Method for spatio-temporal adaptive signal processing in a monopulse shipborne radar with an active phased antenna array

The invention relates to the field of radar, specifically to a method for space-time adaptive signal processing in a monopulse shipborne radar station (radar) with an active phased antenna array (AFAR) and can be used in a wide class of radars with an AFAR installed on mobile platforms for various purposes.

Earlier / 1-14 / in the synthesis of radar signal detectors it was believed that the signal and interference arriving at the input of the radar receiver from the antenna output are functions of the only variable - time.

In reality / 12 / a radar signal is an electromagnetic wave that depends on both time and space coordinates. This dependence is of particular importance when receiving APAR signals. The presence of spatial parameters of the target signal and their difference from similar parameters of interference allows for effective selection (isolation) of signals against the background of interference and significantly improve the quality of detection of useful radar signals. The task of spatio-temporal processing is to search for algorithms and architectures that implement them for optimal processing of spatio-temporal signals against the background of interference.

With the Bayesian procedure for the synthesis of the optimal processing weight vector, associated with inversion of the noise covariance matrix, the algorithm becomes significantly unstable to the errors of the calculations performed. This circumstance forces us to look for alternative filtering methods that do not require inversion of ill-conditioned matrices.

The search for new processing algorithms is often associated with a lack of a priori information about the statistical characteristics of noise and signals. This situation occurs when the temporal processing of the radar signal against the background of its own noise and passive external interference generated by reflections from the earth, sea surface and various meteorological formations. The spectral-correlation properties of these noises are, as a rule, known only approximately, as a result of which one has to resort to their more or less realistic approximation and then, on its basis, carry out the synthesis of processing. The quality of the process filtration obtained in this way directly depends on the degree of closeness of its structure to the Bayesian algorithm in the presence of complete statistical information about the signal-interference environment.

One of the most promising directions for solving the problem of protecting reception channels from combined interference is the formation of deep dips (zeros) in the radiation pattern (DP) of the radar antenna.

Direct adaptation methods associated with inversion or pseudo-inversion of the covariance matrix (CM) of the interference consist in determining the weight vector (amplitude-phase distribution) based on the knowledge of the CM of the noise and the direction vector. The operation of finding the weight vector is rather laborious, requires relatively large time expenditures and is associated with inverting the matrix or solving a system of equations. [eleven]. Direct algorithms include algorithms for direct inversion of the estimated noise CM, recurrent inversion of the sample noise CM (direct iterative refinement of the inverse CM), and an algorithm for sequential noise decorrelation based on the Gram-Schmidt orthogonalization procedure.

It is known that when using direct computational methods with the number of samples used to estimate the QM greater than the doubled number of degrees of control, the loss in the average signal-to-noise ratio when replacing the matrix with its sample estimate does not exceed 3 dB. This is significantly (by several orders of magnitude) less than when using gradient methods. An important advantage of direct methods is the independence of the convergence rate of algorithms from the power ratio and the spatial distribution of interference sources [9, 10]. However, another important problem is the negative impact of many rays reflected from the sea surface in radio-technical radar systems.

Space-time processing (STAP) is a modern development of methods for adapting directional patterns of APAR, which simultaneously processes signals received by a plurality of transceiver elements of an antenna array (spatial domain) and a burst of signals (time domain) in the coherent accumulation interval [12].

Spatio-temporal processing significantly improves the capabilities of radar on a mobile platform. First, this method allows objects to be detected at low speed by improving the suppression of passive interference in the main lobe of the radiation pattern. Secondly, STAP allows detecting targets with a small effective scattering area, which are masked by passive interference received from the side lobes of the pattern. Thirdly, STAP provides target detection in conditions of combined active and passive interference. In addition, STAP has the property of robustness to system errors and non-stationarity of the noise background.

For the first time, space-time processing in a radar application was published by Brennan and Reed [8] in 1973, which described an optimal space-time filter. The work of Klemm [7] is devoted to the development of STAP methods, with an attempt to analyze the degrees of freedom of passive interference as applied to an aircraft-based radar. The development of digital signal processing creates conditions for the implementation of space-time adaptive processing in real time and, as a consequence, an increase in interest in the field of STAP. Modern work on the implementation of spatio-temporal adaptive processing focuses primarily on the development and research of effective computational algorithms. One of the directions in monopulse radar is the use of the sum-difference STAP based on space-beam after Doppler filtering with spaced filters and two-stage zeroing.

