Limited block length optimization method and system of cellular-free communication perception integrated system

Abstract

The invention discloses a limited block length optimization method and system of a cellular-free communication perception integrated system, and belongs to the technical field of wireless communication. The method comprises the following steps: constructing a cellular-free multi-station cooperative communication perception integrated architecture, and establishing a channel estimation and data transmission model; deducing a closed expression of the communication transmission delay, the decoding error probability and the sensing precision Cramer-Rao bound under the limited block length; analyzing a Pareto optimal boundary of communication and sensing performance by adopting a non-dominated sorting genetic algorithm, and determining a block length feasible region; and designing a dynamic optimization algorithm based on a double-depth Q learning network, and realizing collaborative optimization of the communication performance by adaptively adjusting the block length. The method solves the problem that high communication reliability, low time delay and high sensing precision are difficult to consider in a dynamic environment, the overall performance of the system is remarkably improved, and the method is suitable for high-requirement scenes such as automatic driving and industrial Internet of Things.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technologies, and particularly to a method and system for optimizing the finite block length of a cell-free communication and sensing integrated system. Background Art

[0002] As a core technology of the sixth-generation mobile communication system, the communication and sensing integrated technology has shown important application value in key fields such as autonomous driving and industrial Internet of Things. This technology realizes the efficient coordination of communication and sensing functions by sharing spectrum resources and hardware devices, significantly improving the system resource utilization rate. The cell-free multi-station cooperative system adopts a distributed access point architecture and is centrally coordinated by a central processing unit. Its network topology flexibility and spatial diversity gain create new possibilities for the joint improvement of communication and sensing performance, and have received extensive attention in the industry.

[0003] High-reliability and low-latency scenarios require a finite block length transmission mechanism. The communication and sensing performance is constrained by the block length and presents an inherent coupling relationship: shortening the block length can reduce the communication transmission delay, but the reliability will deteriorate due to the decrease in coding gain; increasing the block length can improve the sensing accuracy and communication reliability, but it will introduce non-negligible delay overhead. This contradiction is particularly prominent in dynamic channel environments and mobile target scenarios. The traditional static block length allocation scheme is difficult to adapt to the real-time changing system requirements, lacks consideration of time-varying channel conditions, and fails to take into account the adaptive ability of multi-objective trade-off between communication and sensing, severely restricting the upper limit of system performance. These problems urgently require effective solutions. Summary of the Invention

[0004] The present invention provides a method and system for optimizing the finite block length of a cell-free communication and sensing integrated system. In the current research field of cell-free multi-station cooperative communication and sensing integration, existing work mainly focuses on static resource allocation and performance optimization under ideal channel conditions, often ignoring the dynamic impact of the time-varying characteristics of the channel on system performance in actual scenarios. Especially in high-reliability and low-latency communication scenarios, existing solutions lack in-depth analysis of the coupling relationship between communication and sensing performance under the finite block length effect, resulting in difficulty in achieving dynamic balance among communication reliability, low latency, and sensing accuracy. In response to the above-mentioned urgent technical problems, the method proposed by the present invention has breakthrough innovation value. This solution breaks through the limitations of traditional static resource allocation and solves the key technical problem of difficult to balance communication reliability, low latency, and high-precision sensing in a dynamic environment through an adaptive block length allocation strategy.

[0005] An embodiment of the present invention provides a method for optimizing the finite block length of a cell-free communication and sensing integrated system, including the following steps:

[0006] Step 1: Construct a cell-free communication and sensing integrated architecture including distributed access points, communication users, and sensing targets, and establish a channel estimation scheme and a data transmission model;

[0007] Step 2: Based on the architecture and model constructed in Step 1, derive the communication and sensing performance metrics under finite blocklength, including communication transmission delay, decoding error probability, and the Cramér-Rao bound of sensing accuracy;

[0008] Step 3: Based on the performance metrics derived in Step 2, use the non-dominated sorting genetic algorithm to jointly optimize the communication delay and sensing accuracy, and determine the feasible region of the blocklength that meets the basic performance of communication and sensing;

[0009] Step 4: Based on the feasible region of the blocklength obtained in Step 3, design an adaptive blocklength optimization strategy based on the double deep Q-learning network, and realize the collaborative improvement of communication and sensing performance by dynamically adjusting the blocklength.

