ISAC system robust transceiver design based on product complex sphere manifold method

By adopting product complex spherical manifold method and parallel finite BFGS method in ISAC system, the problems of high computational complexity and poor robustness in ISAC system are solved, and efficient perception and communication performance improvement are achieved to adapt to uncertain conditions. robust design.

CN118984165AInactive Publication Date: 2024-11-19UNIV OF ELECTRONICS SCI & TECH OF CHINA +1

Patent Information

Application Number
CN202411044149.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

When the existing ISAC system transceiver design realizes the coexistence and sharing of radar and communication, there are problems such as high computational complexity, poor robustness, large hardware resource consumption and unstable communication performance, especially in uncertain targets and clutter. When sexuality increases, perception decreases.

Method used

Using the robust transceiver design of ISAC system based on product complex spherical manifold method, by constructing optimization models and converting them into unconstrained fractional planning problems, the transmission waveform and filter are synchronously optimized by parallel finite BFGS method to directly maximize the communication rate and Perceived signal interference plus noise ratio reduces computational complexity and improves robustness.

Benefits of technology

It realizes the improvement of radar-sensing signal interference plus noise ratio and communication rate under low processing complexity, while improving the robustness and communication performance of the system, reducing hardware costs, and maintaining good perception capabilities under uncertain conditions. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118984165A_ABST
    Figure CN118984165A_ABST
Patent Text Reader

Abstract

The invention discloses an ISAC system robust transceiver design based on a product complex sphere manifold method, and belongs to the field of waveform design. According to the invention, a robust ISAC transceiver design model is adopted to solve the problem of insufficient prior sensing information during design of a constant modulus constrained transceiver; the model complies with a CMC for a transmit code and a spherical constraint for a receive filter to jointly maximize an average signal-to-interference plus noise ratio for sensing and an achievable sum rate for communication; the method has the advantages that the method is simple and easy to implement, the constructed model is optimized and solved on the basis of the product complex circle spherical manifold method without relaxation and matrix inversion so as to achieve the optimization target of parallel optimization of transmitting codes and receiving filters, and the complexity of optimization processing is remarkably reduced while the sensing SINR and the communication ASR are improved; compared with a non-robust design, the sensing SINR and the communication ASR are both improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of radar technology, and in particular relates to a robust transceiver design method for an ISAC system based on a product complex sphere manifold method. Background Art

[0002] Transceiver design with constant modulus constraint (CMC) for multiple-input multiple-output (MIMO) integrated sensing and communication (ISAC) systems is a key technology. ISAC systems unify sensing and communication onto a single hardware platform sharing the same frequency band, seeking the performance trade-off and mutual benefit between them. In addition, CMC is enforced on the transmission code to improve the utilization efficiency of the transmitter and facilitate practical applications. Therefore, extensive research has been conducted to solve this problem.

[0003] Existing ISAC transceiver design research is mainly non-robust design, which can be divided into two categories. The first category involves the coexistence design of sensing and communication, designing the sensing coding and communication codebook separately. The second category is the shared design of sensing and communication, in which radar and communication share the same code.

[0004] In the first category, generally, the scheme disclosed in the literature "M. Deng, Z. Cheng, and Z. He, "Spectrally compatible waveform design for large-scale mimo radar beampattern synthesiswith one-bit dacs," IEEE Transactions on Aerospace and Electronic Systems, vol. 58, no. 5, pp. 4729–4744, 2022." aims to maximize the sensing signal-to-interference-noise ratio (SINR) under the communication rate constraint to achieve the coexistence of radar and communication, and proposes an iterative alternating maximization (IAM) method based on the second-order cone program (SOCP) to solve the resulting problem. However, the IAM method is computationally expensive and relaxes the CMC. In order to reduce the computational complexity without CMC relaxation, the paper "Q. Wu, J. Xu, Y. Zeng, DWK Ng, N. Al-Dhahir, R. Schober, and ALS windlehurst, "A comprehensive overview on 5g-and-beyond networks with uavs: From communications to sensing and intelligence," IEEE Journal on Selected Areas in Communications, vol. 39, no. 10, pp. 2912–2945, 2021." proposes an alternating direction multiplier (ADMM) method to simultaneously maximize radar SINR and average communication rate. In addition, in the scheme disclosed in the document "X. Yu, X. Yao, J. Yang, L. Zhang, L. Kong, and G. Cui, "Integrated waveform design for mimo radar and communication via spatiospectral modulation," IEEE Transactions on Signal Processing, vol. 70, pp. 2293–2305, 2022.", the ADMM method is used to mitigate the mutual interference between the frequency diversity array (FDA) sensing and MIMO communication systems. However, the convergence of the ADMM method is proved to be sensitive to the initial parameters.To solve this problem, the paper "P.Kumari, SAVorobyov, and RWHeath, "Adaptive virtual waveform design for millimeter-wave joint communication–radar," IEEE Transactions on Signal Processing, vol.68, pp.715–730, 2020." proposed a new block fading (BF) method to jointly design the sensing code and communication codebook by converting the original problem into an element-by-element phase angle problem. However, the BF method is challenging to handle long sequence design.

[0005] The co-existence design of sensing and communication consumes more ISAC hardware platform resources because it requires separate transmitters and allocates power for both radar code and communication codebooks. In order to reduce hardware cost, the shared design of sensing and communication has attracted much attention.