The advantages of this approach should be considered:

1. Economic attractiveness - minimization of the number of digital receiving channels. Allows to significantly increase reliability and reduce maintenance costs by simplifying the architecture and interconnections of system elements.

2. Data efficiency - the estimation of covariance matrices is performed on a limited sample even in a non-stationary environment in which other STAP architectures lose their efficiency, which they demonstrate with a priori knowledge of the statistical distribution of the passive interference density.

3. Simplification of channel calibration - the problem of calibration of receiving channels is the main task in the construction of adaptive systems with space-time processing, which is associated with matching the signal bandwidth and direction vectors of the antenna elements. Since the channels are not identical in terms of frequency response, the spatial element STAP requires an expensive calibration system. A decrease in the spatial degrees of freedom leads to a significant decrease in the number of generated adaptive weight coefficients. In contrast to this architecture, the sum-difference STAP has two well-structured channels and corresponding direction vectors with a calibration that defines the zero position of the difference channel.

4. Beam Shape Optimization - Adaptive beamforming can result in high sidelobe levels in interference-free angular positions and significant loss of gain in the main beam and a shift in the direction of the main beam, leading to additional redundant degrees of freedom and an estimate of associated errors. The sum-difference STAP with the spatial degree of freedom equal to 1 and the zero position of the differential pattern assumes the unambiguous obtaining of the desired pattern of the pattern, unlike other STAP architectures with a large number of degrees of freedom.

five. Reducing computational complexity - the process of adaptive space-time processing requires significant computing power of technological equipment. The sum-difference STAP significantly reduces the amount of adaptation computation required.

6. Can be applied to existing monopulse radar systems - adaptive suppression of passive interference in a joint "angle-doppler" space can be integrated into existing radars with AFAR or reflector antennas with additional digital processing of the difference channel and modification for the synthesis of a difference channel at the expense of small capital investments with significant improvement combined interference suppression capabilities.

However, there are limitations in the application of sum-difference adaptive space-time processing, namely:

1. Spatial ambiguity - any two-channel system has a significant drawback in the degrees of freedom for effective suppression of passive interference and in any case manifests itself due to the low pulse repetition rate and high Doppler frequencies of passive interference at a high speed of movement of the carrier platform of the radar.

2. Mismatch of radiation patterns - since the presence in the difference radiation patterns of a deep minimum in the space of the side lobes of the total radiation pattern leads to a low efficiency of passive interference suppression. Thus, the total and difference radiation patterns must be optimized.

3. Active Noise Suppression - Noise rejection is not a function of sum-difference adaptive space-time processing. Thus, a preliminary suppression of active interference with the organization of auxiliary channels is necessary.

4. Limited possibility of measurement accuracy - the difference channels are not free from passive interference reflections. Doppler filtering of interference and targets is required.

For these reasons, it is necessary to develop a method for optimizing the total-difference adaptive space-time processing of radio signals in monopulse shipborne radars with APAR, free from these restrictions.

In the prior art, such methods of space-time processing of radio signals to suppress active and passive interference have not been identified.

The objective of the invention is to develop a method for space-time adaptive signal processing in a monopulse shipborne radar with an active phased array antenna (AFAR).

The technical result is an increase in the degree of suppression of active and passive interference in monopulse shipborne radar with AFAR.

The essence of the invention.

The solution to the problem and the achievement of the claimed technical result is ensured by the fact that the method of spatio-temporal adaptive signal processing in a monopulse shipborne radar with an active phased antenna array (AFAR) includes receiving radio signals from a target against the background of active and passive interference using an AFAR. The received signals are digitized. Next, a two-stage adaptive space-time processing of the received radio signals is carried out simultaneously through the sum and difference channels. Then their threshold processing and target detection against the background of active and passive interference. At the first stage of processing, the signals "active interference plus noise" are processed, received by the AFAR elements and free from passive reflections. A spatially covariance matrix "F" of the processed signals is formed. According to the data of the "F" matrix, an adaptive directional diagram (DP) of the AFAR is created in digital form with the creation of deep zeros in the direction of active interference. Further, at the second stage of adaptive signal processing, passive interference is suppressed by Doppler filtering and multi-window signal processing using an adaptive APAA pattern formed at the stage of suppressing active interference.

