Signal transmission method and system supporting multi-group air computing parallel transmission

Abstract

The invention relates to a signal transmission method and system supporting multi-group air computing parallel transmission. According to the method, non-orthogonal superposition transmission is carried out on multiple groups of calculation signals under the condition of sharing time-frequency resources, and cooperative configuration is carried out on transmitting parameters of a wireless terminal and receiving parameters of a central node on the premise that the minimum communication rate constraint of the wireless terminal is met. Each wireless terminal weights the calculation signals according to the emission parameters and then performs non-orthogonal transmission, the central node processes the received superposed signals group by group according to a preset sequence, and the aerial calculation result of each calculation task group is obtained through receiving scaling and interference elimination. According to the method, parallel execution of multiple groups of calculation tasks is realized under the condition of not increasing extra time-frequency resources, calculation signals of multiple calculation task groups can be transmitted in the same wireless channel under the condition of sharing the same time-frequency resource, and corresponding calculation results are obtained in a group-by-group processing mode, so that the calculation efficiency is improved. Therefore, occupation of time-frequency resources among different computing task groups is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, in particular to a signal transmission method and system supporting parallel transmission of multiple groups of over-the-air computation. BACKGROUND

[0002] With the development of wireless communication and distributed computing technology, over-the-air computation (AirComp) as a technology for realizing function computation by utilizing the superposition characteristics of wireless channels has been widely applied in model aggregation, data fusion and other scenarios. Existing over-the-air computation technology usually performs function computation tasks for a single group of terminals, and obtains the target computation result directly at the receiving end by making multiple wireless terminals transmit weighted signals on the same time-frequency resource.

[0003] However, in practical applications, multiple computation tasks often need to be processed simultaneously in the same system. If orthogonal time division or frequency division is used to distinguish different computation tasks, the time-frequency resources will be further divided, thereby reducing the overall spectral efficiency of the system and making it difficult to meet the demand for parallel computation of multiple tasks.

[0004] Non-orthogonal multiple access technology can superimpose multiple user signals on the same time-frequency resource to realize resource reuse. Existing non-orthogonal transmission schemes are mainly designed for data communication scenarios, and their optimization goals are usually to improve communication rate or system throughput, and are not adapted to the characteristics of "signal superposition is computation result" in the over-the-air computation scenario. In the case of parallel transmission of multiple groups of over-the-air computation tasks, if the existing non-orthogonal transmission method is directly used, the superposition of signals between groups will introduce additional interference, making it difficult to ensure both computation accuracy and communication reliability. SUMMARY

[0005] The present application provides a signal transmission method and system supporting parallel transmission of multiple groups of over-the-air computation, which performs non-orthogonal superposition transmission of multiple groups of computation signals under the condition of sharing time-frequency resources, and cooperatively configures the transmission and reception parameters, so as to improve the spectral efficiency while reducing the computation error of multiple groups of computation tasks and improving the overall communication reliability of the system. In the present application, the center node is a network node for receiving signals sent by multiple wireless terminals and performing over-the-air computation processing on the signals.

[0006] The present application is not aimed at improving decoding performance in traditional communication, but at the technical problem of how to obtain the computation results of multiple computation tasks in the presence of inter-group superposition interference when the tasks are executed in parallel on the same time-frequency resource in over-the-air computation.

[0007] To achieve the above-mentioned purpose, the present application adopts the following technical solutions.

[0008] A signal transmission method supporting parallel transmission of multiple sets of in-flight computation, comprising:

[0009] Obtaining channel state information of multiple wireless terminals, and dividing the wireless terminals into at least two sets of computation task groups according to a preset rule, each set of computation task groups corresponding to a set of in-flight computation tasks;

[0010] Under the condition of meeting the minimum communication rate constraint of each wireless terminal, performing collaborative configuration of the transmission parameters of each wireless terminal and the reception parameters of the central node according to a preset computation accuracy optimization target, the collaborative configuration process at least including determining the transmission power coefficients corresponding to each wireless terminal and the reception scaling factors corresponding to the central node, wherein the reception scaling factors can be set for different sets of computation task groups respectively, and the minimum communication rate constraint is used to limit the communication reliability of each set of computation task groups in the decoding process during the parameter collaborative configuration process. In the present application, the transmission power coefficients and the reception scaling factors are not independently set, but are collaboratively configured according to the channel state information of each set of computation task groups, so that while meeting the minimum communication rate constraint of each set of computation task groups, the error between the computation result obtained by the central node and the target computation result is reduced.

