Autonomous collision avoidance is essential for unmanned surface vehicles (USVs) in Maritime Internet of Things (MIoT) systems. Existing deep reinforcement learning (DRL) methods often suffer from poor exploration-exploitation balance and weak safety integration, which may lead to local optima and unsafe actions. To address these issues, a risk-aware role-differentiated dual-actor deep deterministic policy gradient framework (RADA-DDPG) is proposed. Abandoning homogeneous structures, a heterogeneous dual-actor design mitigates local optima by assigning distinct exploration and exploitation roles, dynamically balanced via a Q-value-based switching strategy. Additionally, an embedded risk-critic network directly shapes actions during end-to-end policy optimization, rather than serving merely as an external constraint. Extensive simulations demonstrate that RADA-DDPG outperforms baselines in learning efficiency, safety, and robustness. Furthermore, real-world field experiments with dynamic obstacles validate its feasibility.

Autonomous berthing is a critical step in realizing the full autonomy of unmanned surface vehicles (USVs), which can essentially be regarded as a trajectory-tracking task. It can be further transformed into a problem of nonlinear systems with output constraints. This paper proposes a novel adaptive fixed-time backstepping control scheme based on the barrier Lyapunov function (BLF) for autonomous berthing of USVs. Firstly, a new barrier Lyapunov function is designed to solve the output asymmetric constraint requirement of the autonomous berthing system, and it is also adaptive to the unconstrained system without changing the control structure. Secondly, the convergence of adaptive fixed-time control and bounded tracking of BLF are combined to conquer the long convergence time and nonlinear system uncertainty. Finally, simulation and field tests are conducted to verify our proposed scheme’s superiority.

In ship target detection tasks, challenges arise from heterogeneous ships, complex background interference, and weather conditions. Accurate ship target detection requires deep learning models with strong feature extraction and fusion capabilities. However, previous models still face limitations in addressing these challenges. Firstly, most traditional convolutional networks operate solely in the spatial domain, neglecting the rich global information present in the frequency domain. Additionally, effectively integrating features from different scales is beneficial, yet traditional multi-scale fusion methods typically rely on fixed encoder-decoder architectures, using simple cross-layer connections to integrate local and global features. To overcome these challenges, we propose MPN-YOLO, a network capable of capturing key information in both spatial and frequency domains. The network consists of three main components: a windmill convolution downsampling module, a lightweight space-frequency attention module, and a Multi-scale semantic feature fusion module. We validate the effectiveness of the proposed MPN-YOLO on public datasets. Experimental results demonstrate that, compared to baseline models, MPN-YOLO achieves superior performance, with accuracy and mAP50-95 improving by 8.8% and 5.1%, respectively.

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 (NMPC) with an 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 NMPC’s efficiency. Further, a cyber-physical simulation test framework based on virtual reality is designed to verify the algorithm’s 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%.

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.

An efficient formation-keeping strategy is essential for unmanned surface vehicles (USVs) to achieve complex cooperation missions in the Marine Internet of Things (MIoT) system. However, traditional methods make generating an efficient strategy to adapt to different formation patterns difficult in dynamic MIoT. To address this, we enhance the deep deterministic policy gradient (DDPG) algorithm and propose a novel formation control strategy generation approach. First, we design a generic reward mechanism based on the virtual leader–follower strategy to adapt to different formation patterns, simplify the design process, and optimize the formation control. Then, we adopt the intrinsic curiosity module (ICM) to alleviate the problem of sparse rewards and the prioritized experience replay (PER) mechanism to improve the utilization of experience and accelerate the learning rate. In addition, a Gaussian noise model is integrated into the DDPG approach to simulate various external disturbances, which can improve the robustness of the generated strategy. Finally, we built a virtual simulation environment based on Unity3D and conducted field tests to verify the feasibility and superiority of our approach.

