Design method of vertical large-opening-angle anti-collision sonar

Abstract

The invention discloses a vertical large-opening-angle collision-prevention sonar design method and belongs to the field of sonar design. The method is used for guaranteeing safe navigation of an underwater vehicle. The method mainly relates to a sonar design technology and a signal processing method, through optimization of formation design and improvement of the signal processing method, the obstacle target detection capability under the condition of a large opening angle is improved, especially the strong reverberation background suppression capability is improved, and thus the environmental applicability of the anti-collision sonar is improved.

Description

Technical Field

[0001] The present invention relates to a collision avoidance sonar design technology for ensuring the safe navigation of underwater vehicles. The method mainly involves a sonar design technology and a signal processing method. By optimizing the formation design and improving the signal processing method, the obstacle target detection capability under wide opening angle conditions is improved, especially the strong reverberation background suppression capability is improved, thereby improving the environmental applicability of the collision avoidance sonar. Background Art

[0002] Autonomous underwater vehicles (AUVs), known for their autonomous and flexible nature, are gaining increasing attention both domestically and internationally. When performing tasks such as ocean exploration, scientific research, and environmental monitoring, AUVs, especially unmanned submersibles and automated underwater vehicles, often face unknown and complex underwater environments, posing the risk of collision with other underwater objects. To ensure the safety of AUVs during these missions, the application of Obstacle Avoidance Systems (OAS) is crucial.

[0003] Early collision avoidance systems generally had a small vertical opening angle when detecting obstacles. However, in actual use, especially in complex sea areas, the process of diving and surfacing of the vehicle requires a large vertical safe driving space.

[0004] This method adopts the large-opening-angle collision avoidance sonar design technology. On the one hand, it adopts the optimized formation design technology to ensure a large vertical opening angle. On the other hand, in order to solve the problem of large reverberation interference introduced by the large opening angle, it adopts the signal processing algorithm to achieve reverberation suppression and improve the obstacle target detection capability. Summary of the Invention

[0005] This invention addresses the need for underwater vehicle collision avoidance. Existing collision avoidance technologies suffer from a narrow vertical angle, failing to meet the requirements for safe navigation. This invention proposes a collision avoidance sonar design suitable for underwater vehicles with large vertical angles. The design incorporates a large vertical angle array design and reverberation suppression signal processing technology. This technology ensures the collision avoidance sonar has wide vertical coverage and high target detection capabilities.

[0006] The technical solution provided to solve the purpose of the invention;

[0007] The specific technical solutions of the present invention are as follows:

[0008] The sonar system consists of a receiving transducer and electronic cabin, a transmitting transducer and electronic cabin, and a dry end electronic cabin.

[0009] 1) Transducer design based on vertical multi-subarray

[0010] The receiving transducer adopts a T-shaped linear array with separate transmitting and receiving design, and uses a longitudinal vibration composite rod transducer as the base array element.

[0011] In order to ensure a large vertical opening angle, the receiving array adopts a vertical up and down multi-subarray design. Figure 2 .

[0012] In order to avoid the aliasing caused by multiple sub-arrays, a combination of frequency division and time division is used to distinguish the two beam space signals. That is, the array center frequency f1 is used to transmit the signal in the t1 period.

[0013] S1=exp[j(2πf1t+πfk1t 2 )]t∈[t1,t+T]

[0014] Where k1 is the frequency change rate.

[0015] The center frequency of the next array is f2, and the signal is transmitted during the t2 period.

[0016] S1=exp[j(2πf1t+πfk2t 2 )]t∈[t1,t+T]

[0017] Where k2 is the frequency change rate.

[0018] This design technology can ensure a larger vertical opening angle. For example, if a vertical opening angle of 60° is required.

[0019]

[0020] Considering that actual vehicles typically operate close to the water surface, where reverberation is more pronounced, the vertical array opening angle is designed to be smaller, φ1 = 20° and φ2 = 40°, based on the sound field distribution. If the vehicle is used in deep sea (water depth greater than 1000m), φ1 = 30° and φ2 = 30°, based on the sound field distribution.

[0021] 2) Reverberation suppression based on robust principal component analysis

[0022] After the working frequency band and spatial opening angle are designed, signal processing methods need to be further used to reduce the impact of reverberation. The reverberation suppression in this scenario is modeled as a robust principal component analysis (Robust PCA) problem, that is, the echo matrix M composed of multiple frames can be decomposed into a low-rank reverberation matrix L and a sparse target matrix S: M = L + S, where After beamforming for N frames of data The result of vectorization is N r is the number of sampling points, N θ is the number of beams, which can be further structured as the following optimization problem:

[0023]

[0024] stM=L+S

[0025] Due to the non-convex and non-smooth characteristics of the rank and l0 norm, there are difficulties in optimization. Therefore, the nuclear norm is used to represent the rank of the matrix and the l1 norm is used to represent the sparsity of the matrix. The above optimization problem is converted into the following convex optimization problem that is easy to optimize and solve:

[0026]

[0027] stM=L+S

[0028] where ‖L‖ * =∑ i σ i (L) is the nuclear norm, σ i (·) is the i-th singular value of the matrix L, is the l1 norm after matrix vectorization, and λ is the trade-off coefficient between the low-rank matrix L and the sparse matrix S, which is usually selected based on experience and can be Try it on a similar order of magnitude, where m=max{N r N θ ,N}.

