With the rapid development of wind farm worldwide, monitoring the status of numerous wind turbines becomes the essential work. Abnormal data in wind power curve (WPC) are quite important for wind farm operations and maintenances because they usually reveal wind turbine failures or some extreme conditions. This paper proposes a new algorithm of WPC abnormal data detection and cleaning by image thresholding based on minimization of dissimilarity-and-uncertainty-based energy (MDUE). The basic idea is to transform the scatter data into a digital image and the problem of data cleaning is turned into an image segmentation problem. For all data pixels, the confidences of being classified as normal class are computed and make up a grey level feature image. Then the optimum threshold is determined by searching through the energy space based on intensity-based class uncertainty and shape dissimilarity. Finally, the normal and three types of abnormal data are marked after applying image thresholding to the feature image. The algorithm is compared with several data-based algorithms and a recently published image-based algorithm. A large number of experiments conducted on real-world WPC data collected from 37 wind turbines in two wind farms verified the superior performance of the proposed method.

The gait phase classification method is a key technique to control an exoskeleton robot. Different people have different gait features while wearing an exoskeleton robot due to the gap between the exoskeleton and the wearer and their operation habits, such as the correspondence between the joint angle and the moment at which the foot contacts the ground, the amplitude of the joint angle and others. In order to enhance the performance of the gait phase classification in an exoskeleton robot using only the angle of hip and knee joints, a kernel recursive least-squares (KRLS) algorithm is introduced to build a gait phase classification model. We also build an assist torque predictor based on the KRLS algorithm in this work considering the adaptation of unique gait features. In this paper, we evaluate the classification performance of the KRLS model by comparing with two other commonly used gait recognition methods—the multi-layer perceptron neural network (MLPNN) method and the support vector machine (SVM) algorithm. In this experiment, the training and testing datasets for the models built by KRLS, MLPNN and SVM were collected from 10 healthy volunteers. The gait data are collected from the exoskeleton robot that we designed rather than collected from the human body. These data depict the human-robot coupling gait that includes unique gait features. The KRLS classification results are in average 3% higher than MLPNN and SVM. The testing average accuracy of KRLS is about 86%. The prediction results of KRLS are twice as good as MLPNN in assist torque prediction experiments. The KRLS performs in a good, stable, and robust way and shows model generalization abilities.

Multi-stream signals are increasingly being used in robot-assisted rehabilitation training, where the timely and accurate prediction of a patient's motor intentions is frequently required to provide simultaneous and proportional control strategies. However, existing methods for motion intent prediction typically isolate signals from different modalities, resulting in a loss of signal traceability and interpretability. Therefore, we propose a multi-stream signal-fusion strategy based on knowledge tracing (MSKT). First, we collected surface electromyography, force myography, and vibroarthrography signals from the lower extremity and used different encoders to map these signals into a problem component, concept component, and identity token component. The process of human learning was then simulated using a knowledge-tracing strategy to reconstruct these components into new features. Finally, the reconstructed set of depth features was mapped to joint angles. We validated the performance of the proposed method using motion data from eight participants in different locomotion modes and compared the results to those of a temporal convolution network and long short-term memory networks. The results indicate that the MSKT achieves the lowest root-mean-squared error. This study demonstrates that the proposed knowledge-tracing-based strategy enables accurate continuous estimation for lower-limb exoskeleton rehabilitation robots.

A fast‐running human detection system for the unmanned aerial vehicle (UAV) based on optical flow and deep convolution networks is proposed in this study. In the system, running humans can be detected in real‐time at the speed of 15 frames per second (fps) with an 81.1% detection accuracy. To fast locate the candidate targets, optical flow representing the motion information is calculated with every two successive frames. A series of prior‐processing operations, including spatial average filtering, morphological expansion and outer contour extraction, are performed to extract the regions of interest. A classification model based on small‐kernel convolution networks is proposed to achieve the accurate recognition of the running people in various backgrounds. In the model, small convolutional filters are adopted to accelerate the speed of the data representation. Moreover, a total of 60,000 samples are collected to enhance the robustness of the model to adapt to the complex outdoor UAV scenes. The proposed method is compared with other deep learning frameworks for object detection. Field experiments on UAV videos are performed to verify that the proposed system can effectively detect the running people targets in real‐time.

Dangerous driving behavior is the leading factor of road traffic accidents; therefore, how to predict dangerous driving behavior quickly, accurately, and robustly has been an active research topic of traffic safety management in the past decades. Previous works are focused on learning the driving characteristic of drivers or depended on different sensors to estimate vehicle state. In this paper, we propose a new method for dangerous driving behavior prediction by using a hybrid model consisting of cloud model and Elman neural network (CM-ENN) based on vehicle motion state estimation and passenger’s subjective feeling scores, which is more intuitive in perceiving potential dangerous driving behaviors. To verify the effectiveness of the proposed method, we have developed a data acquisition system of driving motion states and apply it to real traffic scenarios in ShenZhen city of China. Experimental results demonstrate that the new method is more accurate and robust than classical methods based on common neural network.

The indoor environment has brought new challenges for micro Unmanned Aerial Vehicles (UAVs) in terms of their being able to execute tasks with high positioning accuracy. Conventional positioning methods based on GPS are unreliable, although certain circumstances of limited space make it possible to apply new technologies. In this paper, we propose a novel indoor self-positioning system of UAV based on a heterogeneous sensing system, which integrates data from a structured light scanner, ultra-wideband (UWB), and an inertial navigation system (INS). We made the structured light scanner, which is composed of a low-cost structured light and camera, ourselves to improve the positioning accuracy at a specified area. We applied adaptive Kalman filtering to fuse the data from the INS and UWB while the vehicle was moving, as well as Gauss filtering to fuse the data from the UWB and the structured light scanner in a hovering state. The results of our simulations and experiments demonstrate that the proposed strategy significantly improves positioning accuracy in motion and also in the hovering state, as compared to using a single sensor.