Networked cascade control systems (NCCS) are susceptible to performance degradation from network-induced delays, communication constraints, and component failures. To address these interconnected challenges, this paper develops an integrated fault-tolerant H-infinity control framework for discrete-time nonlinear NCCS. A key contribution is a novel, disturbance-aware adaptive event-triggered mechanism (AETM) that uniquely incorporates measurable external disturbances into its triggering logic. This design dynamically reduces communication frequency while preserving performance. Furthermore, a unified co-design of the controller parameters and the AETM’s weighting matrix is established, while stability is rigorously substantiated through theoretical analysis. Simulation results for a boiler control system validate the proposed method’s effectiveness, demonstrating significant improvements in both system robustness and communication efficiency.

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 namedUSV-Canalis 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 theUSV-Canaldataset 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% ofmAP, 95.8% ofmIoU,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 waterwaysNote 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.

Low-cost heading sensors and environmental interference lead to significant heading measurement error (HME) in Autonomous Surface Vehicles (ASVs), necessitating compensation to enhance path-following accuracy. To address this, we propose a compensated event-triggered line-of-sight (CET-LOS) guidance law. The core objective is to rapidly estimate and compensate for HME using an exponential estimation model. Specifically, the exponential model enables fast HME estimation, while an event-triggered mechanism, based on the convergence state of the cross-track error, ensures accurate error estimation. Additionally, the integral term from the integral LOS (ILOS) is utilized to refine the error estimation further. Experimental comparisons of CET-LOS with LOS, ILOS, and adaptive LOS (ALOS) demonstrate its effectiveness. For straight-line paths, the average cross-track error is reduced by 72.3%, 34.5%, and 62.3%, respectively. For complex paths, the reductions are 43.8%, 34.0%, and 29.2%, respectively. These results highlight the superiority of the proposed method.

In order to solve the problems that the production workshop cannot be effectively monitored in real-time, the production information is opaque, and the production data can not be accurately analyzed, this paper proposes a digital twin production workshop monitoring system based on production data, production process and physical machinery and equipment. Firstly, the whole physical production line was modeled from multiple dimensions such as operation logic, assembly relationship, appearance and so on. Then, the data acquisition, data processing and data analysis of the physical production line were carried out through traditional data acquisition technology, Internet of things technology, and edge computing data preprocessing technology. At the end of this paper, a real-time monitoring system of production line is designed to verify the case. The facts prove that the related methods studied in this paper are feasible for the monitoring of production workshop.

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.

This paper mainly studies the modeling, stability analysis and proportional-integral (PI) controller design for a class of discrete-time cascade control systems (CCSs). In the existing CCSs for solving gain of the controller, the controller mostly adopts the proportional (P) controller, but the most commonly used controller in industrial production is PI controller. Based on the actual demand in industry and in order to obtain a better control effect, the design problem of primary and secondary controllers in a class of discrete cascade control is considered, in which the primary controller adopts a PI controller and the secondary controller adopts a P controller. On this basis, the model of cascade control system (CCS) is established. Then, we can create a new Lyapunov functional, the sufficient conditions for the stability of the system is given. Then, the collaborative design method of primary PI controller and secondary P controller is given by using linear matrix inequality (LMI) technique. Finally, a simulation example of a main steam temperature with CCS structure is given to demonstrate the effectiveness of the method. The PI controller is better than the existing P controller, this method has not only the rapidity of P control, but also the ability of integral control to eliminate steady-state errors. Note to Practitioners—The inspiration of this paper comes from the adjustment of plant controller parameters. The PI controller is the most widely used controller in the existing industrial production, but the adjustment of proportional and integral parameters of the controller is a very troublesome thing. In order to obtain a better controller parameters, a professional engineer may spend a whole day to adjust the parameters. This paper can design the solution of the primary and secondary controller parameters to stabilize the CCS at the same time by LMI technique, which can greatly reduce the work of adjusting the controller parameters. Through simulation, we can know that this method can make the system stable, which shows this method is feasible.