[0006] In the second category, in the scheme disclosed in the document "C. Ding, J.-B. Wang, H. Zhang, M. Lin, and GY Li, "Joint mimoprecoding and computation resource allocation for dual-function radar and communication systems with mobile edge computing," IEEE Journal on Selected Areas in Communications, vol. 40, no. 7, pp. 2085–2102, 2022.", the branch and bound (BAB) method is used in the weights to minimize the mean square error (MSE) of the designed sensing beam pattern and the required single-user and communication multi-user interference (MUI), aiming to achieve sensing and communication sharing. However, due to the limited degrees of freedom (DOF) of the sensing beam pattern MSE, balancing sensing and communication performance is challenging. To this end, it is considered to simultaneously maximize the sensing SINR and minimize the communication MUI to achieve radar communication sharing, among which the typical method is gradient projection (GP). However, due to the limited degrees of freedom for minimizing MUI, the communication quality of service (QoS) is reduced. To address this problem, a nonlinear equality constrained alternating direction multiplier method (neADMM) is proposed, which uses constructive interference-based symbol-level precoding (SLP) to obtain better communication QoS to maximize the sensing SINR. However, as the uncertainty of the target and clutter prior information increases, the robust sensing capability decreases. Summary of the invention

[0007] The object of the present invention is to provide a robust transceiver design for an ISAC system based on a product complex sphere manifold method, which can achieve simultaneous improvement of the perceived signal to interference plus noise ratio (SINR) and the communication sum rate (ASR) with low processing complexity.

[0008] The technical solution adopted by the present invention is:

[0009] A robust transceiver design for an ISAC system based on a product complex sphere manifold method includes the following steps:

[0010] Step S1, constructing an optimization model of a robust transceiver of an ISAC system:

[0011]

[0012] Among them, SINR r (s,w) represents the radar-perceived signal-to-interference-plus-noise ratio SINR for the transceiver’s transmit waveform s and filter w, is the upper limit of radar SINR, ASR represents the communication rate of ISAC system, is the upper limit of ASR, the weighting factor ρ∈[0,1] is used to adjust the ratio of radar and communication, f t is the normalized Doppler frequency of the target, θ t is the target arrival direction, f c,p,k is the normalized Doppler frequency of clutter patch k in range bin p, θ c,p,k is the arrival direction of the clutter plate k in the range bin p, represents a uniformly distributed random variable, and f t ,θ t ,f c,p,k and θ c,p,k The prior mean of t ,ψ t ,φ c,p,k and ψ c,p,k Respectively represent f t ,θ t ,f c,p,k and θ c,p,k The uncertainty value, s n (l) represents the transmission waveform of the transmitting antenna n of the transceiver at symbol time l, N t is the number of transmitting antennas of the transceiver, L is the symbol length of the transmitted waveform s;

[0013] Step S2, obtain the probabilistic conjugate complex projection shadow manifold based on the feasible domain of the transmission waveform s and the filter w Convert the optimization model constructed in step S1 into The unconstrained fractional programming problem on the transmit waveform s and the filter w is solved synchronously, and the optimal transmit waveform s and the optimal filter w of the robust transceiver of the ISAC system are obtained based on the optimization results;

[0014] in, M s and M w are the feasible domains of the transmitted waveform s and the filter w, respectively, and their expressions are: Where i is the element index, M is the number of pulses sent by the transceiver within the coherent processing interval device, and N r is the number of receiving antennas of the transceiver.

[0015] Furthermore, in step S2, the converted unconstrained fractional programming problem is iteratively optimized and solved based on the Riemann gradient and the BFGS direction.

[0016] Furthermore, step S2 specifically includes:

[0017] S2-1, project the Euclidean space into the tangent space of the flow, and calculate the parallel Riemann gradient grad of the emission waveform s based on the Euclidean gradient s f(s,w), the parallel Riemann gradient grad of filter w w f(s,w);

[0018] S2-2, based on the parallel Riemann gradient of the transmit waveform s and filter w obtained in the current iteration, obtain the BFGS direction of the current iteration Among them, k is used to identify the number of iterations;

[0019] S2-3, based on BFGS direction exist Update the feasible solution of the transmission waveform s and the filter w, and obtain the updated results of the transmission waveform s and the filter w for the k+1th iteration in, is the updated result for the kth iteration (s k ,w k ) remapping operation;

[0020] S2-4, update the limit iteration intermediate variable (δ,P s ,P w ,Q s ,Q w ), check whether the preset iterative convergence condition is met. If so, use the most recently obtained update result (s k+1 ,w k+1) as the optimal transmit waveform s and the optimal filter w; otherwise, steps S2-1 to S2-4 are repeatedly executed based on the updated limit iteration intermediate variables.

[0021] Furthermore, the iterative convergence condition is set as: the cost function value corresponding to the emission waveform s and the filter w obtained by the last two updates (such as )convergence.

