Neural Networks in Ship Navigation

Spatiotemporal prediction of offshore wind fields based on a hybrid deep learning model for maritime navigation

To prevent the maritime traffic accidents and make scientific decision, scientific and accurate prediction of the traffic flow is useful, which has often been made by neural network. The weight updating methods have played an important role in improving the performance of neural networks. To ameliorate the oscillating phenomenon in training radial basis function (RBF) neural network, a fractional order gradient descent (GD) with momentum method for updating the weights of RBF neural network (FOGDM-RBF) is proposed. Its convergence is proved. The new algorithm is used to predict vessel traffic flow at Xiamen Port. It performs stable and converges to zero as the iteration increases. The results verify the theoretical results of the proposed algorithm such as its monotonicity and convergence. The descending curve of error values by fractional order GDM is smoother than the GD and GDM method. Error analysis shows that the algorithm can effectively accelerate the convergence speed of the GD method and improve its performance with high accuracy and validity. The influence of fractional order, number of hidden layer neurons, tide peak hours, and ship size is analyzed and compared. 1. Introduction As the world shipping becomes more and more busy, the large ship traffic flow leads to frequent maritime traffic accidents, resulting in huge economic losses. Ship traffic flow is a basic system in marine traffic engineering and an important index to measure the construction of marine traffic infrastructure. Its prediction results can provide basis for formulating scientific Port management planning and ship navigation management. Therefore, to ensure the accuracy and rationality of ship traffic flow forecasting is of great significance for improving port infrastructure construction and formulating scientific port management strategies. Many advanced artificial intelligence optimization algorithms have been used for ship traffic flow forecasting, such as artificial neural network (Zhai 2013; Zhang 2015). Neural network can deal with complex nonlinear problems and has achieved some results. However, the neural network itself has some shortcomings, such as slow learning speed, easy to fall into the local extremum, learning and memory instability, etc.

The article explores modern neural network architectures for the automatic detection and recognition of marine surface objects and obstacles of given classes throughout the full image area, applicable for execution in real or near real time on an optoelectronic vision system to au-tomate and improve the safety of civil marine navigation. A formal statement of the problem of automatic detection of objects on images is given. The state-of-the-art algorithms for detecting objects in images based on use of artificial convolutional neural networks were reviewed, their comparison was made and a reasonable choice was made in favor of the most efficient neuralnetwork architecture in terms of computational complexity to recognition accuracy. The subject area is studied, as well as publicly available databases of surface objects suitable for use in the training of algorithms using artificial neural networks. The article concluded that there is insuffi-cient labeled data for training neural network algorithms, as a result of which the authors inde-pendently collected research images and video sequences, prepared and labeled the collected data containing surface marine objects and other obstacles that represent a navigation hazard for ships. Based on the selected neural network architecture, a new neural network algorithm for automatic full-frame detection and recognition of surface objects was developed, and an artificial neural network was trained using the prepared database of images of typical objects. The resulting algorithm was tested by the authors on a validation data set, the quality of its work was estimated using various metrics, and the algorithm’s performance was measured. Conclusions are made about the necessity to expand the collected database of images of typical marine objects, further steps are proposed to improve the accuracy of the developed software and algorithmic complex and its implementation to be used in a marine optoelectronic machine vision system for automa-tion and improving the safety of civil navigation.

This study proposes a novel artificial neural network-based methodology for classifying the incident wave direction during ship navigation using the heave–roll–pitch motion response spectra as input. The proposed model demonstrated a balanced performance with an overall accuracy of approximately 0.888, effectively classifying the wave direction into three major categories: head-sea, beam-sea, and following-sea. The methodology utilizes Response Amplitude Operators derived from linear potential flow theory to generate motion response spectra, which are then used to classify the incident wave direction. The model effectively learns the frequency-distribution characteristics of the response spectrum, enabling wave direction classification without the need for complex inverse analysis procedures. This approach is significant in that it allows wave direction recognition solely based on measurable ship motion responses, without the need for additional external sensors or mathematical modeling. This data-driven approach has strong potential for integration into autonomous ship situational awareness modules and real-time wave monitoring technologies. However, the study simplified the directional domain into three representative groups, and the model was validated primarily using a numerically generated dataset, indicating the need for future improvements. Future research will expand the dataset to include a broader range of sea states, improve directional resolution, and explore continuous wave direction prediction. Additionally, further validation using field-measured data will be conducted to assess the real-time applicability of the proposed model.