[0011] Each wireless terminal performs power weighting on its own computation signal according to the determined transmission power coefficient, and then performs non-orthogonal superposition transmission on the same time-frequency resource;

[0012] After the central node receives the superposition signal, it processes the signals of each set of computation task groups in turn according to a preset inter-group decoding order, the processing order can be determined based on the channel conditions, transmission power levels or system-set priorities of each set of computation task groups, and in the signal processing of each set of computation task groups, the reception scaling factor is used to scale the received signal to obtain the in-flight computation result of the corresponding set of computation task groups;

[0013] After completing the computation result acquisition of the current set of computation task groups, the central node performs interference cancellation on the processed computation signal, and continues to process the signal of the next set of computation task groups until the computation of all sets of computation task groups is completed.

[0014] The beneficial effects of the present application are:

[0015] By performing non-orthogonal superposition transmission of multiple sets of computation signals on the same time-frequency resource, and combining the reception scaling factors set for different sets of computation task groups and the group-by-group interference cancellation processing, the parallel execution of multiple sets of in-flight computation tasks is realized without increasing additional time-frequency resources, thereby reducing the computation error of multiple sets of computation tasks under the shared channel condition while ensuring the minimum communication rate requirement, and improving the spectrum utilization efficiency and communication reliability of the system in the multi-task parallel in-flight computation scenario. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flowchart of a signal transmission method supporting parallel transmission of multiple groups of over-the-air computation according to an embodiment of the present application;

[0017] Figure 2 is an iterative optimization flowchart for cooperative configuration of transmission parameters and reception parameters according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] In the present application, the central node is a network node configured to receive signals transmitted by multiple wireless terminals and perform over-the-air computation processing on the signals. The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be understood that the following embodiments are only used to explain the present application, and are not used to limit the protection scope of the present application.

[0019] In one embodiment, the system includes a central node and multiple wireless terminals. The central node communicates with the multiple wireless terminals through wireless channels, and each wireless terminal performs a corresponding over-the-air computation task according to the computation task group to which it belongs. In one embodiment, each wireless terminal belongs to only one computation task group, and each computation task group corresponds to a group of over-the-air computation tasks.

[0020] As shown in Figure 1 , the central node first acquires channel state information of each wireless terminal, and divides the wireless terminals into at least two computation task groups according to a preset rule. The preset rule can be set based on channel gain size, service demand or computation task priority of the terminals, which is not limited in the present application.

[0021] After determining the computation task groups, the central node performs a parameter cooperative configuration process (also referred to as a resource configuration process) under the premise of satisfying the minimum communication rate constraints of each wireless terminal. The resource configuration process jointly determines the transmission power coefficients corresponding to each wireless terminal and the reception scaling factors corresponding to the central node, with the goal of reducing the overall computation error of multiple computation task groups. The transmission power coefficients are used to weight the computation signals transmitted by the wireless terminals, and the reception scaling factors can be set for different computation task groups to adjust the amplitudes of the superimposed signals received by the central node.

[0022] After resource configuration is completed, each wireless terminal transmits its power-weighted computational signals in a non-orthogonal superposition on the same time-frequency resources. Upon receiving the superimposed signals, the central node processes each group of signals sequentially according to a preset computational task group order. This order can be determined based on the channel conditions, transmit power levels, or system-defined priorities of each computational task group. When processing the signals of a particular computational task group, the central node scales the received signal using a corresponding receive scaling factor to obtain the over-the-air computation result for that task group.

[0023] After obtaining the calculation results of the current computing task group, the central node performs interference elimination on the processed computing signals and continues to process the signals of the next computing task group until the calculations of all computing task groups are completed.