[0022] The technical solution provided by the present invention brings at least the following beneficial effects:

[0023] The present invention aims at the problem of robust transceiver design of CMC and formulates it as the joint maximization of the average SINR for sensing and the achievable sum rate (ASR) for communication, wherein the transmission code is constrained by CMC and the receiving filter is constrained by spherical constraints (SPC); due to CMC and the coupling between the transmission code and the receiving filter, the resulting problem is non-convex and NP-hard. In order to solve this problem, the present invention solves the problem based on the proposed product complex spherical manifold (PCCSM) method without relaxation and matrix inversion. First, the natural characteristics of CMC and SPC are used to construct a PCCSM that meets the requirements of CMC and SPC, and then the problem is converted into an unconstrained coupled fractional programming problem on PCCSM. Finally, a parallel finite BFGS (PL-BFGS) method is derived to optimize the transmission code and the receiving filter in parallel. Compared with the existing processing methods, the main technical advantages of the present invention are reflected in:

[0024] (1) Focus on robust transceiver design, which is different from non-robust transceivers, improving sensing SINR and communication ASR;

[0025] (2) Different from indirectly minimizing MUI, the present invention directly maximizes ASR to bring better communication performance;

[0026] (3) Different from the matrix inversion relaxation method, the present invention derives a PCCSM method without relaxation and matrix inversion, which has better performance and lower computational complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1It is a comparison of the convergence performance of the ISAC system robust transceiver design based on the product complex sphere manifold method (referred to as the PL-BFGS method) proposed in the embodiment of the present invention and the general MM method and the GP method;

[0029] Figure 2 is the space-time plane of CAF of different algorithms, where (a) is radar only (i.e., weighting factor ρ = 1), (b) is the PL-BFGS method with ρ = 0.5, (c) is the general MM method with ρ = 0.003, and (d) is the GP method with ρ = 0.0015;

[0030] Figure 3 The relationship between the communication ASR and SNR of the PL-BFGS method, the general MM method, and the GP method proposed in the embodiment of the present invention;

[0031] Figure 4 is the relationship between the radar SINR and the communication ASR and the weighting factor ρ of the PL-BFGS method proposed in the embodiment of the present invention, where the communication system SNR is 25 dB;

[0032] Figure 5 : is the relationship between the space-time plane of CAF and the weighting factor ρ of the PL-BFGS method proposed in the embodiment of the present invention, wherein ρ=0.01 in (a), ρ=0.2 in (b), ρ=0.4 in (c), ρ=0.6 in (d), ρ=0.8 in (e), and ρ=0.99 in (f);

[0033] Figure 6 is the relationship between the communication ASR and the weighting factor ρ of the PL-BFGS method proposed in the embodiment of the present invention, where the communication system SNR is 25 dB;

[0034] Figure 7 Radar SINR and communication ASR with different target uncertainty values ​​(φ) when using or not using robust ISAC transceiver design t ,ψ t ) relationship diagram;

[0035] Figure 8 Radar SINR and communication ASR with and without robust ISAC transceiver design versus different clutter uncertainty values ​​(φ c ,ψ c ) relationship diagram. DETAILED DESCRIPTION

[0036] To make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be described in detail and completely in conjunction with the drawings in the implementation of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings can be arranged and designed using different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present invention.

[0037] The embodiment of the present invention provides a robust transceiver design for an ISAC system based on a product complex circle-sphere manifold method, solves non-convex problems based on the proposed product complex circle-sphere manifold method (PCCSM) structure, and proposes a Parallel Limited-Broyden-Fletcher-Goldfarb-Shanno (PL-BFGS) method with low computational complexity in the PCCSM structure to design a transmission waveform.

[0038] For the convenience of description, the following definitions are given first:

[0039] The transpose, conjugate, and conjugate transpose operations are respectively represented by (·) T ,(·) * ,(·) H express; Represent the complex number domain and the real number domain respectively; diag([x 1 ,x 2 ,…,x N ]) represents a diagonal matrix, where the diagonal elements are x 1 ,x 2 ,…,x N ; represents the expectation of a random variable; tr{·},vec(·),(·) -1 Represent the trace, column vectorization and inverse of the matrix respectively; Represent the real and imaginary parts of a complex-valued scalar / vector / matrix, respectively; ||·|| F ,|·| respectively represent the matrix binorm and the vector element-wise modulus; ⊙ represents the Hadamard product; the matrices I,1(0) respectively represent the identity matrix and the matrix of all 1s(0).

[0040] As a possible implementation, the ISAC system robust transceiver design based on the product complex sphere manifold method provided in an embodiment of the present invention specifically includes the following steps:

[0041] System Model and Problem Construction:

[0042] In the embodiment of the present invention, for the airborne MIMO-STAP radar model, the airborne co-sited MIMO space-time adaptive processing (STAP) radar is used as an example for description. The radar transmits data at a pulse repetition frequency (PRF) f during a coherent processing interval (CPI). r M pulse trains are sent, and the waveform of each pulse is the same. Symbol time l comes from N t The transmitting waveform of the transmitting antenna is expressed as Then, the transmit waveform matrix in units of L symbol time can be expressed as:

[0043]

[0044] Total space time receiving echo signal It can be expressed as:

[0045]

[0046] Where s = vec(S), i.e., the column vector of the transmit waveform matrix;

[0047] α t is a complex-valued factor that takes into account antenna transmit power, antenna gain, target radar cross section, and other relevant factors;

[0048] in is the time rotation vector, f t =2v t / (λf r ) is the normalized Doppler frequency of the target, v t is the target's speed relative to the radar platform, λ is the wavelength, A(θ t )=b(θ t ) T (θ t ), θ t is the target direction of arrival (DOA), the guidance vector a(θ t ) and b(θ t ) is:

[0049]

[0050] Where a(θ t ) and b(θ t ) are the transmitting and receiving spatial steering vectors respectively. t d r are the receiving antenna spacing and the transmitting antenna spacing, respectively.