[0010] Preferably, Step 1 specifically includes:

[0011] Step 101, based on the cell-free communication and sensing integrated architecture, establish a system scenario model including M distributed access points, K communication users, and 1 sensing target, where the access points are divided into a downlink access point set and an uplink access point set according to the duplex mode, and the users are divided into a downlink user set and an uplink user set;

[0012] Step 102, based on the system scenario model, adopt the quasi-static channel assumption to establish a channel model including large-scale path loss and small-scale Rayleigh fading for the communication channel, and establish a Swerling type I model for the sensing channel;

[0013] Step 103, based on the channel model, design a pilot-assisted minimum mean square error channel estimation scheme for obtaining the communication channel state information, and establish a sensing channel reconstruction mechanism;

[0014] Step 104, based on the channel estimation scheme, construct a composite signal model including communication signals and sensing signals transmitted by the downlink access points, and a received signal model for the uplink and downlink links, where the received signal model includes desired signals and various interference terms.

[0015] Preferably, when deriving the communication and sensing performance metrics in Step 2, it specifically includes:

[0016] Based on the received signal model constructed in Step 1, derive the closed-form expressions of the signal-to-interference-plus-noise ratio for the uplink and downlink links;

[0017] According to the signal-to-interference-plus-noise ratio expression, define the mathematical relationship between the transmission delay and the decoding error probability;

[0018] Based on the sensing channel model in step 1, calculate the Cramér-Rao bound of the sensing accuracy.

[0019] Preferably, the multi-objective joint optimization in step 3 specifically includes:

[0020] Taking the transmission delay and the Cramér-Rao bound defined in step 2 as the objectives, construct a multi-objective optimization problem;

[0021] Use the non-dominated sorting genetic algorithm, combined with the crowding distance operator to solve the Pareto front solution set;

[0022] Generate the feasible region of the block length design according to the Pareto solution set and the basic requirements of communication and sensing performance.

[0023] Preferably, the adaptive block length optimization strategy in step 4 specifically includes:

[0024] Based on the block length feasible region obtained in step 3, define the state space and action space of the double-depth Q-learning network;

[0025] Design a joint reward function, integrating the reliability constraint penalty term derived from the communication delay, sensing accuracy, and upper bound of the decoding error probability defined in step 2;

[0026] Train the network through experience replay and temporal difference objectives, and output the adaptive block length strategy.

[0027] The present invention also provides a cell-free communication and sensing integrated system, including:

[0028] A distributed access point array for cooperatively transmitting communication signals and sensing signals;

[0029] A central processing unit for executing the method according to any one of claims 1-5 to achieve dynamic block length optimization;

[0030] A channel estimation module for obtaining the communication channel state information based on the pilot sequence;

[0031] A sensing interference cancellation module for reconstructing the sensing channel and suppressing the communication signal interference.

[0032] Preferably, the access point uses a uniform linear array antenna with an element spacing of half a wavelength, supporting beamforming of downlink communication and active sensing signals.

[0033] The present invention also provides an application of the method in the field of autonomous driving, which can simultaneously meet the requirements of low-latency communication and high-precision environmental sensing in the scenario of high-speed vehicle movement by dynamically adjusting the block length.

[0034] The present invention also provides an application of the method in industrial Internet of Things for the collaborative optimization of reliable data transmission and device positioning in a complex multi-device environment.

[0035] The present invention also provides a computer-readable storage medium storing a computer program, and when the program is executed by a processor, the steps of the method are implemented.

[0036] Beneficial effects: The finite block length optimization scheme of the cell-free communication and sensing integrated system according to the embodiments of the present invention constructs a comprehensive solution for the technical problem that it is difficult to jointly optimize the communication and sensing performances in a dynamic channel environment in the research of cell-free multi-station cooperative communication and sensing integration. This scheme first establishes a systematic method for communication channel estimation and sensing channel reconstruction, derives closed-form expressions for multi-dimensional performances such as communication transmission delay, decoding error probability, and sensing accuracy under the constraint of finite block length, and reveals the quantitative relationship between the block length and the system performance; secondly, designs a reinforcement learning algorithm based on a double deep Q-network, takes the block length as an optimization variable to generate an adaptive adjustment strategy, and at the same time, by introducing a feasible region dynamic adjustment mechanism, automatically updates the block length optimization range when the channel conditions change to ensure that the system always operates in the best state. This scheme breaks through the limitations of traditional static resource allocation, and through the adaptive block length allocation strategy, solves the key technical problem that it is difficult to balance communication reliability, low latency, and high-precision sensing in a dynamic environment.