[0024] In an optional implementation, the resource configuration process can be performed iteratively, that is, updating the remaining parameters while keeping some parameters fixed until a preset convergence condition is met. Those skilled in the art can adjust the specific implementation method according to the system scale and computational complexity requirements.

[0025] In another alternative implementation, when channel estimation errors exist, the resource configuration process can be set based on statistical channel information to improve the robustness of the system under channel uncertainty conditions.

Claims

1. A signal transmission method supporting parallel transmission of multiple sets of over-the-air computing, characterized in that, include: Step 1: The central node obtains the channel state information between multiple wireless terminals and the central node, and divides the multiple wireless terminals into at least two computing task groups based on the channel state information. Each computing task group corresponds to a set of over-the-air computing tasks. Step 2: Under the condition of sharing the same time-frequency resources, the central node, based on the channel state information, collaboratively determines the transmit power coefficient for amplitude or power weighting of the computation signal sent by the wireless terminal, and the receive scaling factor set for different computation task groups, provided that the minimum communication rate constraints corresponding to each computing task group are met. The receive scaling factor is used to scale the amplitude of the superimposed signal received by the central node to reduce the error between the air computing result obtained by the central node and the corresponding target computing result. Step 3: Each wireless terminal calculates its signal based on the corresponding transmit power coefficient, and then performs non-orthogonal superposition transmission on the same time-frequency resource; Step 4: After receiving the superimposed signal, the central node scales the superimposed signal according to the preset processing order of the computing task groups, based on the receiving scaling factor corresponding to the computing task group, so as to obtain the air computing result corresponding to the computing task group in the presence of interference from signals of other computing task groups. Step 5: After obtaining the over-the-air computation results of the current computation task group, the central node performs interference cancellation operations on the signals corresponding to the processed computation task group, and continues to process the signals of the next computation task group until the over-the-air computation of all computation task groups is completed.

2. The method according to claim 1, characterized in that, The method for determining the transmit power coefficient and the receive scaling factor is based on channel state information and aims to minimize the calculation error, thereby reducing the error between the air calculation results obtained by the central node and the target calculation results.

3. The method according to claim 1, characterized in that, Within a preset parameter update cycle, the central node updates the transmit power coefficient under a fixed receive scaling factor, and updates the receive scaling factor under the condition that the updated transmit power coefficient remains unchanged. The parameter update is terminated when the change in either the transmit power coefficient or the receive scaling factor in two consecutive updates does not exceed a preset threshold.

4. The method according to claim 1, characterized in that, The transmit power coefficient is a real scalar used to perform amplitude or power weighting on the calculated signal transmitted by the wireless terminal to compensate for channel fading and achieve over-the-air superposition calculation.

5. The method according to claim 1, characterized in that, The receiving scaling factor is a scaling parameter corresponding to the calculation task group, used to adjust the amplitude of the superimposed signal received by the central node in order to reduce the intra-group or inter-group error in the calculation results.

6. The method according to claim 1, characterized in that, The minimum communication rate constraint is used to ensure that each wireless terminal can perform over-the-air computing transmission while meeting the basic communication reliability requirements.

7. The method according to claim 1, characterized in that, The processing order of the computing task groups is determined based on the channel conditions, transmit power level, or preset priority of each computing task group.

8. The method according to claim 1, characterized in that, When channel estimation errors exist, the cooperative configuration process is set based on statistical channel information.

9. The method according to claim 1, characterized in that, The aerial computing task includes at least one of model parameter aggregation, data fusion, or function computation tasks.

10. A signal transmission system supporting parallel transmission of multiple sets of over-the-air computing, characterized in that, include: Multiple wireless terminals are configured to weight their respective calculated signals based on the transmit power coefficient determined by the central node, and then transmit them in a non-orthogonal superposition on the same time-frequency resources. The system includes a central node comprising: a channel acquisition module for acquiring channel state information of the wireless terminal; a parameter configuration module for determining the transmit power coefficient and the receive scaling factor for different computing task groups under the premise of meeting the minimum communication rate constraint; and a signal processing module for processing, receiving, scaling, and canceling interference of the received superimposed signals in a preset order to obtain the over-the-air computing results corresponding to multiple computing task groups.

Images