Such a space-time adaptive two-stage signal processing allows suppressing active interference in time and then using the results of the first stage of adaptive processing to suppress passive interference. The consequence of this is an increase in the degree of suppression of active and passive interference in monopulse shipborne radars with AFAR by at least 15%. At the same time, the signal processing algorithm is simplified and the time for suppression of active and passive interference is reduced by more than 2 times.

The essence of the invention is illustrated by the drawings shown in FIG. 1 to FIG. 3.

FIG. 1 shows a two-stage block diagram of adaptive signal processing by zeroing combined (active plus passive) interference; in fig. 2 is a block diagram of the first stage of adaptive signal processing to suppress active interference; in fig. 3 is a block diagram of the second stage of adaptive signal processing to suppress passive interference.

Disclosure of the essence of the invention

As shown in FIG. 1 to FIG. 3, a method of spatio-temporal adaptive signal processing in a monopulse shipborne radar with an active phased antenna array (AFAR), includes receiving radio signals from a target against a background of active and passive interference using AFAR 1. Received AFAR 1 signals are converted into digital form (in the figures, shown). Next, two-stage 2-3 adaptive spatio-temporal processing of the received radio signals is carried out along the sum and difference channels simultaneously. Then their threshold processing and target detection against the background of active and passive interference. At the first 2 stage of processing, the signals "active interference plus noise", received by the AFAR elements and free from passive reflections, are processed. Form 2.2 spatially covariance matrix "F" of the processed signals. According to the matrix "F" create 2.1 in digital form adaptive radiation pattern (DP) AFAR with the creation of deep zeros in the direction of active interference. Further, at the second 3 stages of adaptive signal processing, 3.1 passive interference is suppressed by Doppler filtering and multi-window processing of 3.2 signals using the adaptive APAA pattern formed at stage 2 of active interference suppression.

The formation of the adaptive radiation pattern (DP) of the AFAR with the creation of deep zeros in the direction of active interference is carried out at the first 2 stage of processing in the azimuthal and elevation planes by window convolution (Fig. 2) of signals using the Baylis distribution coefficients.

Multi-window signal processing at the second 3 processing stage is carried out (Fig. 3) using a plurality of parallel matched filters 3.2 with subsequent coherent processing of the filtering results and the formation of 3.1 covariance matrix of passive interference. In turn, the formation of the covariance matrix is ​​carried out by the sliding window method with the grouping of the matrix discrete according to the Kronecker multiplication rule.

At the first stage 2 of adaptive processing, the signals are first subjected to Doppler filtering using a bank of space-time filters. The filters are formed by cascading the calculators of the spatial beamforming weights of each pulse and the temporal Doppler filters in each formed beam. The filtered signals are then adaptively summed to produce the output response of a single Doppler frequency bin. The process is performed for each bin in the range of the normalized Doppler frequency. The joint procedure of beamforming and after the Doppler filtering of the coherent burst provides significant suppression of active interference, provided that they are preliminary localized in space before adaptation and subsequent adaptation to passive interference in each target Doppler bin.

Formally specified architecture 3.1 of the STAP T processor_m dimension MN × K for adaptive processing of m bin (M is the number of pulses in a burst, N is the number of AFAR elements, K is the dimension of the sample) can be implemented on the basis of two types described by expressions (1) and (2).

<img file="RU2735216C2_D0001.tif" />

In expression (1) F_m - M × K_t is a matrix of Doppler filters, and <img file="RU2735216C2_D0002.tif" /> - matrix of diagram formation N × K_s, K = K_tK_s

This type of algorithm (2) processing 2.1 (Fig. 1 - Fig. 2) signals is separable (separable), since it contains a plurality of beamforming calculators for each pulse of the burst and a plurality of Doppler filters in each formed beam of the pattern.