[0037] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. Description of the Drawings

[0038] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:

[0039] Figure 1 is a flowchart of the finite block length optimization method for the cell-free communication and sensing integrated system according to the embodiments of the present invention;

[0040] Figure 2 shows the dynamic adjustment process of the optimal block length strategy of the system under different reliability standard constraints, as well as the communication and sensing performances corresponding to the optimal block length strategy. Detailed Embodiments

[0041] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0042] Based on the problems raised in the background art, the existing cell-free communication and sensing integrated system is difficult to achieve the collaborative optimization of communication delay, reliability, and sensing accuracy in a dynamic environment, and the traditional static block length allocation scheme cannot adapt to the time-varying characteristics of the channel. Therefore, the present invention explores a finite block length optimization method for a cell-free communication and sensing integrated system to solve the above technical problems.

[0043] Figure 1 FIG. is a flowchart of a finite block length optimization method for a cell-free communication and sensing integrated system according to an embodiment of the present invention.

[0044] As Figure 1 shown, the finite block length optimization scheme includes the following steps:

[0045] Step 1, construct a cell-free communication and sensing integrated architecture including distributed access points, communication users, and sensing targets, and construct a communication and sensing architecture, a channel estimation scheme, and a data transmission model under finite block lengths.

[0046] In an embodiment of the present invention, step 1 specifically includes:

[0047] Step 101, establish a scenario model of a cell-free multi-station cooperative communication and sensing integrated system, including M distributed access points, K single-antenna communication users, and 1 sensing target to be sensed. Each access point is configured with N uniform linear array antennas, and the element spacing is half of the operating wavelength. For a signal with an incident angle of θ, its normalized array steering vector a(θ) is defined as an N-dimensional column vector, and the expression of the nth element is n is an integer with a value in [0, N-1]. According to the duplex mode of the access points, the M access points can be divided into a downlink access point set and an uplink access point set According to the communication requirements of the users, the communication users can be divided into a downlink user set and an uplink user set wherein, the downlink access points simultaneously transmit downlink user communication signals and active sensing signals, and the uplink access points receive uplink user communication signals and echo signals reflected by the sensing target. It is defined that all access points are connected to the central processing unit through a backhaul link, and a time-frequency resource sharing mechanism is adopted to form a centralized control network architecture.

[0048] Step 102, adopt the quasi-static channel hypothesis, and the communication and sensing channels remain unchanged within the block length L. The communication channel considers large-scale path loss and small-scale Rayleigh fading. The downlink channel from the mth downlink access point to the kth downlink user can be modeled as a row vector of dimension N, following a complex Gaussian distribution with a mean of 0 and a variance of λ dl,k,m I N and m is a value in [1, M dlan integer, and k is an integer taking values in [1, K dl an integer. The channel from the i-th uplink user to the n-th uplink access point can be modeled as a row vector of dimension N, following a complex Gaussian distribution with mean 0 and variance λ ul,i,n I N , where i is an integer taking values in [1, K ul , and n is an integer taking values in [1, M ul .

[0049] The sensing channel adopts the Swerling type I model assumption, and the target radar cross-section area is constant within the coherent processing interval. The sensing channel from the m-th downlink access point through the target reflection to the n-th uplink access point can be defined as where the coefficient considers the radar cross-section area and the two-way path loss. Among them, θ m represents the angle of departure from the m-th downlink access point to the target, and θ n represents the angle of arrival from the target to the n-th uplink access point.

[0050] Step 103, design a communication channel estimation scheme and a sensing channel reconstruction mechanism. The system uses a pilot-assisted minimum mean square error estimation method to obtain the channel state information of communication. In the uplink, all users send mutually orthogonal pilot sequences. After all access points receive the pilot sequences, the central processing unit estimates the channel between the uplink users and the uplink access points. The channel can be estimated as an N-dimensional complex Gaussian variable with variance β ul I N ; in the downlink, a symmetric design is adopted. All downlink access points send mutually orthogonal pilot sequences, and all downlink users can estimate the downlink communication channel. The channel can be estimated as an N-dimensional complex Gaussian variable with variance β dl I N . At the same time, the uplink access point can estimate the line-of-sight channel between it and the downlink access point for subsequent sensing interference cancellation.

