This work presents a framework that allows Unmanned Surface Vehicles (USVs) to avoid dynamic obstacles through initial training on an Unmanned Ground Vehicle (UGV) and cross-domain retraining on a USV. This is achieved by integrating a Deep Reinforcement Learning (DRL) agent that generates high-level control commands and leveraging a neural network based model predictive controller (NN-MPC) to reach target waypoints and reject disturbances. A Deep Q Network (DQN) utilized in this framework is trained in a ground environment using a Turtlebot robot and retrained in a water environment using the BREAM USV in the Gazebo simulator to avoid dynamic obstacles. The network is then validated in both simulation and real-world tests. The cross-domain learning largely decreases the training time () and increases the obstacle avoidance performance (70 more reward points) compared to pure water domain training. This methodology shows that it is possible to leverage the data-rich and accessible ground environments to train DRL agent in data-poor and difficult-to-access marine environments. This will allow rapid and iterative agent development without further training due to the change in environment or vehicle dynamics.
Obstacle detection for autonomous navigation through semantic image segmentation using neural networks has grown in popularity for use in unmanned ground and surface vehicles because of its ability to rapidly create a highly accurate pixel-wise classification of complex scenes. Due to the lack of available training data, semantic networks are rarely applied to navigation in complex water scenes such as rivers, creeks, canals, and harbors. This work seeks to address the issue by making a one-of-its-kind River Obstacle Segmentation En-Route By USV Dataset (ROSEBUD) publicly available for use in robotic SLAM applications that map water and non-water entities in fluvial images from the water level. ROSEBUD provides a challenging baseline for surface navigation in complex environments using complex fluvial scenes. The dataset contains 549 images encompassing various water qualities, seasons, and obstacle types that were taken on narrow inland rivers and then hand annotated for use in semantic network training. The difference between the ROSEBUD dataset and existing marine datasets was verified. Two state-of-the-art networks were trained on existing water segmentation datasets and tested for generalization to the ROSEBUD dataset. Results from further training show that modern semantic networks custom made for water recognition, and trained on marine images, can properly segment large areas, but they struggle to properly segment small obstacles in fluvial scenes without further training on the ROSEBUD dataset.
Underwater docking is critical to enable the persistent operation of Autonomous Underwater Vehicles (AUVs). For this, the AUV must be capable of detecting and localizing the docking station, which is complex due to the highly dynamic undersea environment. Image-based solutions offer a high acquisition rate and versatile alternative to adapt to this environment; however, the underwater environment presents challenges such as low visibility, high turbidity, and distortion. In addition to this, field experiments to validate underwater docking capabilities can be costly and dangerous due to the specialized equipment and safety considerations required to conduct the experiments. This work compares different deep-learning architectures to perform underwater docking detection and classification. The architecture with the best performance is then compressed using knowledge distillation under the teacher-student paradigm to reduce the network's memory footprint, allowing real-time implementation. To reduce the simulation-to-reality gap, a Generative Adversarial Network (GAN) is used to do image-to-image translation, converting the Gazebo simulation image into a realistic underwater-looking image. The obtained image is then processed using an underwater image formation model to simulate image attenuation over distance under different water types. The proposed method is finally evaluated according to the AUV docking success rate and compared with classical vision methods. The simulation results show an improvement of 20% in the high turbidity scenarios regardless of the underwater currents. Furthermore, we show the performance of the proposed approach by showing experimental results on the off-the-shelf AUV Iver3.