The development of Unmanned Surface Vehicles (USVs) has become a key focus in marine robotics, fueling the need for navigation systems capable of performing complex and delicate tasks with speed and precision. However, the end-to-end path tracking process often encounters challenges in learning efficiency, and generalization, and varying environmental conditions. To achieve sample-efficient and robust USV navigation in dynamic maritime environments, the paper proposes a novel hierarchical multi-view latent imitation learning (IL) architecture. By formulating a latent IL objective, the framework disentangles diverse navigation modalities through continuous variables, preventing mode collapse and enhancing behavioral adaptability to non-stationary conditions. High-dimensional multi-view observations are transformed via a ViT-based backbone into compressed latent features to minimize redundant environmental information. These representations are processed by a Mamba-based action encoder, which leverages selective state-space modeling to capture long-term temporal dependencies with high computational efficiency. A UNet-based decoder subsequently forecasts optimal action sequences by synthesizing spatial maps to infer critical environment-agent relationships. This preliminary multi-view latent IL-based trajectory ensures precise tracking and dynamic stability while adhering to physical vehicle constraints. Experimental results validate that this end-to-end approach achieves robust path planning effectiveness, obstacle avoidance capability, and model training efficiency in complex, multi-modal maritime scenarios.

In the realm of unmanned surface vehicle (USV) operations, leveraging environmental factors to enhance situational awareness has garnered significant academic attention. Developing vision systems for USVs presents considerable challenges, mainly due to variable observational conditions and angular vibrations caused by hydrodynamic forces. The paper proposed a novel MDGAN-DIFI network for end-to-end multi-object tracking (MOT), specifically designed for camera systems mounted on USVs. Beyond enhancing traditional MOT models, the proposed MDGAN-DIFI includes preprocessing modules designed to enhance the efficiency of processing input signal quality. Initially, a Deep Iterative Frame Interpolation (DIFI) module is used to stabilize frames in the spatiotemporal domain. Next, an enhanced generative adversarial network (GAN) model is applied to reduce motion blur affecting objects within the field of view. Finally, a YOLO-CSSA architecture combines dual infrared (IR) and RGB data streams to maintain consistent performance across diverse environmental conditions. By synthesizing intermediate frames and restoring blurred details, the framework seeks to stabilize object motion trajectories and recover distinctive appearance features prior to tracking. This approach directly tackles the main causes of tracking failure in maritime environments, such as motion discontinuities and visual degradation. Experimental results demonstrate that the proposed approach outperforms conventional methods in multi-object tracking on USVs, achieving a maximum accuracy (MOTA) of 47.0% and an IDF1 score of 50.1% under challenging operational conditions. Consequently, the proposed multi-object tracking network provides a more robust foundation for subsequent detection and data association processes.

Precise and robust three-dimensional object detection (3DOD) presents a promising opportunity in the field of mobile robot (MR) navigation. Monocular 3DOD techniques typically involve extending existing two-dimensional object detection (2DOD) frameworks to predict the three-dimensional bounding box (3DBB) of objects captured in 2D RGB images. However, these methods often require multiple images, making them less feasible for various real-time scenarios. To address these challenges, the emergence of agile convolutional neural networks (CNNs) capable of inferring depth from a single image opens a new avenue for investigation. The paper proposes a novel ELDE-Net network designed to produce cost-effective 3D Bounding Box Estimation (3D-BBE) from a single image. This novel framework comprises the PP-LCNet as the encoder and a fast convolutional decoder. Additionally, this integration includes a Squeeze-Exploit (SE) module utilizing the Math Kernel Library for Deep Neural Networks (MKLDNN) optimizer to enhance convolutional efficiency and streamline model size during effective training. Meanwhile, the proposed multi-scale sub-pixel decoder generates high-quality depth maps while maintaining a compact structure. Furthermore, the generated depth maps provide a clear perspective with distance details of objects in the environment. These depth insights are combined with 2DOD for precise evaluation of 3D Bounding Boxes (3DBB), facilitating scene understanding and optimal route planning for mobile robots. Based on the estimated object center of the 3DBB, the Deep Reinforcement Learning (DRL)-based obstacle avoidance strategy for MRs is developed. Experimental results demonstrate that our model achieves state-of-the-art performance across three datasets: NYU-V2, KITTI, and Cityscapes. Overall, this framework shows significant potential for adaptation in intelligent mechatronic systems, particularly in developing knowledge-driven systems for mobile robot navigation.

Due to the necessity for lightweight and efficient network models, deploying semantic segmentation models on mobile robots (MRs) is a formidable task. The fundamental limitation of the problem lies in the training performance, the ability to effectively exploit the dataset, and the ability to adapt to complex environments when deploying the model. By utilizing the knowledge distillation techniques, the article strives to overcome the above challenges with the inheritance of the advantages of both the teacher model and the student model. More precisely, the ResNet152-PSP-Net model’s characteristics are utilized to train the ResNet18-PSP-Net model. Pyramid pooling blocks are utilized to decode multi-scale feature maps, creating a complete semantic map inference. The student model not only preserves the strong segmentation performance from the teacher model but also improves the inference speed of the prediction results. The proposed method exhibits a clear advantage over conventional convolutional neural network (CNN) models, as evident from the conducted experiments. Furthermore, the proposed model also shows remarkable improvement in processing speed when compared with light-weight models such as MobileNetV2 and EfficientNet based on latency and throughput parameters. The proposed KD-SegNet model obtains an accuracy of 96.3% and a mIoU (mean Intersection over Union) of 77%, outperforming the performance of existing models by more than 15% on the same training dataset. The suggested method has an average training time that is only 0.51 times less than same field models, while still achieving comparable segmentation performance. Hence, the semantic segmentation frames are collected, forming the motion trajectory for the system in the environment. Overall, this architecture shows great promise for the development of knowledge-based systems for MR’s navigation.