[0051] For the sensing channel, the central processing unit eliminates the line-of-sight channel between the downlink access point and the uplink access point, and uses the target position information and the prior knowledge of the radar cross-section area obtained in the previous time slot for channel reconstruction. Among them, the covariance matrix of the channel reconstruction error between the m-th downlink access point and the n-th uplink access point can be expressed as a matrix of dimension N×N

[0052] Step 104, the composite signal x dl,m [l] transmitted by the downlink access point can be modeled as a linear superposition of the communication signal and the sensing signal, where l is an integer taking values in [1, L]. For the sensing symbol, the power allocated at the m-th downlink access point is The conjugate beamforming method is adopted to design the sensing beamforming vector; for the communication symbol, the power allocated by the m-th downlink access point to the k-th downlink user is The maximum ratio transmission criterion is adopted to design the communication precoding vector, and the precoding normalization coefficient ξ is used dl To ensure the power constraint of precoding.

[0053] Step 105, in the downlink communication link, using the communication and sensing superimposed signal described in Step 104, the received signal of the k-th downlink user at the l-th symbol Can be specifically expressed as

[0054]

[0055] Where Is the desired signal of the k-th downlink user, Represents the interference term of the signals of other downlink users on the k-th downlink user, Represents the interference term caused by the downlink channel estimation error, Is the interference term of the sensing signal on the downlink communication signal, Represents the interference term caused by the uplink communication user; n dl Represents an additive Gaussian noise term with a mean of 0 and a variance of Of.

[0056] In the uplink communication link, the central processing unit processes the received signal of the uplink access point using the processing vector. The signal of the i-th uplink user at the l-th symbol after processing can be expressed as

[0057]

[0058] Where Is the desired signal of the i-th uplink user, Represents the interference term of the signals of other uplink users on the i-th uplink user, Represents the interference term caused by the uplink channel estimation error, Is the interference term of the sensing signal on the uplink communication signal, Represents the interference term caused by the downlink communication signal; Represents the interference term caused by the uplink noise, and the uplink noise can be modeled as an additive Gaussian noise interference with a mean of 0 and a variance of Of.

[0059] Step 106, After the sensed echo signal reaches the uplink access point after being reflected by the target, the central processing unit can perform the following processing: Eliminate the targetless channel components, including constant environmental noises such as the line-of-sight channel and the fixed reflector channel; in addition, eliminate the communication signal interference by using the transmitted signal and the channel estimation result. After processing, at the l-th symbol, the sensed signal sent by the m-th downlink access point to the n-th uplink access point can be expressed as

[0060]

[0061] where is the ideal sensed signal component from the m-th downlink access point, and w n contains interference terms such as the sensed channel reconstruction error, the residual noise of communication signal cancellation, and the sensed noise. It is defined that the power of the sensed interference term satisfies

[0062] Step 2, Deduce the expressions of the communication and sensing performance metrics under finite blocklength, including the communication transmission delay, the decoding error probability, and the Cramér-Rao bound of the sensing accuracy.

[0063] In an embodiment of the present invention, Step 2 specifically includes:[[]]

[0064] Step 201, Using the received signal model in Step 105, establish the signal-to-interference-and-noise ratio (SINR) expressions for the uplink and downlink. For the k-th downlink user, according to formula (1), the closed-form expression of the received signal SINR can be expressed as

[0065]

[0066] where denotes taking the expectation, denotes the interference and noise term of the k-th downlink user, and β dl,k,m denotes the variance of the estimated channel between the m-th downlink access point and the k-th downlink user, and θ dl,k,m denotes the variance of the channel estimation error between the m-th downlink access point and the k-th downlink user, and λ ul,i,k denotes the channel variance between the i-th uplink user and the k-th downlink user.

[0067] For the i-th uplink user, according to formula (2), the closed-form expression of the SINR can be expressed as

[0068]

[0069] where tr(·) denotes taking the trace of the matrix, denotes the interference and noise term of the i-th uplink user, and β ul,i,ndenotes the variance of the estimated channels between the n-th uplink access point and the i-th uplink user, θ ul,i,n denotes the variance of the channel estimation error between the n-th uplink access point and the i-th uplink user.

[0070] Step 202, for the t-th user, the transmission delay D t can be defined as

[0071]

[0072] where B represents the system bandwidth, ε t is the decoding error probability of the t-th user, and its upper bound can be derived as

[0073]

[0074] where is the complementary cumulative distribution function, denotes the pilot length L p as a proportion of the block length, R t is the achievable rate of the t-th user. Substituting the formulas (4) and (6) in Step 201, the expressions for the transmission delay of the communication user and the upper bound of the decoding error probability can be obtained.