[0051] The airborne radar system is subject to signal-dependent interference (clutter) formed by collecting echoes from 2P+1 range bins adjacent to the cell under test (CUT), and each range bin includes N c clutter patches, α c,p,k ,θ c,p,k and f c,p,k =2V a sin(θ c,p,k ) / (λf r ) are the complex amplitude, DOA and normalized Doppler frequency of the kth clutter plate in the pth range bin, respectively. in is the shift matrix:

[0052]

[0053] Specifically, for ease of expression, let u(f c,p,k ),A(θ c,p,k ),C(f c,p,k ,θ c,p,k ) is represented by u c,p,k ,A c,p,k ,C c,p,k ;

[0054] is additive white Gaussian noise (AWGN), which follows distributed.

[0055] In practice, the exact f t ,θ t ,f c,p,k and θ c,p,k The prior knowledge of is not available. Therefore, the embodiment of the present invention considers a robust design and assumes that f t ,θ t ,f c,p,k and θ c,p,k are independent uniformly distributed random variables with corresponding prior means and Right now:

[0056]

[0057] Among them, φ t ,ψ t ,φ c,p,k and ψ c,p,k Respectively represent f t ,θ t ,f c,p,k and θ c,p,k The uncertainty value of .

[0058] With filter The output SINR of the airborne MIMO-STAP radar can be expressed as:

[0059]

[0060] Among them, w represents the weight vector of the filter, that is, the weight vector of the transmitting end of the MIMO-STAP radar, α t and α c,p,k are uncorrelated zero-mean random variables with variance and Furthermore, based on statistical algebra operations, the average output SINR can be rewritten as:

[0061]

[0062] in

[0063]

[0064] in

[0065] Communication model:

[0066] Assume that the number of downlink users U is served by the ISAC base station (BS) at the communication end. Then, the signal received by the communication user in L symbol time is given by

[0067] Y=HS+Z (9)

[0068] in, is the received signal matrix, is the Rayleigh distribution channel matrix between BS (base station) and user, U is the number of users, Z = [z 1 ,...,z L ] is the AWGN matrix variable, i.e., the communication system noise, where and represents the noise variance.

[0069] Let x(l) = [x 1 (l),....,x U (l)] T ∈Ω U The symbol Ω representing the desired symbol of a communication user at symbol time l is a set of communication constellation points independently selected from the phase shift keying (PSK) constellation. To improve communication performance, minimizing the so-called "multi-user interference (MUI) energy" is widely used, which can be written as

[0070]

[0071] in In addition, the received signal of the u-th user at symbol time l can be expressed as

[0072]

[0073] where z u (l) is the corresponding noise.

[0074] Then, the SINR of the uth communicating user is defined as:

[0075]

[0076] in, is the noise power of the u-th communication user.

[0077] Then, the achievable sum rate (ASR) of the communication model can be expressed as

[0078]

[0079] Typically, the metric of minimizing MUI is widely used in ISAC transceiver design, where the problem is modeled as

[0080]

[0081] where "constraints" contains the CM or similar constraints under consideration, and let J(s,w) denote the cost function for simplicity. The optimization problem given in formula (14) can be solved using the alternating optimization (AO) method, i.e., iteratively using matrix inversion to obtain w, and using the gradient projection (GP) method and the general optimization minimization (MM) method to solve for s.

[0082] Problem construction:

[0083] A robust ISAC transceiver design is constructed by designing the transmit waveform s and the filter w to maximize both radar SINR and communication ASR, which can be written as:

[0084]

[0085] Among them, ρ∈[0,1] is the weighting factor used to adjust the ratio of radar and communication, and are the theoretical upper limits of radar SINR and communication ASR, respectively. In particular, without loss of generality, the constraint ||w|| 2 =1.

[0086] It is worth noting that the ISAC transceiver design in the embodiment of the present invention has the following improvements:

[0087] (i) It considers the robust transceiver design under the inaccurate perceptual prior information (i.e., the constraints in (15)), which is more application-oriented;

[0088] (ii) Different from minimizing MUI indirectly, directly maximizing ASR can bring better communication performance. In addition, the normalization of SINR is introduced, and SINR and ASR reduce the difficulty of selecting the weight factor ρ∈[0,1].

[0089] Due to the CM constraints, the resulting problem is complex and non-convex, which prevents a direct closed-form solution. Typically, the usual matrix inversion for optimizing the filter w is accompanied by matrix inversion and singular value decomposition, which is O((MN r L) 3 ), which is time-consuming. To address this problem, an embodiment of the present invention proposes a solution method based on the product complex spherical manifold method, which achieves better performance at a lower computational cost.

[0090] Preferably, when solving the coupled non-convex fractional programming problem shown in formula (15), the embodiment of the present invention first constructs the PCCSM as a solution space that accommodates the transmission waveform s and the filter w; then, the problem is converted into a UCFP problem (unconstrained fractional programming problem) on the PCCSM structure; finally, the PL-BFGS method for synchronously optimizing s and w is derived.