[0029] For the above optimization problem, its augmented Lagrangian function can be written as:

[0030]

[0031] Where ρ>0 is the penalty coefficient, and the initial ρ value is set to 1. is the Lagrange multiplier, <Λ,MLS>=tr(Λ T (MLS)) is the inner product of two matrices, ‖·‖ F is the Frobenius norm.

[0032] After the above derivation, the optimization problem is solved through the following iterative process:

[0033]

[0034] Λ k+1 =Λ k +ρ k (ML k+1 -S k+1 )

[0035] where UΣV H =svd(ML k+1 -ρ -1 Λk ) is the result of singular value decomposition, H ε [·] is the contraction operator, H ε [X] = sgn(X)⊙max{|X|-εE,[0]}, where sgn(·) is the sign function, ⊙ represents matrix element multiplication, E is an all-one matrix, and [0] is an all-zero matrix.

[0036] The determination of the original residual is set as follows:

[0037]

[0038] The dual residual is determined as follows:

[0039]

[0040] where δ p , δ d , is the set threshold, the underwater acoustic signal echo signal is weak, combined with the signal echo intensity, the characteristic setting: δ p ∈[0.1*10 -5 ,10*10 -5 ], δ d ∈[0.1*10 -4 ,10*10 -4 ].

[0041] The iterative algorithm stops when the original residual and the dual residual are less than the set threshold. k+1 and S k+1 These are the low-rank matrix and sparse matrix to be solved. The sparse matrix is the result after reverberation suppression. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 System composition diagram.

[0043] Figure 2 Transducer structure diagram.

[0044] Figure 3 Collision avoidance sonar sound image based on the invention; wherein a) conventional processing technology and b) processed target sound image. DETAILED DESCRIPTION

[0045] This method is used to design collision avoidance sonar. The receiving array adopts an upper and lower array design. The spatial opening angle is guaranteed to be no less than 60 degrees. The detection sound map is formed as follows: Figure 3 a), where the targets are located at 30m and 120m respectively.

[0046] Perform reverberation suppression on the received data, where ρ = 1, λ = 0.005, δ p =10 -5 , δ d =10-4 A total of 100 frames of data were processed, and the comparison chart after processing is as follows: Figure 3 b) As can be seen, the target acoustic image has low signal-to-noise ratio before processing, and the target is completely separated in the sparse matrix. This demonstrates the excellent performance of the sonar design and signal processing method, enabling the design of a sonar system with wide vertical angles for obstacle detection.

Claims

1. A method for designing a vertically wide-angle collision avoidance sonar, characterized by: The sonar system used in this method is composed of a receiving transducer and an electronic cabin, a transmitting transducer and an electronic cabin, and a dry end electronic cabin; 1) Transducer design based on vertical multi-subarray The receiving transducer adopts T-shaped linear array, with separate transmitting and receiving design; To ensure a wide vertical opening angle, the receiving array adopts a vertical up-and-down multi-subarray design. The method of combining frequency division and time division is used to distinguish the two beam space signals; that is, the center frequency f1 is used to transmit the signal at time t1. s1=exp[j(2πf1t+πfk1t 2 )]t∈[t1,t+T] Wherein, k1 is the frequency change rate; The center frequency of the lower array is f2, and the signal is transmitted during the t2 period; S1=exp[j(2πf1t+πfk2t 2 )]t∈[t1,t+T] Where k2 is the frequency change rate; Vertical opening angle 2) Reverberation suppression based on robust principal component analysis The echo matrix M composed of multiple frames is decomposed into a low-rank reverberation matrix L and a sparse target matrix S: M = L + S, where After beamforming for N frames of data The result of vectorization is N r is the number of sampling points, N θ is the number of beams, which is constructed as the following optimization problem: Due to the non-convex and non-smooth characteristics of the rank and l0 norm, there are difficulties in optimization. Therefore, the nuclear norm is used to represent the rank of the matrix and the l1 norm is used to represent the sparsity of the matrix. The above optimization problem is converted into the following convex optimization problem that is easy to optimize and solve: where ‖L‖ * =∑ i σ i (L) is the nuclear norm, σ i (·) is the i-th singular value of the matrix L, is the l1 norm after matrix vectorization, λ is the trade-off coefficient between the low-rank matrix L and the sparse matrix S, and is taken Where m = max{N r N θ ,N}; For the above optimization problem, write its augmented Lagrangian function: Where ρ>0 is the penalty coefficient, and the initial ρ value is set to 1. is the Lagrange multiplier, <Λ,MLS>=tr(Λ(MLS)) is the inner product of two matrices, ‖·‖ F is the Frobenius norm; After the above derivation, the optimization problem is solved through the following iterative process: L k+1 =L k +r k (ML k+1 -S k+1 ) where UΣV H =svd(ML k+1 -ρ -1 Λ k ) is the result of singular value decomposition, H ε [·] is the contraction operator, H ε [X] = sgn(X)⊙max{|X|-εE,[0]}, where sgn(·) is the sign function, ⊙ represents matrix element multiplication, E is an all-one matrix, and [0] is an all-zero matrix; The determination of the original residual is set as follows: The dual residual is determined as follows: where δ p , δ d , is the set threshold; The iterative algorithm stops when the original residual and the dual residual are less than the set threshold. k+1 and S k+1 That is the low-rank matrix and sparse matrix to be solved; the sparse matrix is the result after reverberation suppression.

2. The method according to claim 1, characterized in that When the vertical opening angle is 60 degrees, φ1 = 20° and φ2 = 40° are set in non-deep sea conditions; when the vehicle is in deep sea, that is, the water depth is greater than 1000m, φ1 = 30° and φ2 = 30° are set.

Images