[0075] Step 203, define the bistatic distance parameter d m,n as the sum of the distances from the m-th downlink access point and the n-th uplink access point to the target. The Cramér-Rao bound of this parameter can be expressed as where is the estimated bistatic distance parameter, is the Fisher information matrix between the m-th downlink access point and the n-th uplink access point, (·) * denotes the conjugate operation. The signals of the downlink access points can be assumed to be independent of each other and are transmitted to the central processing unit through the backhaul link for centralized processing. The average Cramér-Rao bound of M ul uplink access points can be obtained as

[0076]

[0077] where the Bessel information matrix of the n-th uplink access point is f is the system carrier frequency, and c represents the speed of light.

[0078] Step 3, use the non-dominated sorting genetic algorithm for multi-objective joint optimization of communication and sensing, quantify the trade-off relationship between communication delay and sensing accuracy, and determine the feasible range of block length that meets the basic performance of communication and sensing.

[0079] In an embodiment of the present invention, Step 3 specifically includes:

[0080] Step 301. For the high-reliability and low-latency communication requirements in the cell-free system, considering the communication reliability under finite block lengths as a constraint and minimizing the weighted transmission delay as the communication optimization objective, define the optimization function f c where

[0081]

[0082] Here, D k represents the transmission delay of the k-th downlink user, and D i represents the transmission delay of the i-th uplink user. The downlink weight coefficient w dl and the uplink weight coefficient w ul reflect the importance differences of different links. ε t(th) represents the reliability threshold of the t-th user, and t is an integer taking values in [1, K dl +K ul .

[0083] Step 302. Aiming to minimize the Cramér-Rao bound of the bistatic distance parameter, construct a sensing performance optimization problem and define the sensing optimization function f s as

[0084]

[0085] Step 303. There is a trade-off relationship between the communication optimization function f c and the sensing optimization function f s under finite block lengths. Define the multi-objective optimization problem f as

[0086]

[0087] Use the non-dominated genetic algorithm to solve the multi-objective optimization problem f. Adopt the Latin hypercube sampling method to generate the initial population and construct a uniformly distributed spatial grid within the feasible region [L min , L max of the block length. Among them, the minimum block length L min is determined by the pilot length, communication reliability, and sensing accuracy requirements, and the maximum block length L max is determined by the transmission delay constraint and the maximum processing capacity of the system.

[0088] Perform the non-dominated sorting operation, sort all the solutions in the current population according to the Pareto dominance relationship, and obtain multiple front ranks. Calculate the crowding distance of each solution in each front rank, and define the crowding distance operator f[k] as

[0089] f[k] = Δf c [k] + Δf s [k], #(14)

[0090] where k represents the sorting index of the current individual within its affiliated frontier layer; define the communication performance crowding distance operator as f c [k] represents the communication function value corresponding to index k, represents the maximum communication function value of this layer, represents the minimum communication function value of this layer; define the sensing performance crowding distance operator as f s [k] represents the sensing function value corresponding to index k, represents the maximum sensing function value of this layer, represents the minimum sensing function value of this layer.

[0091] Perform genetic operations. Use the simulated binary crossover operator to perform crossover operations on the selected parent individuals, and set the crossover probability to 0.9. Perform polynomial mutation operations on the offspring individuals generated by crossover, and set the mutation probability to 0.1. Combine the parent population and the offspring population, sort and screen them according to the frontier level and crowding distance, and retain the optimal non-dominated solutions to form a new generation population. Repeat the above optimization process until the preset maximum number of iterations is reached, and finally output the Pareto front solution set to obtain the block length feasible region that meets the basic communication and sensing performance requirements

[0092] Step 4: Construct the reinforcement learning framework of the double deep Q-learning network, define the state space, action space, and reward function, and obtain the adaptive optimal block length strategy through algorithm convergence and fine-tuning.

[0093] In an embodiment of the present invention, step 4 specifically includes:

[0094] Step 401: Adopt the double deep Q-learning network architecture. Define that at the t-th moment, the current state s t is the block length information L with a dimension of 1 t ; use the Pareto optimal solution set obtained in step 303 as the value range of the action space, and define the action space The double deep Q-learning network adopts a fully connected structure, where the input layer corresponds to the state space and the output layer corresponds to the action space.

[0095] Step 402: Construct the immediate reward function r for joint communication and sensing t as

[0096] r t = c - (w1D t + w2CRB t + p t ), #(15)

[0097] where Dt Indicates the current transmission delay, CRB t Indicates the current Cramer-Rao bound, where w1 and w2 are the weight coefficients for communication and sensing respectively, c is the baseline constant for rewards, and p t is the reliability constraint penalty term, satisfying

[0098]

[0099] where p represents the penalty coefficient.