[0091] In general, constrained optimization problems can be transformed into unconstrained optimization problems on a specially designed manifold space that contains all constraints. The feasible domain of the transmit waveform s and the filter w in formula (15) is equivalent to the product of the complex circle and the complex sphere manifold, where the size of the PCCSM is N t L+LMN r , expressed as:

[0092]

[0093] Among them, M s and M w are the feasible domains of the transmitted waveform s and the filter w, respectively. Specifically, M s and M w is defined as follows:

[0094]

[0095] Near each point, the smooth manifold can be locally linearized. The linearization at is called the tangent space and is defined as follows:

[0096]

[0097] Among them, the tangent vector ζ(s,w) ∈T (s,w) Represents any point on the smooth curve c The moving direction of the probability conjugate complex shadow manifold (PCCSM) It can be decomposed into the product of two tangent spaces:

[0098]

[0099] In s∈M s and w∈M w The tangent spaces at are:

[0100]

[0101] Among them, s and w are the tangent vectors at s and w respectively, then the point The tangent space at can be expressed as:

[0102]

[0103] In order to obtain The cost function gradient on needs to be selected A convenient choice is to assign The induced Riemannian metric transforms it into Euclidean space Therefore, the embodiment of the present invention uses the normal complex Euclidean inner product to define each The measure of is:

[0104]

[0105] in, Used to indicate two different points;

[0106] Based on the above metrics, PCCSM becomes a Euclidean space is a Riemann submanifold of .

[0107] In order to add two vectors in different tangent spaces, the transport operation Trans(·)←(·) is introduced to convert the tangent vector ζ (s,w) from Move to another point As given below:

[0108]

[0109] in,

[0110] Preferably, the ISAC system robust transceiver design based on the product complex sphere manifold method (PL-BFGS method) provided in the embodiment of the present invention mainly includes four steps: 1) calculating parallel Riemann gradients; 2) finding BFGS directions; 3) updating feasible solutions; 4) updating limit iteration intermediate variables; each step is specifically:

[0111] 1) Calculate parallel Riemann gradients

[0112] First, the cost function f(s,w) in equation (15) is rewritten as:

[0113]

[0114] in, Channel of user u The received signal x of user u u =[x u (1),...,x u (L)] T , represents the noise of the u-th user;

[0115] Then, the Riemann gradient gradf(s,w) is defined as the tangent space An element on , satisfying:

[0116]

[0117] in is at (s,w) along The directional derivative of the direction can be expressed as:

[0118]

[0119] Where Grad f(s,w) is the Euclidean gradient of f(s,w), Proj (s,w) (Gradf(s,w)) means from Gradf(s,w) to Orthogonal projection of

[0120] Thus we get The Riemann gradient grad f(s,w) on can be expressed as:

[0121]

[0122] in, and Represent the Riemann gradients with respect to s and w respectively.

[0123] The Euclidean gradient of s and w is calculated as follows:

[0124]

[0125] Among them, the parameter t 1 =s H Ξ(w)s, , t 3 =H u sx u , t 5 =w H Θ(s)w,

[0126] Finally, the Riemann gradient of s and w is s f(s,w),grad w f(s,w) is calculated by projecting from Euclidean space to tangent space:

[0127]

[0128] 2) Calculate the BFGS direction:

[0129] At the kth iteration, the BFGS direction is expressed as:

[0130]

[0131] Among them, H sk and H wk is the corresponding Hessian matrix, and the subscript or superscript k is used to indicate the number of iterations.

[0132] Further, define the intermediate variable (ps k ,pw k ) and (qs k ,qw k )for:

[0133]

[0134] To avoid scaling the BFGS direction, the intermediate variables (ps k ,pw k ) and (qs k ,qw k ) for normalization (i.e. ps k =ps k / ||ps k || 2 ,qs k =qs k / ||qs k || 2 ), pw k and qw kThen, the approximate Hessian matrix can be further expressed as:

[0135]

[0136] Among them, the auxiliary I represents the identity matrix.

[0137] The expression of formula (35) has a recursive property, with Hs k+1 Take it as an example, and recursively expand it m times, we can get:.

[0138]

[0139] In fact, the search direction d sk By calculating Hs k grad s f(s k ,w k ) is obtained, namely:

[0140]

[0141] For the sake of simplicity, auxiliary variables are introduced:

[0142]

[0143] You can get:

[0144]

[0145] in, By substitution and replace The above formula can be rewritten as:

[0146]

[0147] in, Then substitute The above formula can be expressed as:

[0148]

[0149] Similarly, substitute pass The above formula can be further expressed as:

[0150]

[0151] Finally, transform the above formula into:

[0152]

[0153] Direction of BFGS It can be obtained by inverting the above formula. Similarly, The calculation is similar.

[0154] 3) Update feasible solution:

[0155] exist On, will (s k ,w k ) is updated to (s k+1 ,w k+1 ) process can be expressed as:

[0156]

[0157] Here α k is the step size factor.

[0158] The remapping operation can be expressed as:

[0159] R (s,w) (ξ (s,w) )=(Ret s (ξ s ),Ret w (ξ w )) (45)

[0160] Right now

[0161] HereRet s (ξ s ) and Ret w (ξ w ) are:

[0162]

[0163] in, Ret s The input parameter ξ of () s Different element values ​​of are different elements of s.

[0164] Usually, the step size α k It is found by the Armijo linear search method, that is, to find a minimum non-negative integer h ≥ 0 that satisfies:

[0165]

[0166] Among them, σ,τ∈(0,1), and <grad;f(s k ,wk),(dsk,dw k )> represents the inner product of corresponding vectors.

[0167] 4) Update the intermediate variables of limit iteration: The PL-BFGS method stores the intermediate variables (i.e. (ps k ,pw k ),(qs k ,qw k ),δ k ), limiting it to the storage space length l m To save memory space and computing resources, we introduce m The storage sequence is as follows:

[0168]

[0169] in, Represents a single storage space that has not yet been stored.