The second type of algorithm (2) processing 2.2 (Fig. 1 - Fig. 2) is formed by choosing a certain set of outputs of a separable processor, in which F - M × M, G - N × N, and J_m - MN × K forming matrices of filter subsets in the "angle-doppler" space.

The transformations in this algorithm are also separable since the selection matrices are separable. Both types of processor architectures can implement a two-stage noise cancellation procedure.

At the first 2 stage of processing signals received by the AFAR 1 elements (active interference plus noise), free from interfering reflections, the spatial covariance matrix is ​​estimated (3

<img file="RU2735216C2_D0003.tif" />

In each subinterval of coherent accumulation, the matrix Ф_jn is used to construct adaptive patterns of AFAR 1 with the formation of deep zeros in the direction of active interference as part of the general algorithm of spatial-beam transformation. In the second stage 3, the resulting beam response, with suppression of active interference signals, is used to adapt the suppression of passive interference.

The resulting responses of the active interference cancellation beams are then used in the second stage 3 of the space-time processing of passive interference cancellation.

<img file="RU2735216C2_D0004.tif" />

Where <img file="RU2735216C2_D0005.tif" /> - the desired AFAR beam pattern in the absence of active interference.

Based on these provisions, the proposed method implements a two-step 2-3 computational architecture of sum-difference adaptive processing based on the implementation of the optimization method by a biological algorithm for the quantum behavior of a population of particles.

Stage 1 (Fig. 1 - Fig. 2) Synthesis of adaptive total and difference AFAA directivity patterns.

At this stage, the process of adapting the total and difference radiation patterns is aimed at generating optimal beamforming weights in order to minimize errors between the required and actual received signals and maximize the signal-to-noise ratio in the direction of sight with suppression in the direction of interference. Many adaptation methods are known based on algorithms LMS, CMA, SMI, MVDR, GA, PSO [6]. Comparative analysis shows that adaptive beamforming using the PSO biological algorithm provides 0 dB gain in the main lobe of the pattern with interference rejection down to -20 dB to -50 dB and maintains low side lobes. In addition, the main advantage of PSO over stochastic and other optimization methods is the relative simplicity, which consists in the ability to tune only the velocity operator to find the optimum of a function in hyperspace. Several modifications of the classical PSO are known in relation to antenna arrays and its integration with other methods in which attempts are made to increase the convergence rate and the efficiency of searching for an extremum. However, as was proved by Bergh [6], PSO does not guarantee the global convergence of the algorithm in accordance with the chosen optimization criterion.

A plane antenna array 1 with boundaries lying in the xy plane is considered, the coordinates of each element (subarray) of the AFAR 1 correspond to

<img file="RU2735216C2_D0006.tif" />

where d_x= d_y= d is the distance between elements (sublattices). Then the antenna array factor is determined by the well-known expression

<img file="RU2735216C2_D0007.tif" />

where I_mn - excitation coefficients and <img file="RU2735216C2_D0008.tif" />

The formation of the total radiation pattern is performed using the real numerical coefficients of the Taylor series expansion

<img file="RU2735216C2_D0009.tif" />

The formation of difference radiation patterns in the azimuth and elevation planes is performed using the Baylis distribution coefficients

<img file="RU2735216C2_D0010.tif" />

And from the condition of symmetry

<img file="RU2735216C2_D0011.tif" />

Based on the position of quantum mechanics and the trajectory of particle motion in the PSO (Particle Swarm Optimization) algorithm [5], a new proposed version of PSO is used - PSO with quantum-like behavior of particles moving in a multiparameter search space with a velocity vector dynamically changing according to the presence of their own information and interaction with other particles QPSO (Quantum Behaved Particle Swarm Optimization) [3].

The QPSO algorithm has some characteristics that are different from PSO and make it more attractive from the point of view of adapting the APAR DN. First, the introduction of an exponential distribution of particle positions leads to global convergence of the algorithm. Second, the addition of the best average position is a clear improvement in QPSO. In the classical PSO, each particle tends to the global best position without regard to the movement of the rest of the population. In the QPSO algorithm with the presence of the average best position, the population of particles never leaves the particle lagging behind the optimization movement, which is much closer to the intellectual behavior of society and increases the optimization efficiency of QPSO many times over.