[0100] Step 403: Initialize the main network parameters θ and the target network parameters θ - . Set the capacity of the experience replay buffer to 10 5 , set the target network update period to C = 100, and the discount factor to γ = 0.9. Adopt the ε-greedy strategy, set the initial exploration probability ε = 1, linearly decay by 0.0005 per training step, with a lower limit of 0.01. Define the training convergence condition as the average reward change rate over 200 consecutive training epochs being less than 0.5%.

[0101] In each training cycle, collect the current block length state s from the current communication and sensing environment t , and obtain the Q-values Q(s t , a t ; θ) corresponding to each action. The main network is responsible for generating the action selection strategy in real time, selecting the action a with a random probability ε t , otherwise output the action corresponding to the maximum Q-value Execute the selected action a t After that, calculate the immediate reward r through the reward function defined in Step 401 t , and observe the next state s t+1 , and store the sample (s t , a t , r t , s t+1 ) into the experience replay buffer. The target network is independent of the main network and calculates the temporal difference target y t , defined as

[0102]

[0103] where the target network parameter θ - is synchronized with the main network θ - ← θ every C steps through a hard update method. Define the mean squared error loss function as After training reaches the convergence condition, output the optimal block length strategy.

[0104] Step 404, execute the test and fine-tuning cycle. In each test cycle, use the network model to obtain the optimal block length decision Collect the current channel state information and monitor the current communication transmission delay D in real time t , the decoding error probability ε t and the sensing accuracy CRB t . Set the network fine-tuning trigger condition to the failure of the reliability constraint in consecutive T test test cycles.

[0105] Based on the current channel conditions, re-execute the Pareto optimization in step 303 to generate a new feasible region of block lengths Update the action space of the double deep Q-learning network to the new feasible region and perform fine-tuning tests while maintaining the exploration probability ε. Adopt a sliding window evaluation mechanism to statistically calculate the performance compliance rate ρ of the most recent T fine fine-tuning cycles,

[0106]

[0107] wherein, represents the calculation of the probability of the event occurring within the parentheses. When ρ≥99%, it is determined that the fine-tuning operation converges, and at this time, the system reaches a stable working state.

[0108] The following details the finite block length optimization scheme of the cell-free communication and sensing integrated system of the present invention through a specific embodiment.

[0109] Suppose there is a cell-free cooperative communication and sensing integrated system with a range of 300×300 meters, in which 16 access points are evenly distributed, including 8 downlink access points and 8 uplink access points, jointly providing services for 4 randomly distributed downlink users and 4 uplink users in the area. The system operates in the 3.9 GHz band, the maximum transmit power of the downlink access point is 1 W, the transmit power of the uplink user is fixed at 50 mW, and the noise power of the communication receiving end is -114 dBm. The simulation targets a high-reliability and low-latency scenario, uses 10 pilot symbols for channel estimation, the maximum allowable block length is 120 symbols, and realizes the co-optimization of communication and sensing by dynamically adjusting the block length within the range of 20 to 120 symbols. The decoding error probability threshold is set to 10 -3 to 10 -9 .

[0110] Figure 2It shows the finite block length optimization effect of the cell-free communication and sensing integrated system in the embodiments of the present invention, and reveals the quantitative relationship between the decoding error probability threshold, the optimal block length and the Cramer-Rao lower bound under different antenna configurations. The three solid lines in the figure respectively correspond to the optimal block length change curves with the number of antennas per access point being N = 4, 6, and 8, revealing the quantitative relationship between the block length and communication reliability in a dynamic channel environment; the three groups of bar charts show the Cramer-Rao lower bound values under the corresponding antenna configurations, reflecting the change trend of sensing accuracy with the reliability threshold. It can be observed from the illustrated results that as the number of antennas increases, the system can support fewer block length resources under the same detection error probability threshold, and at the same time obtain a lower Cramer-Rao lower bound value. In addition, the data in the figure shows that as the reliability requirement increases, that is, when the reliability threshold moves towards 10 -9 moves, the feasible region of the block length significantly narrows, and the change range of the optimal block length is compressed from [50, 120] symbols to [70, 120] symbols. For the scenario with antenna configuration N = 6, as the reliability threshold increases, the square root value of the sensing Cramer-Rao bound decreases from 0.122 m to 0.094 m, with an increase of about 23%. This constraint relationship reflects the technical necessity of the dynamic feasible region adjustment mechanism proposed in the present invention, which not only ensures communication reliability but also stabilizes the fluctuation of sensing accuracy.

