by the semantic constraint index set, thereby updating the channel matrix estimate.

[0068] The iteration control module is used to determine whether the iteration termination condition is met. If it is not met, the semantic decoding and prior extraction module, Doppler compensation module and semantic constraint channel reconstruction module are controlled to continue iterating.

[0069] Application Examples:

[0070] In an "emergency communication support" mission, a ground control station is deployed at the emergency command center, and multiple drones are conducting aerial reconnaissance missions over the disaster area. The first drone is operating over the disaster area, with its onboard optical camera and infrared thermal imager continuously collecting environmental perception data.

[0071] S1. The ground control station receives the pilot signal uploaded by the UAV through a large-scale MIMO antenna array, performs initial channel estimation, and obtains the initial channel matrix estimate.

[0072] S2. The ground control station decodes the received semantic information matrix based on the initial channel estimate. The semantic information of the UAV includes flight speed, horizontal angle of arrival range, and pitch angle of arrival range obtained through the fusion of onboard GPS, IMU, and visual odometry. The ground control station extracts this information through an inverse mapping operation to form a "space-state" semantic prior.

[0073] S3. At this time, the UAV is flying at a relatively high speed (e.g., 150 m / s), generating a significant Doppler frequency offset. The ground control station calculates the maximum Doppler frequency offset based on the flight speed in the semantic prior, constructs a Doppler compensation matrix, and performs phase compensation on the subsequently received pilot signals to eliminate the influence of the Doppler effect.

[0074] S4. The ground control station uses the horizontal and pitch angles of arrival ranges in the semantic prior to construct a semantic constraint index set, which greatly reduces the search space of the matching pursuit algorithm from the entire angle domain dictionary to the local range of the semantic prior constraints. Channel sparse reconstruction is performed in this reduced search space to obtain the updated channel matrix estimate.

[0075] S5. Determine the termination condition of the iteration: If the difference between the channel estimation of the current iteration and the previous iteration is less than the preset threshold, or the maximum number of iterations is reached, then terminate the iteration and output the final decoded semantic information (such as key semantics such as "building collapse area identification" and "trapped personnel thermal signal detection") and the final channel estimation value; otherwise, return to step S2 to continue iterative enhancement.

[0076] Through the above iterative process, even in complex low-altitude environments such as sudden strong winds and multipath interference, the system can filter out bit errors caused by channel distortion and accurately recover the core semantic information transmitted back by the UAV, ensuring reliable communication between the emergency command center and the UAV.

[0077] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A joint estimation method for semantics and channel in a low-altitude MIMO system, characterized in that, Includes the following steps: S1. Based on the pilot signals uploaded by each UAV received by the ground control station, perform initial channel estimation on the channel matrix of the uplink between each UAV and the ground control station to obtain initial channel state information. S2. Based on the initial channel state information, decode the semantic information matrix uploaded by each UAV received by the ground control station to obtain the semantic information of each UAV; perform an inverse mapping operation on the semantic information to extract the flight speed, horizontal angle of arrival range and pitch angle of arrival range of each UAV to form a "space-state" semantic prior. The "space-state" semantic prior includes the first The drone obtained in the next iteration flight speed drones Horizontal angle of arrival range of the resolvable path between the ground control station and the ground control station and horizontal central angle Pitch reach range and pitch center angle ; S3. Based on the flight speed in the "space-state" semantic prior, calculate the maximum Doppler frequency offset of the uplink between each UAV and the ground control station, construct the Doppler compensation matrix, and perform Doppler phase compensation on the received pilot signal. S4. Based on the horizontal angle of arrival range and pitch angle of arrival range in the "space-state" semantic prior, determine the horizontal angle of arrival index and pitch angle of arrival index respectively, and construct a semantic constraint index set; within the search space defined by the semantic constraint index set, use the matching pursuit algorithm to perform channel sparse reconstruction on the Doppler-compensated pilot signal, and update the channel matrix estimate of the uplink between each UAV and the ground control station. The semantic constraint index set Build it as follows: ， in, Indicates the corresponding horizontal angle of arrival and pitch angle of arrival Index in the angle field dictionary The range of horizontal arrival angles. Range of pitch angles; S5. Determine whether the iteration termination condition is met. If it is met, output the final decoded semantic information and the final channel matrix estimate; otherwise, return to step S2 with the updated channel matrix estimate to continue the iteration.

2. The joint estimation method for semantics and channel in a low-altitude MIMO system according to claim 1, characterized in that, The maximum Doppler frequency offset mentioned in step S3 according to Calculate, where, The carrier wavelength is used; a frequency domain Doppler compensation matrix is ​​constructed based on the maximum Doppler frequency offset. For the received pilot matrix Compensation is performed, and the compensated pilot matrix is ​​as follows: The frequency domain Doppler compensation matrix Diagonal matrix of time-domain Doppler deflection The time-domain Doppler deflection diagonal matrix, obtained through discrete Fourier transform, is as follows: ， in, Indicates the subcarrier spacing. For the number of subcarriers, For subcarrier index, For the first The maximum Doppler frequency shift obtained in the next iteration This represents the diagonalization operation that forms a diagonal matrix.