[0170] Usually, the cautious update formula (CUF) is used to update the intermediate variables. First, according to formulas (33) to (34), (ps k ,pw k ) and (qs k ,qw k ); Then, the updating and storage of the intermediate variables depends on the number of expansions m and the length of the storage space l m The relationship between CUF and Given as:

[0171]

[0172] The discrimination threshold of CUF is set as:

[0173]

[0174] Among them, the function is strictly monotonically increasing with respect to t, and In this embodiment, Without loss of generality, we can use the intermediate variable related to s (i.e., ps k 、qs k , δ k ) as an example, the updating and storage rules of intermediate variables are as follows:

[0175] (i) If Then accept this update and according to the expansion times m and storage space length l m The relationship between them can be divided into two cases:

[0176] (i.1): Calculation And define the auxiliary amount for:

[0177]

[0178] (i.2): If m<l m , then the currently updated variable ps is continuously updated k 、qs k , δ k Add to established storage spaces until they are filled.

[0179] First, the existing m elements of the storage variable sequence are transformed into the tangent plane at the new iteration point, that is:

[0180]

[0181] Then, ps k Insert from the right side of the array into an empty position [] m+1 , as shown below:

[0182]

[0183] Similarly, The updates are as follows:

[0184]

[0185] Then qs k Insert into empty position [] m+1 , as shown below:

[0186]

[0187] Finally, the scalar δ k Directly update to:

[0188]

[0189] (i.3): If m ≥ l m , discarding the previously stored data in sequence and storing the newly updated data accordingly, as shown below:

[0190]

[0191] and

[0192]

[0193] Finally, the scalar δ k Directly update to:

[0194]

[0195] (i.4): After executing (i.1), (i.2) and (i.3), set m=m+1.

[0196] (ii): If Reject this update, δ k and The original m storage spaces still need to be converted to the new tangent plane accordingly. The specific steps are as follows:

[0197]

[0198] and

[0199]

[0200] In summary, when optimizing and solving the cost function constructed based on formula (25) to obtain the optimal transmission waveform s and filter w, in the embodiment of the present invention, the specific solution process includes:

[0201] 201, set the initial value of the transmission waveform s and the filter w Search times threshold h max , storage space length l m , parameters σ and τ, and convergence threshold μ; where σ, τ∈(0,1); 202, initialization iteration number k=0, expansion number m=0;

[0202] 203, calculate the emission waveform s at the kth iteration k , filter weight vector w k The Riemann gradient gradf(s k ,w k ), and calculate the auxiliary quantity based on the Riemann gradient

[0203] 204, when |f(s k ,w k )-f(s k-1 ,w k-1 )|<μ, perform the following operations:

[0204] 204-1, based on the BFGS direction acquisition algorithm, obtain the kth iteration transmission waveform s k and filter w k The corresponding search direction

[0205] 204-2, let the search step length of the current iteration be α k =1, and initialize the search times of the current iteration h=0, and update the transmission waveform s and filter w according to formula (44), and obtain the updated result (s k+1 ,w k+1 );

[0206] 204-3, when formula (48) holds and h≤h max, execute step 204-4; otherwise, execute step 204-5;

[0207] 204-4, let α k =τ h α k , and then update (s according to formula (44) k+1 ,w k+1 ), update the search times h=h+1 and continue to execute step 204-3;

[0208] 204-5, based on the currently obtained update result (s k+1 ,w k+1 ) calculates the corresponding Riemann gradient gradf(s k+1 ,w k+1 );

[0209] According to formulas (33) and (34), the two intermediate auxiliary quantities of the filter w of the transmit waveform s are updated respectively: (ps k ,pw k ) and (qs k ,qw k );

[0210] Then update the discriminant factor according to formula (50) and (51): and the discrimination threshold ε k ;

[0211] 204-6, if According to the formula Calculate the intermediate variable δ k , and update the auxiliary amount according to formula (52) After the update is completed, execute step 204-7;

[0212] like Then keep the current intermediate variable δ k and auxiliary amount The data of the current m storage spaces are unchanged, and the data of the current m storage spaces are converted to the new cutting plane accordingly, and the process goes to step 204-8;

[0213] 204-7, if m<l m , then according to formulas (53) to (57) the intermediate variables δ,P are stored s ,P w ,Q s ,Q w ; if m ≥ l m , then, according to formulas (58) to (60), store δ,P s ,P w ,Q s ,Q w ;

[0214] 204-8, update m=m+1, and continue to execute step 204-6;

[0215] 204-9, output the currently calculated search direction

[0216] In the above step 204-1, the direction acquisition algorithm is used to obtain The specific steps include:

[0217] 1) Input: Among them, the parameter t = min (m, l m );

[0218] 2) Initialize auxiliary amount r = 0 t , about the auxiliary variables (ζ s ,ζ w )=gradf(s k ,w k );

[0219] renew

[0220] 3) When the storage space index i=t, t-1, ..., 1 and i>0, perform the following steps:

[0221] 3-1) Order

[0222] 3-2) Update

[0223] 3-3) Calculation After the calculation is completed, go to step 5);

[0224] 4) When i=1,2,...,t, perform the following steps:

[0225] 4-1) Calculation

[0226] 4-2) Update After the calculation is completed, go to step 5);

[0227] 5) Order And output.

[0228] Example

[0229] In order to verify the proposed solution, the waveform is optimized by using the existing solution and the solution proposed in the embodiment of the present invention.