[0111] In summary, the present invention aims at the technical problems that it is difficult to jointly optimize the communication and sensing performance in a dynamic channel environment and the static resource allocation cannot balance high-reliability communication and high-precision sensing in a cell-free multi-station cooperative communication and sensing integrated system, and constructs a comprehensive solution. The present invention first establishes a systematic method for communication channel estimation and sensing channel reconstruction, constructs a multi-dimensional performance model including communication transmission delay, decoding error probability and sensing accuracy, and derives a closed-form expression under the constraint of finite block length; secondly, designs a reinforcement learning algorithm based on a double deep Q network to jointly optimize the communication and sensing performance by dynamically adjusting the block length, and introduces a dynamic feasible region adjustment mechanism to ensure that the system always operates in the best state. The solution proposed by the present invention breaks through the limitations of traditional static resource allocation and effectively solves the key technical problems that it is difficult to balance communication reliability, low latency and high-precision sensing in a dynamic environment. Therefore, the present invention has certain practical value.

[0112] In an embodiment of the present invention, the cell-free communication and sensing integrated system includes the following modules:

[0113] 1. Distributed access point array:

[0114] It adopts 16 uniform linear array (ULA) antennas with an element spacing of half a wavelength and a working frequency band of 3.9 GHz.

[0115] The downlink access points (8 in number) transmit composite signals (downlink communication signal + sensing signal), and the uplink access points (8 in number) receive the uplink communication signal and the sensing echo signal.

[0116] Support maximum ratio transmission (MRT) precoding and conjugate beamforming, with a transmit power of 1W (downlink) and 50mW (uplink).

[0117] 2. Central processing unit:

[0118] Run a double deep Q - learning network (DDQN), with the input being the current block - length state (20 - 120 symbols) and the output being the dynamically adjusted optimal block length.

[0119] Generate the block - length feasible region (such as [50, 120] symbols) through the non - dominated sorting genetic algorithm (NSGA - II), ensuring that the communication delay < 10ms and the sensing accuracy CRB < 0.1m.

[0120] 3. Channel estimation module:

[0121] Based on the minimum mean - square error (MMSE) estimation, using 10 orthogonal pilot symbols, the estimated error covariance matrix is β dl I N (downlink) and β ul I N (uplink).

[0122] 4. Sensing interference cancellation module:

[0123] Utilize the prior information of the target position in the previous time slot to reconstruct the sensing channel G m,n , eliminate the interference of the communication signal, and the residual noise power

[0124] In an embodiment of the present invention, the composite signal transmitted by the downlink access point is modeled as a linear superposition of the communication signal and the sensing signal; after the uplink access point receives the signal and performs matched filtering, it is jointly processed by the central unit to suppress multi - user interference.

[0125] In an embodiment of the present invention, for the autonomous driving application scenario:

[0126] Dynamic block - length adjustment: When the vehicle speed changes, the block length adaptively adjusts from 70 symbols (low speed) to 50 symbols (high speed), ensuring that the delay < 5ms.

[0127] Sensing accuracy guarantee: Through the optimization of the Cramer - Rao bound, the target distance estimation error ≤ 0.05m (when the number of antennas N = 8).

[0128] Reliability constraint: The decoding error probability ε t ≤10 -6 , and it is verified in real - time through formula (9).

[0129] In one embodiment of the present invention, the industrial Internet of Things application scenario:

[0130] Multi-device collaboration: In a factory environment with 50 devices, the block length feasible region is from 50 symbols to 80 symbols, and it is adaptively adjusted according to communication and sensing requirements:

[0131] Communication: Bit error rate ε t ≤10 -6 (verified by formula (9)).

[0132] Sensing: The device positioning error < 0.3m (after CRB optimization).

[0133] Interference suppression: The sensing interference cancellation module reduces the residual power of the communication signal to below -120 dBm.

[0134] In one embodiment of the present invention, the implementation of the computer-readable storage medium:

[0135] The stored program includes the following modules:

[0136] 1. Channel estimation module: Performs MMSE estimation and outputs the channel matrix.

[0137] 2. NSGA-II optimization module: The input is the signal-to-interference-plus-noise ratio (formulas (4)(6)) and CRB (formula (10)), and the output is the Pareto optimal block length range.

[0138] 3. DDQN decision module: The state s t is the current block length L t , and the action a t is the adjusted block length, and the reward function is calculated according to formula (15).