Considering that collision accidents happen sometimes, it is necessary to predict the collision risk to ensure navigation safety. With the information construction in maritime and the popularity of automatic identification system application, it is more convenient to obtain ship navigation dynamics. How to obtain ship encounter dynamic parameters through automatic identification system information, assess ship collision risk, find out dangerous target ships, and give early warning and guarantee for ship navigation safety, is a problem that scholars have been studying. As an index to measure the degree of ship collision risk, CRI, namely, collision risk index, is usually obtained by calculating ship encounter parameters and comprehensive analysis. There are many factors that affect CRI, and the values of many parameters depend on expert judgment. The corresponding CRI has nonlinear and complex characteristics, which is highly correlated with the time sequence. In order to enhance the prediction accuracy and efficiency, PSO-LSTM neural network is applied in the paper to predict CRI. Experiments show that PSO-LSTM neural network can effectively predict collision risk and provide a reference for navigation safety.

Sea ice mapping plays an integral role in ship navigation and meteorological modeling in the polar regions. Numerous published studies in sea ice classification using synthetic aperture radar (SAR) have reported high classification rates. However, many of these focus on numerical results based on sample points and ignore the quality of the inferred sea ice maps. We have designed and implemented a novel SAR sea ice classification algorithm where the spatial context, obtained by the unsupervised IRGS segmentation algorithm, is integrated with texture features extracted by a residual neural network (ResNet) and, using regional pooling, classifies ice and water. This algorithm is trained and tested on a published dataset and cross-validated using leave-one-out (LOO) strategy, obtaining an overall accuracy of 99.67% and outperforming several existing algorithms. In addition, visual results show that this new method produces sea ice maps with natural ice–water boundaries and fewer ice and water errors.

This paper presents an efficient deep learning framework for long-term monitoring of acoustic events from hydrophone big data. The large-scale noisy ONC (Ocean Networks Canada) data may contain rare acoustic events, which can be automatically recognized by utilizing a deep convolutional neural network. Few works have been reported in the area of deep learning for the recognition of different kinds of marine mammal calls which however is crucial for many applications such as marine navigation. In this proposed scheme, deep learning feature sets are adopted and processed by a support vector machione (SVM) classifier. The proposed method is tested with 28685 minutes of data, spanning a single year with 5573 whale calls/acoustic events, and using a human operator's annotations. The experimental results show that the average accuracy rate of recognition using deep feature learning are 98.69% (two-class) and 94.48% (multi-class), respectively, for the proposed recognition scheme, which outperforms the MFCC-based method.

Many ocean applications such as control optimization of wave energy converters and ship navigation rely on ocean wave prediction to improve the efficiency and safety of offshore operations. The two main types of ocean wave prediction methods are deterministic sea wave prediction (DSWP) and same point prediction. DSWP are based on several upstream wave measurements, which aid in prediction but are expensive due to the amount of measuring equipment necessary. Furthermore, the majority of DSWP are physical models that work on certain sea states such as deep water condition and are based on a knowledge of ocean physics and the assumptions of linear wave theory. On the other side, same-point prediction solely uses past measurements to estimate future sea elevation, which reduces measurement costs but adds extra complexity. For same-point prediction, researchers in the existing literature apply data-driven models such as auto-regressive models or neural networks; nevertheless, neural networks do not perform better. In this paper, we examine the performance of Temporal Convolutional Networks (TCNs) and traditional Artificial Neural Networks (ANNs) for short-term ocean wave prediction on the same point wave measurement. With dilated causal convolutional layers and residual blocks, TCNs are designed to extract features from complex long-term patterns and capture the nonlinearity of the wave time series data. Wave time series data are simulated using the Pierson-Moskowitz spectrum with different levels of non-linearity for the experiment. The experiment results illustrate the efficacy of two deep learning models as influenced by nonlinearity and noise levels, but TCNs do not outperform ANNs in the same point wave prediction.

In today's extremely competitive shipping era, especially in the field of unmanned ships, the safety of ship navigation and the rapid growth of operational demand have forced ship designers to improve the automatic control performance of ships, in which ships Heading control system is most important. PID is an integral part of ship heading control system. However, the three parameters Ki, Kp and Kd of traditional PID need to be manually set and can't be adjusted in real time, which lacks flexibility. For a ship with such a large inertia device, PID control easily lead to serious overshoot and deviation. Therefore, this paper separately studies the performance of vanilla PID algorithm, fuzzy PID algorithm and back propagation (BP) neural network PID algorithm applied to unmanned ship autopilot. The three parameters can be adjusted in real time by combining Fuzzy algorithm and neural network. The simulation results show that the performance of neural network PID is better than the former two controllers in both cases of single command and continuous tracking. In the case of adjusted by a single command, the overshoot is only 41% of the PID control, and the adjustment time is only PID control 27%, so improved PID controller based on neural network has a bright future.

