Autonomous underwater vehicle (AUV) plays an increasingly important role in ocean exploration. Existing AUVs are usually not fully autonomous and generally limited to pre-planning or pre-programming tasks. Reinforcement learning (RL) deep reinforcement have been introduced into the AUV design research improve its autonomy. However, these methods still difficult apply directly actual system because of sparse rewards low efficiency. In this paper, we proposed a interactive method for path...

As a widely used segmentation scheme, Markov random field (MRF) utilizes k-means clustering to calculate the initial model for sidescan sonar image segmentation. However, noise and intensity inhomogeneity nature of images, results have low accuracy, motivating us use machine learning methods initialize MRF. Meanwhile, an extreme (ELM), supervised algorithm derived from single-hidden-layer feedforward neural networks, learns faster than randomly generated hidden-layer parameters is superior...

An effective autonomous underwater vehicle (AUV) simulation system can greatly improve development efficiency and reduce the cost risk of actual equipment operation. In this paper, a comprehensive is developed using Mission Oriented Operating Suite (MOOS) Unreal Engine 4 (UE4). The former provides an open-source framework application components, which are widely used in field robots. latter well-known game engine that has realistic effects various plugins. As far as we know, there few...

Autonomous underwater vehicles (AUVs) are becoming increasingly popular for ocean exploration, military and industrial applications. Motion control of AUV is the key to completing these missions. Sliding mode (SMC) method has a good performance in motion system AUV. Nevertheless, when not fully or previously known, classical techniques entirely suitable SMC tuning. Moreover, variation drag lift coefficients very sensitive speed, it difficult achieve accurate by fixed parameters at different...

Precise positioning of AUV plays an important role in the efficient and reliable underwater operation. The extended Kalman filter (EKF) is most commonly used method, because this algorithm easy to implement. However, EKF only effective for nonlinear systems with approximate linearity, then truncation error introduced. When initial state large or system model has high nonlinearity, estimation effect poor convergence rate slow. In order overcome shortcomings EKF, Ienkaran Arasaratnam Simon...

The challenge in the Autonomous Underwater Vehicle (AUV) navigation technology is how to ensure precise localization accurately. Extended Kalman filter (EKF) most widely used method. Despite its long successful application, EKF has a number of problems for instance, system motion model usually not appropriate be described as linear system. This paper proposed modified algorithm which combined Least Squares with Filter (LS-EKF). experimental results show that accuracy AUV based on LS-EKF...

Autonomous Underwater Vehicle (AUV) is a typical system which characterized by nonlinearity, uncertainty of parameters, and it also affected many external disturbances. Therefore, difficult to achieve the accurate AUV motion control, includes depth heading control so on. In order improve stability robustness AUV, this paper researches model AUV's motion, analyzes characteristics main existing methods for including PID, variable structure (VSC) fuzzy logic (FLC). Then, new method controller...

As we all known, autonomous navigation system is the basic of AUV movement, besides, accurate position estimation and reliable measurement are keys it. In order to avoid accumulating errors in long voyage by using algorithm, can depend on GPS. Due influence diving floating process, GPS data would slip. The slips affect accuracy seriously cause for path planning. Little be accepted, while can't tolerate large up several hundred meters. This paper proposed filter which combined with algorithm...

Deep reinforcement learning has achieved significant success in many fields, but will confront sampling efficiency and safety problems when applying to robot control the real world. Sim-to-real transfer was proposed make use of samples simulation overcome gap between In this paper, we focus on improving Progressive Neural Network — an effective sim-to-real method, by proposing Interactive Learning (IPNL). IPNL integrates progressive network interactive (interactive RL) which learns from...

Autonomous underwater vehicle (AUV) is developed as a powerful tool with capability of autonomous navigation to conduct benthal survey and exploration. It's kind expensive equipment experimental data obtained in one voyage also very important. So an on-board self-rescue system indispensable ensure AUV escaping from dangers rising up the surface safely. In this paper solution proposed on basis behavior-based decision-making, computed torque control fuzzy logic implement reliable operation....

With applications in the academic, military, and maintenance oil industrial Autonomous Underwater Vehicles (AUVs) are increasing important. AUV is a kind of complicated equipment, its motion control extremely crucial. At present, centralized method used by most AUVs share many deficiencies: complex controller, large computation, difficulties system upgrading, etc. This paper presents distributed System for our multi-thruster platform C-RANGER. It includes architecture algorithm. Simulation...

In order to complete nonlinear path following smoothly and accurately, this paper proposes a method utilizing the Serret-Frenet Line-of-Sight (LOS) guidance with adaptive compensation in horizontal plane. All regular paths are feasible. Our takes three steps accomplish following. First, law calculates desired yaw angle. Then an is added on which considering uncertainty input saturation. Last, PID controller extended cope tracking velocity control. Simulations outfield experiments conducted...

As an underwater detection sensor, side-scan sonar plays important role in marine survey, mineral exploration, archaeology and so on. During the use of sonar, classifiication mosaicking collected images is essential most cases. There are two main contributions our work. On one hand, we propose a supervised learning method based on kernel-based extreme machine (KELM) to perform image classification. single-hidden layer feedforward neural network, ELM has hidden output layer. It been proved...

For underwater target exploration, multiple Autonomous Underwater Vehicles (AUVs) have shown significant advantages over single AUVs. Aiming at Multi-AUV task allocation, which is an important issue for collaborative work in environments, this paper proposes a allocation method based on the Differential Evolutionary Gray Wolf Optimization (DE-GWO) algorithm. Firstly, working process of system was analyzed, and model objective function were established. Then, we combined strong global search...

In this paper, we propose a new structure that combines Grey Prediction with Line of Sight (GP-LOS) to predict yaw angle in order achieve path following for autonomous underwater vehicle (AUV). The proposed can be described by two stages. First, use grey prediction the position coordinates next time. Second, taking advantage principles Sight, obtain desired heading advance. Then take classical PID as control algorithm AUV because it has easily implemented and provide reliable performance....