[0139] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or N embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0140] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0141] Any process or method description in the flowchart or described otherwise herein can be understood to represent a module, segment, or portion of code including one or N executable instructions for implementing a customized logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations where functions may be executed not in the order shown or discussed, including in a substantially simultaneous manner according to the involved functions or in a reverse order, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

Claims

1. A finite blocklength optimization method for a cell-free communication and sensing integrated system, characterized in that It includes the following steps: Step 1: Construct a cell-free communication and sensing integrated architecture including distributed access points, communication users, and sensing targets, and establish a channel estimation scheme and a data transmission model; Step 2: Based on the architecture and model constructed in Step 1, derive the communication and sensing performance metrics under finite blocklength, including the communication transmission delay, the decoding error probability, and the Cramér-Rao bound of the sensing accuracy; Step 3: Based on the performance metrics derived in Step 2, use the non-dominated sorting genetic algorithm to jointly optimize the communication delay and the sensing accuracy for multi-objective optimization, and determine the feasible region of the blocklength that meets the basic performance of communication and sensing; Step 4: Based on the feasible region of the blocklength obtained in Step 3, design an adaptive blocklength optimization strategy based on a double deep Q-learning network, and achieve the coordinated improvement of communication and sensing performance by dynamically adjusting the blocklength.

2. The method according to claim 1, wherein The specific content of Step 1 includes: Step 101, based on the cell-free communication and sensing integrated architecture, establish a system scenario model including M distributed access points, K communication users, and 1 sensing target, where the access points are divided into a downlink access point set and an uplink access point set according to the duplex mode, and the users are divided into a downlink user set and an uplink user set; Step 102, based on the system scenario model, adopt the quasi-static channel hypothesis, establish a channel model including large-scale path loss and small-scale Rayleigh fading for the communication channel, and establish a Swerling type I model for the sensing channel; Step 103, based on the channel model, design a pilot-assisted minimum mean square error channel estimation scheme for obtaining the communication channel state information, and establish a sensing channel reconstruction mechanism; Step 104, based on the channel estimation scheme, construct a composite signal model including communication signals and sensing signals transmitted by the downlink access points, and a received signal model for the uplink and downlink links, where the received signal model includes desired signals and various interference terms.

3. The method according to claim 2, wherein When deriving the communication and sensing performance metrics in Step 2, it specifically includes: Based on the constructed received signal model, derive the closed-form expressions of the signal-to-interference-plus-noise ratio for the uplink and downlink links; According to the signal-to-interference-plus-noise ratio expression, define the mathematical relationship between the transmission delay and the decoding error probability; Based on the sensing channel model, calculate the Cramér-Rao bound of the sensing accuracy.

4. The method according to claim 1, wherein The multi-objective joint optimization in Step 3 specifically includes: Taking the transmission delay and the Cramér-Rao bound defined in Step 2 as the objectives, construct a multi-objective optimization problem; Use the non-dominated sorting genetic algorithm, combined with the crowding distance operator, to solve the Pareto front solution set; Generate the feasible region of the blocklength design according to the Pareto solution set and the basic requirements of the communication and sensing performance.

5. The method according to claim 1, characterized in that The adaptive blocklength optimization strategy in Step 4 specifically includes: Based on the feasible region of the blocklength obtained in Step 3, define the state space and the action space of the double deep Q-learning network; Design a joint reward function, integrating the reliability constraint penalty term derived from the communication delay, the sensing accuracy, and the upper bound of the decoding error probability defined in Step 2; Train the network through experience replay and temporal difference targets, and output the adaptive blocklength strategy.

6. A cell-free communication and sensing integrated system, characterized in that, It includes: A distributed access point array for cooperatively transmitting communication signals and sensing signals; A central processing unit for executing the method according to any one of claims 1-5 to achieve dynamic blocklength optimization; A channel estimation module, which obtains communication channel state information based on a pilot sequence; A sensing interference cancellation module, which is used to reconstruct a sensing channel and suppress communication signal interference.

7. The system according to claim 6, characterized in that, The access point adopts a uniform linear array antenna with an element spacing of half a wavelength, and supports beamforming of downlink communication and active sensing signals.

8. Use of the method according to any one of claims 1-5 in the field of autonomous driving, characterized in that, By dynamically adjusting the block length, the requirements of low-latency communication and high-precision environment perception are simultaneously met in the scenario of high-speed vehicle movement.

9. Use of the method according to any one of claims 1-5 in the industrial Internet of Things, characterized in that, It is used for collaborative optimization of reliable data transmission and device positioning in complex multi-device environments.

10. A computer-readable storage medium storing a computer program, characterized in that, When the program is executed by a processor, it implements the steps of the method according to any one of claims 1-5.

Images