Recognition of obstacle type based on visual sensors is important for navigation by unmanned surface vehicles (USV), including path planning, obstacle avoidance, and reactive control. Conventional detection techniques may fail to distinguish obstacles that are similar in visual appearance in a cluttered environment. This work proposes a novel obstacle type recognition approach that combines a dilated operator with the deep-level features map of ResNet50 for autonomous navigation. First, visual images are collected and annotated from various different scenarios for USV test navigation. Second, the deep learning model, based on a dilated convolutional neural network, is set and trained. Dilated convolution allows the whole network to learn deep features with increased receptive field and further improves the performance of obstacle type recognition. Third, a series of evaluation parameters are utilised to evaluate the obtained model, such as the mean average precision (mAP), missing rate and detection speed. Finally, some experiments are designed to verify the accuracy of the proposed approach using visual images in a cluttered environment. Experimental results demonstrate that the dilated convolutional neural network obtains better recognition performance than the other methods, with an mAP of 88%.

This paper attempts to solve a challenge in online relative optimal path planning of unmanned surface vehicles (USVs) caused by current and wave disturbance in the practical marine environment. The asymptotically optimal rapidly extending random tree (RRT*) method for local path optimisation is improved. Based on that, an online path planning (OPP) scheme is proposed according to the USV's kinematic and dynamic model. The execution efficiency of RRT* is improved by reduction of the sampling space that is used for randomly learning environmental knowledge. A heuristic sampling scheme is proposed based on the proportional navigation guidance (PNG) method that is used to enable the OPP procedure to utilise the reference information of the global path. Meanwhile, PNG is used to guide RRT* in generating feasible paths with a small amount of gentle turns. The dynamic obstacle avoidance problem is also investigated based on the International Regulations for Preventing Collisions at Sea. Case studies demonstrate that the proposed method efficiently plans paths that are relatively easier to execute and lower in fuel expenditure than traditional schemes. The dynamic obstacle avoidance ability of the proposed scheme is also attested.

The development of unmanned systems (UMS) for naval combat poses a profound challenge to existing conventions regarding the treatment of the shipwrecked and wounded in war at sea. Article 18 of the 1949 Geneva Convention II states that warring parties are required to take “all possible measures” to search for and collect seamen left in the water after each engagement. The authors of the present paper analyze the ethical basis of this convention and argue that the international community should demand that UMS intended for roles in war at sea be provided with the capacity to make some contribution to search and rescue operations.