Thus, each individual particle is represented by an elementary particle moving in quantum space and the probability of finding a particle in the position of space <img file="RU2735216C2_D0012.tif" /> when performing calculations at iteration t is determined by the selected probability density function | ψ (x, t) |², the shape of which determines the potential field of the particle. In accordance with the Monte Carlo method, the particle moves according to the following rules described by expressions (10-12)

<img file="RU2735216C2_D0013.tif" />

if randν≥0.5

<img file="RU2735216C2_D0014.tif" />

if randν &lt;0.5,

Where

<img file="RU2735216C2_D0015.tif" />

parameter α determines the compression-extension ratio, <img file="RU2735216C2_D0016.tif" /> and randν are random numbers uniformly distributed in the interval [0,1], mbest is a global virtual point called the main direction or the best average calculated by the formula

<img file="RU2735216C2_D0017.tif" />

The tensile compression ratio changes over time according to the rule represented by the ratio (14)

<img file="RU2735216C2_D0018.tif" />

where α₀ and α₁ are the initial and final values ​​of α, respectively, T is the maximum value of computational iterations, t is the number of the current iteration of the search for an extremum. The convergence of the algorithm is achieved in the case when each particle is at its local point of attraction p_i= (p_{i, 1}, p_{i, 2},…, P_{i, D}), whose coordinates for each iteration are determined by the expression

<img file="RU2735216C2_D0019.tif" />

Where <img file="RU2735216C2_D0020.tif" /> c₁ and with₂ - two positive coefficients of particle acceleration, <img file="RU2735216C2_D0021.tif" /> and <img file="RU2735216C2_D0022.tif" /> - two random numbers evenly distributed on the interval [0,1].

The procedure for applying QPSO to the process of adapting the total and difference directional patterns of AFAR 1 at the first stage to the combined interference background is performed independently in the total and difference pattern in accordance with the iterative optimization algorithm and the maximum number of iterations 50.

The fitness functions, on the one hand, are set from the condition of ensuring the maximum signal-to-noise ratio in the direction of sighting of the total RP, on the other hand, minimizing the output power with restrictions on the gain in the differential RP. Namely, for the total channel:

<img file="RU2735216C2_D0023.tif" />

for difference channels:

<img file="RU2735216C2_D0024.tif" />

where w are independent weight coefficients in the total and difference azimuth and elevation channels, R is the statistical estimate of the covariance matrix with the diagonal noise procedure in the interval preceding the emission of the probe pulse, v is the direction vector taking into account the amplitude weighting by the Taylor A and Baylis coefficients [B^E, B^H]. Thus, the active interference suppression procedure is performed using the unlimited degrees of freedom for each pulse repetition period.

The signals of the sum and difference channels are processed in the entire possible range of Doppler frequencies by matched filters using the overlap-add FFT windowed convolution method [4]. In the proposed system, the total number of parallel Doppler filters in each channel is K = 2M + 1, in which MF₀ - corresponds to matched signal filtering with zero Doppler frequency shift, MF₁ - for a signal with a Doppler shift F₁ and MF_M for the Doppler shift F_M... Since the direction of movement of the target is a priori unknown, the filters are arranged symmetrically with respect to F₀, that is, there is a filter for both F_k and for F_-k k = 1.2 ... M.

Stage 2 (Fig. 1 - Fig. 3). Post-Doppler adaptive sum-difference space-time processing.

A multi-window post-Doppler STAP 3.1 is used, which includes a plurality of parallel matched filters in processing a burst of impulse echo signals in each channel [2]. Each filter in the channels represents a window for K_t sub-intervals for the coherent accumulation interval M. The adaptation process operates with a small K_t sub-interval from K_s responses of the sum and difference channels. Thus, the dimension of the problem is determined by the dimension in K = K_sK_t and is K_s= 3 and K_t&lt;&lt; M (K_t is chosen in the range 2-4), which actually achieves a significant reduction in dimension. After each matched filter, the processor performs pipelined adaptation separately for each subinterval with subsequent coherent processing of the results of all impulses of the subintervals (0: K_t-1.1: K_t, ..., MK_t : M-1). The vector of weight coefficients for each subinterval m is calculated by the formula for the criterion for maximizing the signal-to-noise ratio