[0230] The simulation experiment conditions are set as follows:

[0231] For the radar system: Consider an airborne co-sited MIMO-STAP radar with Nt = 6 transmitting antennas and N r = 6 receiving antennas. The number of pulses during CPI is M p =8, the number of symbols is L = 6. The wavelength is λ = 0.15, and the spacing between array elements is d t =λ / 2 and d r =λ / 2. The platform speed is V a =150m / s, PRF is f r =4000Hz. Without loss of generality, only one range bin adjacent to the CUT is considered, i.e. P = 0. The DOA of the target is The normalized Doppler frequency is The signal-to-noise ratio (SNR) is 10dB, the clutter-to-noise ratio (CNR) is 30dB, and the power of the complex AWGN is And suppose there exists N c = 181 clutter patches, and average Evenly distributed from 0° to 180°. Then, the average normalized Doppler frequency Accordingly In addition, we also considered

[0232] For the communication system: Assume that the number of downlink users is U=2, and randomly extract the symbol matrix X required by the communication users from the orthogonal PSK (QPSK) constellation. The signal-to-noise ratio of the communication system is 10 dB.

[0233] The existing solutions involved in this embodiment are:

[0234] 1) ISAC robust transceiver design using GP method: Reference: CGTsinos, A.Arora, S.Chatzinotas, and B.Ottersten, "Joint transmit waveform and receive filter design for dual-function radar communication systems," IEEE Journal of Selected Topics in Signal Processing, vol. 15, no. 6, pp. 1378–1392, 2021.

[0235] 2) ISAC robust transceiver design using MM method: Reference: L.Wu, P.Babu, and D.P.Palomar, "Transmit waveform / receive filter design for mimo radar with multiple waveform constraints," IEEE Transactions on Signal Processing, vol.66, no.6, pp.1526–1540, 2018.

[0236] Figure 1 The residual term of the method proposed in the embodiment of the present invention (referred to as PL-BFGS method, corresponding to the proposed shown in the figure) is shown The convergence curve of the outer layer iteration number. It can be seen that as the outer layer iteration number increases to 30 times, the residual term Finally, it converges to 7.029e-05, meeting the iteration stop condition.

[0237] Figure 2 The relationship between radar SINR and the number of iterations of the PL-BFGS method proposed in the embodiment of the present invention and the general MM method and GP method is shown. The results in the figure show that the PL-BFGS method proposed in the embodiment of the present invention achieves a higher radar SINR value at a more efficient rate compared with the general MM method and GP method. Generally, the radar SINR obtained when the proposed PL-BFGS method converges is 0.8865dB and 1.6193dB higher than that of the general MM method and GP method, respectively.

[0238] Figure 3 The relationship between the communication ASR and SNR of the proposed PL-BFGS method and the general MM method and GP method is shown. The results in the figure show that the proposed PL-BFGS method achieves a higher communication ASR. Generally, when the communication system SNR is 29dB, the ASR obtained by the proposed PL-BFGS method is 0.3415bits / s / Hz and 0.4827bits / s / Hz higher than that of the general MM and GP methods, respectively.

[0239] Figure 4 The relationship between radar SINR and communication ASR of the proposed PL-BFGS method and the weighting factor ρ is shown. The results show that a controllable trade-off between radar and communication performance can be achieved by adjusting the weighting factor from ρ = 0.01 to ρ = 0.99. That is, the expected ISAC transceiver design effect can be achieved under different weight ratios. Correspondingly,

[0240] Figure 5 The space-time plane of CAF is plotted against the weighting factor ρ. Figure 6 The relationship between the communication ASR and the weighting factor is plotted. It can be concluded that with the increase of ρ, the radar detection performance is significantly improved, and the communication ASR is reduced accordingly, that is, the weighting factor ρ can effectively control the radar and communication performance to meet the requirements.

[0241] Figure 7 It is shown that both the robust ISAC transceiver design and the non-robust ISAC transceiver design show that the output radar SINR increases with the target uncertainty value (φ t ,ψ t ) increases, and the communication ASR is stable within a certain range, which is as expected. However, the radar SINR of the robust ISAC coding can achieve higher SINR gain compared with the non-robust ISAC transceiver design. Generally, when ρ = 0.9 and (φ t =0.2,ψ t =2), the radar SINR of the robust ISAC transceiver design is about 0.19 dB higher than that of the non-robust design. Therefore, the proposed robust ISAC transceiver design using PL-BFGS method is robust to target uncertainty compared to the non-robust ISAC transceiver.

[0242] Figure 8 The radar SINR and communication ASR are shown with different clutter uncertainty values ​​(φ c ,ψ c ), regardless of whether the robust ISAC transceiver design using the proposed PL-BFGS method is used. Similar to the previous section, both the robust ISAC transceiver design and the non-robust ISAC transceiver design show that the output radar SINR increases with the target uncertainty value (φ c ,ψ c ) increases, and the communication ASR is stable within a certain range. However, it can be clearly seen that with the increase of the clutter uncertainty value (φ c ,ψ c ), the output radar SINR of the robust ISAC transceiver design is much better than that of the non-robust ISAC code. c =0.15,ψ c =1.5), the radar SINR of the robust ISAC transceiver design is about 13.836 dB higher than that of the non-robust design. However, when ρ = 0.6, the communication ASR also has a certain growth trend (i.e., in (φ c =0.2,ψ c=2), which is about 0.613 bits / s / Hz, which should be due to the increase in communication ratio caused by the decrease in radar SINR. Overall, the proposed robust ISAC transceiver design using PL-BFGS method also shows robustness to clutter uncertainty compared to the non-robust ISAC transceiver.