3. The joint estimation method for semantics and channel in a low-altitude MIMO system according to claim 1, characterized in that, In step S4, the search space of the matching pursuit algorithm in each iteration is limited to... Within, the optimal index value Select as follows: ， in, For the first The observation residual vector of the second matching search For the first angle field dictionary List, Represents finding vectors and dot product operation, Expressing the request The index at which the inner function reaches its maximum value. This indicates the absolute value operation.

4. The joint estimation method for semantics and channel in a low-altitude MIMO system according to claim 1, characterized in that, The iteration termination condition in step S5 is that the iteration terminates when any of the following conditions are met: S5.1 Normalized difference between the channel matrix estimates of the current iteration and the previous iteration Less than the preset iteration termination threshold ; The normalized differences are as follows calculate; in, Represents finding a matrix The Frobenius norm, and They represent the first Second and third The estimated channel matrix obtained in the next iteration; S5.2 The current iteration count has reached the preset maximum iteration count. .

5. The joint estimation method for semantics and channel in a low-altitude MIMO system according to claim 4, characterized in that, The iteration termination threshold according to Settings, in which The preset constant false alarm rate, Represents the Gaussian error function. For noise variance, This refers to the number of antennas at the ground station. For the first The estimated channel matrix obtained in the next iteration; The maximum number of iterations Maximum processing latency allowed by the system Constraints, satisfaction ; in To initialize processing delay, This represents the processing delay for a single iteration.

6. The joint estimation method for semantics and channel in a low-altitude MIMO system according to claim 1, characterized in that, The inner loop stopping criterion of the matching pursuit algorithm in step S4 is: the number of currently estimated paths reaches the preset maximum number of paths for low-altitude UAV communication links. Or the energy of the current observed residual is less than the preset stop decision threshold. ; The stop judgment threshold according to set up, As a preset penalty factor, This represents the noise variance.

7. A joint semantic and channel estimation system for a low-altitude MIMO system, characterized in that, A method for jointly estimating semantics and channel in a low-altitude MIMO system as described in any one of claims 1 to 6; the system comprises: The initial channel estimation module is used to perform initial channel estimation on the uplink channel matrix between each UAV and the ground control station based on the pilot signals uploaded by each UAV received by the ground control station, and to obtain initial channel state information. The semantic decoding and prior extraction module is used to decode the semantic information matrix uploaded by each UAV received by the ground control station based on the initial channel state information or the updated channel matrix estimate, to obtain the semantic information of each UAV, and to perform an inverse mapping operation on the semantic information to extract the flight speed, horizontal angle of arrival range and pitch angle of arrival range of each UAV, forming a "space-state" semantic prior. The Doppler compensation module is used to calculate the maximum Doppler frequency offset and construct the Doppler compensation matrix based on the flight speed in the "space-state" semantic prior, and to perform Doppler phase compensation on the received pilot signal. The semantic constraint channel reconstruction module is used to construct a semantic constraint index set based on the horizontal angle of arrival range and the pitch angle of arrival range in the "space-state" semantic prior, and to perform sparse channel reconstruction using the matching pursuit algorithm within the search space defined by the semantic constraint index set, and update the channel matrix estimate. The iteration control module is used to determine whether the iteration termination condition is met. If it is not met, the semantic decoding and prior extraction module, Doppler compensation module and semantic constraint channel reconstruction module are controlled to continue iterating.

8. An emergency communication support system, characterized in that: The method for joint estimation of semantics and channel in a low-altitude MIMO system according to any one of claims 1 to 6; the system comprises: Ground control stations are deployed at emergency site command centers or mobile command platforms; Unmanned aerial vehicles (UAVs) are aerial patrol drones or communication relay drones that perform emergency missions. When the regular ground communication link is interrupted or unavailable, semantic information is transmitted through the uplink between each UAV and the ground control station. The joint iteration of semantic decoding and channel estimation is used to enhance and ensure reliable communication between the emergency command center and the UAVs.

9. An emergency communication support system according to claim 8, characterized in that: The semantic information includes semantic feature information generated by semantically encoding disaster site environmental perception data collected by optical cameras, infrared thermal imagers, or lidar sensors carried by the UAV. The flight speed, horizontal angle of arrival range, and pitch angle of arrival range in the "space-state" semantic prior are obtained by fusing the UAV's onboard GPS, inertial measurement unit, and visual odometry and then embedded into the semantic information and uploaded.

Images