TL;DR
This paper introduces a novel iterative-convex optimization method for designing wide-beam sonar transmit patterns, crucial for autonomous underwater vehicles (AUVs) needing to map large areas. Their method outperforms standard convex optimization and disciplined convex-concave programming (DCCP) in achieving desired beampatterns with fewer iterations.
What Problem Does This Paper Address?
Autonomous underwater vehicles (AUVs) rely on sonar systems like forward-looking sonars (FLS) and side-scan sonars (SSS) for navigation, obstacle avoidance, and seabed mapping. To effectively survey large areas, these sonar systems need to transmit acoustic beams with a wide field of view. Designing these "wide beam" transmit patterns for phased antenna arrays is challenging. Traditional beamforming techniques often focus on narrow beams for target detection or direction-of-arrival estimation. This paper addresses the specific problem of optimizing transmit beamforming for wide beams (greater than 20 degrees) with uniform output power, a requirement for efficient area coverage in sonar imaging.
Methods
The authors propose a new iterative-convex optimization method for designing wide beam sonar transmit patterns. This method builds upon existing convex optimization techniques but introduces an iterative update step for the complex weights of the array elements.
Here's a breakdown:
- Standard Convex Optimization: This approach aims to find optimal weights by minimizing an objective function that is convex. It guarantees a global optimum but can be computationally intensive and might struggle with certain constraints.
- Disciplined Convex-Concave Programming (DCCP): This method extends convex optimization to handle problems that can be decomposed into convex and concave parts. It's used here to allow for more flexibility in the beampattern constraints, particularly when the phase of the weights can vary.
- Iterative-Convex Method (Proposed): This is the core contribution. It starts with an initial set of weights and iteratively refines them. In each iteration, it uses convex optimization to find a new weight vector and then updates the phase of the weights based on the difference between the current and previous beampatterns. This iterative phase update is inspired by previous work but replaces a Fast Fourier Transform (FFT) with convex optimization. The key advantage claimed is achieving better results with fewer iterations compared to DCCP.
Datasets
Not discussed in the paper. The experiments were conducted using simulated data based on defined beampattern masks.
Key Insights and Ideas
- Wide Beams are Crucial for AUVs: The paper firmly establishes the need for wide-beam sonar for efficient underwater mapping and exploration by AUVs.
- Iterative Phase Refinement is Key: The iterative update of the complex weights, particularly the phase, is shown to be more effective than a single-shot convex optimization for achieving desired wide-beam patterns.
- Convex Optimization's Power for Sonar: While convex optimization has been applied to radar and communications, its application to wide-beam sonar transmit beamforming is less explored, and this paper demonstrates its utility.
- Phase Quantization Sensitivity: The study reveals that phase quantization is more critical than amplitude quantization for maintaining beampattern performance, with 8 bits of phase quantization being sufficient for near-ideal results.
Key Results
The paper compares their iterative-convex method against standard convex optimization and DCCP across different beampattern mask designs (rectangular and ramp-shaped) and beamwidths (20°, 50°, and 70°).
- Improved Accuracy: The iterative-convex method consistently achieved lower Mean Absolute Error (MAE) and Passband Ripple (PB-Ripple) compared to standard convex optimization and DCCP. For example, in the 70° ramp-shaped mask scenario, the iterative-convex method achieved an MAE of 0.095 dB, significantly better than Convex (4.254 dB) and DCCP (0.589 dB).
- Efficiency in Iterations: For the same number of iterations, the proposed iterative-convex method produced beampatterns closer to the desired ones than DCCP.
- Beamwidth Control: The method demonstrated good control over the beamwidth, with the iterative-convex method often achieving beamwidths closer to the target than other methods.
- Quantization Robustness: The analysis showed that the proposed method is robust to quantization errors. Approximately 8 bits for phase and 5 bits for amplitude quantization were sufficient to achieve performance comparable to the non-quantized case.
Limitations and Future Work
- Narrowband Assumption: The current method is designed for narrowband signals. The authors suggest extending it to work on broadband signals as a future research direction.
- Simulated Data: The experiments were conducted using simulated data. Investigating the influence of transmission elements experimentally is proposed for future work.
- Computational Cost: While the iterative-convex method is more efficient in terms of iterations for achieving good results, it can still be computationally more intensive than standard convex optimization due to the iterative nature. However, the authors note this is acceptable as optimization is typically done offline.
Summary updated