This paper presents a novel online learning-based fault detection designed for underwater robotic thruster health monitoring. In the fault detection algorithm, we build a mathematical model between the control variable and the propeller speed by fitting collected online work status data to the model. To improve the accuracy of online modeling, a multi-center PSO algorithm with memory ability is utilized to optimize the modeling parameters. Additionally, a model online update mechanism is designed to accommodate the model to the change of thruster work status and sea environment. During the operation, propeller speed of the underwater robot is predicted through the online learning-based model, and the model residuals are used for thruster health monitoring. To avoid false alarm, an adaptive fault detection strategy is established based on model online update mechanism. The proposed method has been extensively evaluated using different underwater robotics, through a sea trial data simulation, a pool test fault detection experiment and a sea trial fault detection experiment. Compared with fixed model-based method, speed prediction MAE of the online learning model is at least 37.9% lower than that of the fixed model. The online learning-based method show no misdiagnosis in experiments, while the fixed model-based method is misdiagnosed. Experimental results show that the proposed method is competitive in terms of accuracy, adaptability, and robustness.

The present study proposes a low-energy consumption multipoint sampler carried by a deep-sea landing vehicle (DSLV) to meet the requirements of time series sampling in local areas and location series sampling in wide areas, and an optimization method of sampling structure based on least-squares support-vector machine (LSSVM) surrogate model and a multi-objective particle swarm optimization (MOPSO) algorithm. First, the overall structure and core components, such as the multipoint sampler’s sampling structure, were designed. The optimization variables were the cone angle, sampling tube inner diameter, and sampling tube inner hole length, which were determined by considering the force with which the sampling structure penetrates the seafloor sediment. Then, the sampling process was simulated by the finite element method-smoothed particle hydrodynamics (FEM-SPH) method, while the accurate LSSVM model of force required for sampling and sampling tube volume was established. Finally, the MOPSO algorithm was used for multi-objective optimization of model parameters of sampling structure. The optimal model of sampling structure that can provide theoretical support for the optimal design of multipoint sampler effectively reduces energy consumption and improves sampling efficiency by force required for sampling 25.89% lower and sampling tube volume 34.81% higher than the original model.

AMHSs (Automated Material Handling Systems) are widely used in major Fabs (semiconductor fabrication plants). The OHT in an AMHS is responsible for handling the FOUP (Front Opening Unified Pod) within the Fabs. Due to the unidirectional track, the movement path of the OHT aims to avoid congested areas caused by operations or malfunctions as much as possible, to improve the overall FOUP handling efficiency. To do so, we propose a dynamic path planning method, MI-DDQN (Map Information Double Deep Q-Network), driven by deep reinforcement learning and based on map information. Firstly, we design and establish a map information state space model based on the core elements of the OHT path planning in the AMHS. Then, we design an OHT motion simulator to simulate the position coordinate transformation of the OHT, providing real-time coordinate update data for the OHT during the algorithm training process. We design a deep reinforcement learning algorithm structure based on map information model and a convolutional neural network model structure and use the algorithm to train the network model. Finally, the designed task generation module and OHT motion simulator are used to randomly generate the starting position and task position of the OHT during the training process to enhance the richness of the data. The addition of a “fault” OHT verifies the method’s ability to plan routes in complex road conditions such as congestion that may occur at any time.

In this paper, we introduce PortLaneNet, an optimized lane detection model specifically designed for the unique challenges of enclosed container terminal environments. Unlike conventional lane detection scenarios, this model addresses complexities such as intricate ground markings, tire crane lane lines, and various types of regional lines that significantly complicate detection tasks. Our approach includes the novel Scene Prior Perception Module, which leverages pre-training to provide essential prior information for more accurate lane detection. This module capitalizes on the enclosed nature of container terminals, where images from similar area scenes offer effective prior knowledge to enhance detection accuracy. Additionally, our model significantly improves understanding by integrating both high- and low-level image features through attention mechanisms, focusing on the critical components of lane detection. Through rigorous experimentation, PortLaneNet has demonstrated superior performance in port environments, outperforming traditional lane detection methods. The results confirm the effectiveness and superiority of our model in addressing the complex challenges of lane detection in such specific settings. Our work provides a valuable reference for solving lane detection issues in specialized environments and proposes new ideas and directions for future research.

To ensure the safety and energy efficiency of autonomous sampling operations for a deep-sea landing vehicle (DSLV), the Safety Energy-Dynamic Window Approach (SE-DWA) algorithm was proposed. The safety assessment sub-function formed from the warning obstacle zone and safety factor addresses the safety issue arising from the excessive range measurement error of forward-looking sonar. The trajectory comparison evaluation sub-function with the effect of reducing energy consumption achieves a reduction in path length by causing the predicted trajectory to deviate from the historical trajectory when encountering “U”-shaped obstacles. The pseudo-power evaluation sub-function with further energy consumption reduction ensures optimal linear and angular velocities by minimizing variables when encountering unknown obstacles. The simulation results demonstrate that compared with the Minimum Energy Consumption-DWA algorithm, the SE-DWA algorithm improves the minimum distance to an actual obstacle zone by 68% while reducing energy consumption by 11%. Both the SE-DWA algorithm and the Maximum Safety-DWA (MS-DWA) algorithm ensure operational safety with minimal distance to the actual obstacle zone, yet the SE-DWA algorithm achieves a 24% decrease in energy consumption. In conclusion, the path planned by the SE-DWA algorithm ensures not only safety but also energy consumption reduction during autonomous sampling operations by a DSLV in the deep sea.