[0243] Through the above simulations, the performance advantages of the ISAC system robust transceiver design based on the product complex sphere manifold method disclosed in the embodiment of the present invention for the ISAC system robust transceiver system are verified: on the one hand, compared with the existing method, the embodiment of the present invention obtains better perception and communication performance with less than half the computational complexity; on the other hand, compared with the non-robust design, the embodiment of the present invention has better robustness in scenarios with uncertain prior information.

[0244] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0245] The above are only some embodiments of the present invention. For those skilled in the art, several modifications and improvements can be made without departing from the creative concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. Design of robust transceiver for ISAC system based on product complex sphere manifold method, characterized by: The following steps are involved: Step S1, constructing an optimization model of a robust transceiver of an ISAC system: Among them, SINR r (s,w) represents the radar-perceived signal-to-interference-plus-noise ratio SINR for the transceiver’s transmit waveform s and filter w, is the upper limit of radar SINR, ASR represents the communication rate of ISAC system, ASR * is the upper limit of ASR, the weighting factor ρ∈[0,1] is used to adjust the ratio of radar and communication, f t is the normalized Doppler frequency of the target, θ t is the target arrival direction, f c,p,k is the normalized Doppler frequency of clutter patch k in range bin p, θ c,p,k is the arrival direction of the clutter plate k in range bin p, represents a uniformly distributed random variable, and f t ,θ t ,f c,p,k and θ c,p,k The prior mean of t ,ψ t ,φ c,p,k and ψ c,p,k Respectively represent f t ,θ t ,f c,p,k and θ c,p,k The uncertainty value, s n (l) represents the transmission waveform of the transmitting antenna n of the transceiver at symbol time l, N t is the number of transmitting antennas of the transceiver, L is the symbol length of the transmitted waveform s; Step S2, obtain the probabilistic conjugate complex projection shadow manifold based on the feasible domain of the transmission waveform s and the filter w Convert the optimization model constructed in step S1 into The unconstrained fractional programming problem on the transmit waveform s and the filter w is solved synchronously, and the optimal transmit waveform s and the optimal filter w of the robust transceiver of the ISAC system are obtained based on the optimization results; in, M s and M w are the feasible domains of the transmitted waveform s and the filter w, respectively, and their expressions are: Where i is the element index, M is the number of pulses sent by the transceiver within the coherent processing interval device, and N r is the number of receiving antennas of the transceiver.

2. The ISAC system robust transceiver design based on the product complex sphere manifold method as claimed in claim 1, characterized in that: In step S2, the converted unconstrained fractional programming problem is iteratively optimized and solved based on the Riemann gradient and the search direction.

3. The ISAC system robust transceiver design based on the product complex sphere manifold method as claimed in claim 2, characterized in that: Step S2 specifically includes: S2-1, project the Euclidean space into the tangent space of the flow, and calculate the parallel Riemann gradient grad of the emission waveform s based on the Euclidean gradient s f(s,w), the parallel Riemann gradient grad of filter w w f(s,w); S2-2, based on the parallel Riemann gradient of the transmitted waveform s and the filter w obtained in the current iteration, obtain the search direction of the current iteration Among them, k is used to identify the number of iterations; S2-3, based on search direction exist Update the feasible solution of the transmission waveform s and the filter w, and obtain the updated results of the transmission waveform s and the filter w for the k+1th iteration in, is the updated result for the kth iteration (s k ,w k ) remapping operation, α k is the step size factor of the current iteration; S2-4, after updating the limit iteration intermediate variable, check whether the preset iteration convergence condition is met. If so, use the most recently obtained update result (s k+1 ,w k+1 ) as the optimal transmit waveform s and the optimal filter w; otherwise, steps S2-1 to S2-4 are repeatedly executed based on the updated limit iteration intermediate variables.

4. The ISAC system robust transceiver design based on the product complex sphere manifold method as claimed in claim 3, characterized in that: The iterative convergence condition is set as follows: the cost function values ​​corresponding to the transmitted waveform s and the filter w obtained by the two most recent updates converge.

5. The ISAC system robust transceiver design based on the product complex sphere manifold method as claimed in claim 5, characterized in that: The cost function is: in, U is the number of users, and the channel matrix of user u is I L is an L-dimensional unit vector, h u is the Rayleigh distribution channel of user u, the intermediate quantity Intermediate Intermediate is the expectation of the random variable, C(f c,p,k ,θ c,p,k ) is the normalized Doppler frequency, represents the variance, Time Turning Vector A(θ t )=b(θ t ) T (θ t ), a(θ t ) and b(θ t ) are the transmitting and receiving spatial steering vectors respectively, The received signal x of user u u =[x u (1),...,x u (L)] T , represents the noise of the u-th user.

Citation Information

Patent Citations

  • MIMO radar waveform optimization method for rapid manifold modeling

    CN113642181A

  • RIS-assisted backscatter communication perception integration method

    CN116248173A

  • Combined beam forming design method for hybrid RIS-assisted millimeter wave ISAC system

    CN118118072A

Cited By

  • Large deformation plate state parameter cross convergence mapping causal test method and system

    CN119249920A

  • A method and system for cross-convergence mapping causal test of large deformation plate state parameters

    CN119249920B