The canal waterways of China still play an important role in the logistics and transportation industry. Unmanned technology helps to reduce costs and improve the safety of navigation. Real-time environmental awareness is vital to making unmanned surface vehicles (USVs) come true. This paper proposes a new lightweight environmental awareness method based on deep convolutional neural networks ( DCNN ) and a mixed attention mechanism for USVs in canals, which can simultaneously perform ship detection, segmentation, and surface and background segmentation tasks. The features of the ships, surface, and background are extracted by a shared feature extraction backbone network and hybrid attention mechanism, which improves the efficiency of visual environmental awareness. In addition, a dataset named USV-Canal is constructed to enrich the features of canal waterways for environmental awareness, which contains typical canal scenes and 3443 ship objects. To improve the generalization, multiple public datasets are mixed with the USV-Canal dataset to build an integrated dataset to train our model, which boasts diversity in scene types and ship classes. The comparative and field experiments’ results show that 40.9% of mAP , 95.8% of mIoU, and 5 frames per second ( FPS ) inference speed can be achieved, and have good generalization, which can meet the requirements of environmental awareness of low-speed ships in canal waterways Note to Practitioners —The trained and validated model can ultimately be deployed on unmanned surface vehicles, and the required hardware platform is NVIDIA’s Jetson Nano, which is used for real-time perception of surrounding ships and navigable surfaces during navigation. The information can be integrated into the guidance, navigation, and control (GNC) system of USVs, achieving obstacle avoidance and ensuring safe navigation. It is vital to make autonomous navigation come true.
23 October 2024 in IEEE Transactions on Intelligent Transportation Systems
The collision avoidance in real-time is crucial for unmanned surface vehicles (USVs) in a complex environment. Traditional methods make it hard to ensure the balance of control decisions. To balance safety and practicality, a collision avoidance algorithm based on deep reinforcement learning (DRL) and a two-level incentive reward based on the principle of complementarity is proposed. To address the vital sparse reward problem of Deep Deterministic Policy Gradient (DDPG), the trajectory evaluation function of the dynamic window algorithm (DWA) is referred to construct the primary reward strategy, and a secondary incentive reward is constructed based on velocity obstacle (VO) to eliminate potential collision risks. To improve the efficiency of training, the electronic chart (EC) and Unity3D are used to build an immersive simulation platform. Based on it, simulations are made to verify the performance. In addition, field experiments are first conducted in various encounter scenarios to verify the effectiveness. The results show that it can take safe collision avoidance actions and get practical paths in various situations.
14 October 2024 in IEEE Internet of Things Journal
Autonomous collision avoidance technology is the core of unmanned surface vehicles (USVs). Deep reinforcement learning (DRL) is a new approach to avoid collision for USVs. However, most research is based on the assumption of a fixed number of obstacles and ignores the collision prediction to improve safety. To address this problem, a novel “prediction-decision” collision avoidance model based on the deep deterministic policy gradient (DDPG) is proposed. First, a radiation-shaped state space is designed to make the DDPG that can be used in time-varying scenarios with stochastic obstacles. Then, the velocity obstacle (VO) is combined with the state space for training to realize the collision prediction. Subsequently, reward functions are designed using a reward-shaping technique to improve training efficiency and safety. Finally, virtual simulation experiments based on Unity3D and field tests are conducted to verify the algorithm’s performance. The results show that it can take safe collision avoidance actions in unknown environments and with generalization ability.
The design, simulation, and testing of intelligent trajectory-tracking control in narrow waters are essential issues for unmanned surface vehicles (USVs). Due to limited actuators, spatial constraints, and obstacles in narrow waters, the reference trajectory for USVs has various curves. This presents significant challenges to the accuracy and computational load of trajectory tracking. Therefore, a novel event-triggered-based nonlinear model predictive control with artificial reference trajectory (ENMPC-ART) method is proposed. The artificial reference decision variables are integrated into the quadratic trajectory planning of reference trajectory and motion control of USVs to reduce the cross-track error. An event-triggered mechanism is designed to improve NMPCs efficiency. Further, a cyber-physical simulation test framework based on virtual reality is designed to verify the algorithms performance and enhance the immersion. Finally, the proposed ENMPC-ART shows significant improvements through virtual simulations and field tests, such as the maximum cross-track error being reduced by 16% and the computation time being reduced by 20.2%.