Guangliang Li

Sensor fault diagnosis of autonomous underwater vehicle based on extreme learning machine

Autonomous underwater vehicles (AUVs) work in complex marine environments, and sensors play an important role in AUV systems. Therefore, research on sensor failure diagnosis technology is important for improving the reliability of AUV systems. In this paper, a new method combining phase space reconstruction and extreme learning machine (ELM) is proposed. This method is applied to predict sensor output to achieve sensor fault diagnosis for AUVs. The results of the simulation experiments based on sea trial data shown that the proposed method can diagnose sensor faults and recover the signal after faults occur in a period of time.

Nonlinear path following of autonomous underwater vehicle considering uncertainty

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 the horizontal plane. All regular paths are feasible. Our method takes three steps to accomplish the path following. First, the guidance law calculates the desired yaw angle. Then an adaptive compensation is added on the desired yaw which considering uncertainty and input saturation. Last, the PID controller is extended to cope with the yaw tracking and velocity control. Simulations and outfield experiments are conducted to verify the feasibility and superiority of the novel approach.

Heading control for an Autonomous Underwater Vehicle using ELM-based Q-learning

Heading control is an important part of Autonomous Underwater Vehicle (AUV) control. But its control performance is restricted to the uncertainty environments, and lack of understanding of dynamic characteristics of AUV. As a model-free method, the Q-learning achieves its control motivation by interacting with the environment and maximizing a reward, so suits the complicated applications in heading control of AUV. However, Q Learning algorithms are not competent for continuous space problems. So, Extreme Learning Machine(ELM) is proposed to guarantee the generalization performance and work with continuous states and actions. In this paper, the method of using ELM based Q Learning is proposed for heading control. The results have shown that the proposed method for heading control has good performance.

Side-scan sonar image segmentation using Kernel-based Extreme Learning Machine

Autonomous Underwater Vehicles (AUVs) are important platform for oceanographic survey. AUVs have been widely applied to many fields, such as the ocean research, oil and gas exploitation, mineral resources investigation, fishing and military. People can obtain important ocean information by segmenting, classifying and recognizing sonar image of AUV. So studying side-scan sonar image is significant. Markov Random Field (MRF) is an efficient method for segmentation of side-scan sonar image. However, MRF may not work well for side-scan sonar image obtained from complex environment. In these images, pixel values do not change obviously. In this paper, an innovative segmentation method based MRF and Kernel-based Extreme Learning Machine (K-ELM) is proposed for real side-scan sonar image segmentation. This method has been validated on the real sonar images. Experimental results demonstrate that the proposed method outperforms MRF in classification accuracy.

PCA and Kernel-based extreme learning machine for side-scan sonar image classification

As an important role of oceanographic survey, side-scan sonar image classification has attracted much attention in the past two decades. Due to the special properties of sonar image, traditional approaches are difficult to get good classification accuracy, so their implementation in real world is blocked. In this paper, a novel classification system based on kernel-based extreme learning machine (KELM) and principle component analysis (PCA) is proposed. Experimental results demonstrate that the proposed method can get better stability and higher classification accuracy than traditional approaches such as support vector machine (SVM).

Classification and mosaicking of side scan sonar image

As an underwater detection sensor, side-scan sonar plays an important role in marine survey, mineral exploration, underwater archaeology and so on. During the use of side-scan sonar, classifiication and mosaicking of collected images is essential in most cases. There are two main contributions in our work. On the one hand, we propose a supervised learning method based on kernel-based extreme learning machine (KELM) to perform image classification. As a single-hidden layer feedforward neural network, ELM has one hidden layer and one output layer. It has been proved that ELM provides better performance in classification and regression at shorter consumed time than some others, such as traditional support vector machine (SVM), without complex parameter adjustment. However, the weights of ELM hidden layer are randomly produced and the classification results of ELM are different because of this. To solve this problem, the kernel-based ELM was proposed, in which the hidden layer was processed with a kernel function to eliminate randomness. On the other hand, the side-scan sonar images and the classified image data will be geo-referenced mosaicked using positions produced by extended Kalman filter (EKF) with sensor data from an autonomous underwater vehicle (AUV). To eliminate gaps in the mosaicking images, image dilation is adopted in our work. Experimental results demonstrate that the proposed classification method works well, and the proposed image mosaicking method is applicable when concerns real side-scan sonar images.