<img file="RU2735216C2_D0025.tif" />

Where <img file="RU2735216C2_D0026.tif" /> - sample covariance matrix of passive interference with dimension K_sK_t× K_sK_tequal to

<img file="RU2735216C2_D0027.tif" />

Sample χ₁ covers the discretes of the range interval surrounding the analyzed sample, with the exception of the sample itself and adjacent preceding and subsequent protection samples <img file="RU2735216C2_D0028.tif" /> For all analyzed discrete receiving strobe to reduce the amount of computations, the covariance matrix is ​​estimated by the sliding window method with grouping [1].

Vector <img file="RU2735216C2_D0029.tif" /> - calculated according to Kronecker's multiplication rule

<img file="RU2735216C2_D0030.tif" />

where w = ƒ_dT_r - the product of the Doppler frequency of the response of the matched filter and the pulse repetition rate.

The results of the adaptive processing of stage 2 are then detected in block 4 and transmitted to the threshold processing of signals with a constant probability of false alarms (not shown in the figures). As a result of thresholding, useful signals from the target are detected, free of active and passive interference. Further, the detected signals from the target are transmitted to trajectory processing and displayed on the radar screen.

Industrial applicability

The invention was developed at the level of a mathematical model of space-time adaptive signal processing for a monopulse shipborne radar station (radar) with an active phased antenna array (AFAR).

A technological sample of a device that implements the proposed method of adaptive signal processing is being developed

Literature

1. Yoshinari Iwakura, Junichiro Suzuki, Hiroyoshi Yamada, An Efficient Sliding Window Processing for the Covariance Matrix Estimation, Graduate School of Science and Technology, Niigata University. Japan 2006

2. SD Blunt. J. Jakabosky, J. Metcalf, J. Stiles, B. Himed, "Multiwaveform STAP," IEEE Radar Conf., Ottawa, Canada, Apr.-May 2013.

3. LD Coelho, A quantum particle swarm optimizer with chaotic mutation operator, Chaos, Solitons and Fractals 37 (2008).

4. Selesnick, IW &amp; Burrus, CS "Fast Convolution and Filtering" Digital Signal Processing Handbook Ed. Vijay K. Madisetti and Douglas B. Williams Boca Raton: CRC Press LLC, 1999

five. J. Kennedy, R. Mendes, Population structure and particle swarm performance, 2002, pp. 1671-1676.

6. M. Rangaswamy, F. Lin, and K. Gerlach, "Robust adaptive signal processing methods for heterogeneous radar clutter scenarios", Signal Processing, vol. 84, pp. 1653-1665,2004.

7. KLEMM, R .: 'Principles of space-time adaptive processing' (IEE Publishing, London, UK, 2002, 2nd edn.)

eight. BRENNAN, LE and REED, LS: 'Theory of Adaptive Radar', IEEE Trans.Aerosp. Electron. Syst, March 1973, 9, pp. 237-252

nine. A.P. Trifonov Joint discrimination of signals and assessment of their parameters / Trifonov A.P., Shinakov Yu.S. - M .: Radio and communication, 1986 .-- 266 p.

ten. V.V. Karavaev Statistical theory of passive location / Karavaev V.V., Sazonov V.V. - M .: Radio and communication, 1987 .-- 237 p.

eleven. Spatial-temporal signal processing / Kremer I.Ya., Kremer A.I., Petrov V.M. and etc.; Ed. Kremer I. Ya. - M .: Radio and communication, 1984 .-- 224 p.

12. Volosyuk V.K., Gulyaev Yu.V. and others. Modern methods of optimal processing of spatio-temporal signals in active, passive and combined active-passive radio systems // Radio engineering and electronics, 2014, volume 59, no. 2, p. 109-131.

13. US Patent No. US9971027 B1, G01S 13/5244, G01S 7/2813, G01S 13/5246, H04B 7/0848;

fourteen. USA, patent No. US 6720910 B2, G01S 13/52, G01S 13/00