Shallowwater depth plays an important role in marine development, navigation safety, and environmental protection. It is an efficient and economical way to obtain water depth by remote sensing technology. At present, most empirical models based on multispectral image usually obtain water depth by the relationship between the sea surface reflectance (SSR) (a single pixel) and <italic>in situ</italic> water depth, it is a one-to-one correspondence between the reflectance and depth. However, seafloor substrate and inherent optical properties (IOP) will also have contribution to the SSR. In this article, we propose an adjacent pixels multilayer perceptron model (APMLP) model using adjacent pixels to weaken the influence of seafloor substrate and IOP.Datasets on Oahu Island (Sentinel-2B, LIDAR <italic>in situ</italic> data) and Saint Thomas Island (Sentinel-2A, LIDAR <italic>in situ</italic> data) are used to establish and verify the model. The APMLP model are also compared with the multilayer perceptron model (MLP) model, BP neural network model, and Log-ratio model. The overall root-mean-square error (RMSE) of APMLP model on Oahu Island is 0.72 m, which is much better than the other three models (MLP 1.07 m, BP 1.05 m, Log-ratio 1.52 m). Similar results are obtained from the Saint Thomas Island dataset, RMSE of APMLP model is 1.56 m, better than the other three (MLP 1.91 m, BP 1.89 m, Log-ratio 2.39 m). The study confirms that considering adjacent pixels in an artificial neural network model can effectively improve the performance of water depth retrieval.

Cruise ships navigate in a complex environment, and collision avoidance path planning is crucial. In this paper, we propose a collision avoidance path planning method for cruise ships based on the Parametrized Deep Q-Networks (PDQN) algorithm, which combines DQN and DDPG to deal with the discrete-continuous hybrid action space problem. The principle is to input the states into a network that determines the continuous action parameters, and then splice the states and parameters into a Q-network to obtain the maximum Q-value action. In the cruise ship collision avoidance path planning, the state of the cruise ship navigation environment is represented and input into the neural network; a multilayer neural network is constructed as the strategy network, which outputs the cruise ship actions based on the input state; the environmental feedback and new state are observed through the execution of the actions, and the parameters of the strategy network are updated by using the strategy gradient algorithm, etc., in order to maximize the cumulative rewards; and the optimization of the network parameters after several iterations of training is carried out, which improves the decision-making performance of the cruise ship. The experimental results show that the method can effectively deal with the complex scenarios in cruise ship navigation, provides an efficient and reliable solution for cruise ship collision avoidance path planning, and significantly improves the safety and collision avoidance ability of cruise ship navigation.

We present a survey on marine object detection based on deep neural network approaches, which are state-of-the-art approaches for the development of autonomous ship navigation, maritime surveillance, shipping management, and other intelligent transportation system applications in the future. The fundamental task of maritime transportation surveillance and autonomous ship navigation is to construct a reachable visual perception system that requires high efficiency and high accuracy of marine object detection. Therefore, high-performance deep learning-based algorithms and high-quality marine-related datasets need to be summarized. This survey focuses on summarizing the methods and application scenarios of maritime object detection, analyzes the characteristics of different marine-related datasets, highlights the marine detection application of the YOLO series model, and also discusses the current limitations of object detection based on deep learning and possible breakthrough directions. The large-scale, multiscenario industrialized neural network training is an indispensable link to solve the practical application of marine object detection. A widely accepted and standardized large-scale marine object verification dataset should be proposed.

In the 21st century marine navigation has become dominated by satellite positioning systems and automated navigational processes. Today, global navigation satellite systems (GNSS) play a central role in the process of carrying out basic navigational tasks, e.g. determining the coordinates of a vessel’s position at sea. Since satellite systems are being used increasingly more often in everyday life, the signals they send are becoming more and more prone to jamming. Therefore there is a need to search for other positioning systems and methods that would be as accurate and fast as the existing satellite systems. On the other hand, the automation process makes it possible to conduct navigational tasks more quickly. Due to the development of this technology, all kinds of navigation equipment can be used in the process of automating navigation. This also applies to marine radars, which are characterised by a relatively high accuracy that allows them to replace satellite systems in performing classic navigational tasks. By employing M-estimation methods that are used in geodesy as well as simple neural networks, a software package can be created that will aid in automating navigation and will provide highly accurate information about a given object’s position at sea by making use of radar in comparative navigation. This paper presents proposals for automating the process of determining a vessel’s position at sea by using comparative navigation methods that are based on simple neural networks and geodetic M-estimation methods.

Machine learning-based multimodal fusion recognition of passenger ship seafarers’ workload: A case study of a real navigation experiment

A time-varying ensemble model for ship motion prediction based on feature selection and clustering methods