In this study, we focus on the Multi-robot Coverage Path Planning (MCPP) problem for maritime Search And Rescue (SAR) missions using a multiple Autonomous Underwater Vehicle (AUV) system, with the ultimate purpose of efficiently and accurately discovering the target from sonar images taken by Side-Scan Sonar (SSS) mounted on the AUVs. Considering the specificities of real maritime SAR projects, we propose a novel MCPP method, in which the MCPP problem is transformed into two sub-problems: Area partitioning and single-AUV coverage path planning. The structure of the task area is first defined using Morse decomposition of the spike pattern. The area partitioning problem is then formulated as an AUV ordering problem, which is solved by developing a customized backtracking method to balance the workload and to avoid segmentation of the possible target area. As for the single-AUV coverage path planning problem, the SAR-A* method is adopted, which generates a path that preferentially visits the possible target areas and reduces the number of turns to guarantee the high quality of the resulting sonar images. Simulation results demonstrate that the proposed method can maintain the workload balance and significantly improve the efficiency and accuracy of discovering the target. Moreover, our experimental results indicate that the proposed method is practical and the mentioned specificities are useful for discovering targets.

Task allocation is crucial for autonomous underwater vehicle (AUV) collaboration in multi-AUV maritime search and rescue missions. In real projects, there are possible target areas existing in task areas, which are not expected to be divided. Motivated by such a special situation, this paper proposes an area partitioning method to allocate the task to multiple AUVs and maintain the possible target area as a whole. First, the spatial structure of the task area is defined by the spiked Morse decomposition, which divides the task area according to a set of angles. Then, we perform a variational transformation to determine the optimal angles using the AUV order. Next, a customized backtracking method is introduced to determine the optimal AUV order which divides the task area among the multiple AUVs without disturbing the possible target areas. The proposed methodology is validated under various challenging scenarios using a different number of AUVs. The empirical results show that the divided possible target areas and workload variance were superior to the comparison methods. This indicates that the proposed method can generate stable solutions that effectively reduce the segmentation of possible target areas and keep the workload of the multiple AUVs balanced.

Multi-modal fusion can exploit complementary information from various modalities and improve the accuracy of prediction or classification tasks. In this paper, we propose a parallel, multi-modal, factorized, bilinear pooling method based on a semi-tensor product (STP) for information fusion in emotion recognition. Initially, we apply the STP to factorize a high-dimensional weight matrix into two low-rank factor matrices without dimension matching constraints. Next, we project the multi-modal features to the low-dimensional matrices and perform multiplication based on the STP to capture the rich interactions between the features. Finally, we utilize an STP-pooling method to reduce the dimensionality to get the final features. This method can achieve the information fusion between modalities of different scales and dimensions and avoids data redundancy due to dimension matching. Experimental verification of the proposed method on the emotion-recognition task using the IEMOCAP and CMU-MOSI datasets showed a significant reduction in storage space and recognition time. The results also validate that the proposed method improves the performance and reduces both the training time and the number of parameters.

Multi-modal fusion can achieve better predictions through the amalgamation of information from different modalities. To improve the performance of accuracy, a method based on Higher-order Orthogonal Iteration Decomposition and Projection (HOIDP) is proposed, in the fusion process, higher-order orthogonal iteration decomposition algorithm and factor matrix projection are used to remove redundant information duplicated inter-modal and produce fewer parameters with minimal information loss. The performance of the proposed method is verified by three different multi-modal datasets. The numerical results validate the accuracy of the performance of the proposed method having 0.4% to 4% improvement in sentiment analysis, 0.3% to 8% improvement in personality trait recognition, and 0.2% to 25% improvement in emotion recognition at three different multi-modal datasets compared with other 5 methods.

Underwater point cloud registration using Multibeam Echosounder (MBES) data is fundamental to marine exploration and seafloor mapping. However, MBES point clouds present unique challenges compared to terrestrial Light Detection and Ranging (LiDAR): high noise levels, low overlap rates, and strongly anisotropic distributions caused by the strip-like sonar scanning pattern. These characteristics degrade existing registration algorithms, which predominantly assume locally isotropic point distributions. To address these challenges, this paper proposes Anisotropic Kernel Point Convolution (A-KPConv), a novel operator tailored to the strip-like structure of MBES point clouds. A-KPConv uses Principal Component Analysis (PCA) to estimate local geometric principal directions and constructs an affine transformation that adapts the convolution kernel shape and orientation to align with the local geometry, thereby shifting feature extraction from isotropic aggregation to structure-aware feature learning along the principal structural directions. Building upon this operator, we integrate A-KPConv into Predator—a framework for low-overlap registration—to develop A-Predator, in which the standard isotropic KPConv in the first three encoder layers is replaced with A-KPConv so that structure-aware feature learning is performed where geometric information is most salient. Extensive experiments on the public Dotson-east dataset and a self-collected LiQuan Lake (LQL) MBES dataset demonstrate that A-Predator achieves the highest registration recall among the evaluated methods. On Dotson-east, recall improves from 31.63% to 59.55% under 10% overlap, with consistently low translation and rotation errors. Ablation studies support the effectiveness of A-KPConv relative to the evaluated anisotropic operators, and cross-dataset experiments suggest more effective transfer than the evaluated baselines from Dotson-east to the rescaled LQL-MBES data without fine-tuning.

Multi-modal information fusion integrates data from various sensors, distinct sources, or different modalities, such as audio, images, and text, to achieve a more comprehensive and accurate understanding and analysis. This paper proposes a Semi-Tensor Product-based Block Term Decomposition of Multilinear (STP-BTDM) pooling method and applies it to sentiment analysis and emotion recognition. Unlike prior factorized multilinear approaches, STP-BTDM introduces block-term decomposition with a block-diagonal core tensor, yielding a globally sparse yet locally dense structure and enabling modality-specific independent subspace learning. The technique first introduces the Semi-Tensor Product-based Block Term Decomposition (STP-BTD) model to obtain globally sparse and locally dense weight tensors. Subsequently, by combining the multilinear pooling model, the STP-BTDM method is presented. This approach allows each modality to be controlled by only one block of the block-diagonal core, with different blocks being independent during training. The model can represent full multilinear interactions in a computationally efficient manner. Moreover, the introduction of sparsity constraints in the core tensor of STP-BTDM enhances the generalization performance of multilinear pooling. The resulting locally dense yet globally sparse characteristics make the model highly flexible. Finally, our experiments with the STP-BTDM method on the CMU-MOSI dataset for sentiment analysis and the IEMOCAP dataset for emotion recognition demonstrate its superior performance and effectiveness across both tasks.