To understand surrounding scenes accurately, the semantic segmentation of images is vital in autonomous driving tasks, such as navigation, and route planning. Currently, convolutional neural networks (CNN) are widely employed in semantic segmentation to perform precise prediction in the dense pixel level. A recent trend in network design is the stacking of small convolution kernels. In this work, small convolution kernels (3 × 3) are decomposed into complementary convolution kernels (1 × 3 + 3 × 1, 3 × 1 + 1 × 3), the complementary small convolution kernels perform better in the classification and location tasks of semantic segmentation. Subsequently, a complementary convolution residual network (CCRN) is proposed to improve the speed and accuracy of semantic segmentation. To further locate the edge of objects precisely, A coupled Gaussian conditional random field (G-CRF) is utilized for CCRN post-processing. Proposal approach achieved 81.8% and 73.1% mean Intersection-over-Union (mIoU) on PASCAL VOC-2012 test set and Cityscapes test set, respectively.

Data centers are fundamental facilities that support high-performance computing and large-scale data processing. To guarantee that a data center can provide excellent properties of expanding and routing, the interconnection network of a data center should be designed elaborately. Herein, we propose a novel structure for the interconnection network of data centers that can be expanded with a variable coefficient, also known as a variable expanding structure (VES). A VES is designed in a hierarchical manner and built iteratively. A VES can include hundreds of thousands and millions of servers with only a few layers. Meanwhile, a VES has an extremely short diameter, which implies better performance on routing between every pair of servers. Furthermore, we design an address space for the servers and switches in a VES. In addition, we propose a construction algorithm and routing algorithm associated with the address space. The results and analysis of simulations verify that the expanding rate of a VES depends on three factors: n, m, and k where the n is the number of ports on a switch, the m is the expanding speed and the k is the number of layers. However, the factor m yields the optimal effect. Hence, a VES can be designed with factor m to achieve the expected expanding rate and server scale based on the initial planning objectives.

With the rapid development of control technology, increasing applications are using model predictive control (MPC) for deviation correction in vertical drilling. However, the accuracy of the predictive model is affected by the uncertainty of the stratum, which results in model mismatch and a reduction in control performance. In this paper, an intelligent compensating method is proposed for MPC-based deviation correction with stratum uncertainty in a vertical drilling process to increase control accuracy. First, a trajectory extension model is introduced as the predictive model for MPC, and the uncertainty of the stratum is discussed. Then, the compensation for the MPC is acquired based on a Gaussian fitting method and hybrid bat algorithm. Finally, based on the actual drilling data, a simulation is performed to demonstrate the effectiveness of the proposed method.

This paper concerns with trajectory azimuth control in directional drilling. The motion process of the drill bit and a series of stabilizers are described, and a state-space model of the trajectory azimuth is constructed. The scheme of the trajectory azimuth control system is designed based on the equivalent input disturbance approach. An internal model is inserted to track the drill bit to improve the quality of the drilling trajectory. A state observer is combined with a low-pass filter to estimate the trajectory azimuth by measuring the azimuth of the bottom hole assembly (BHA). The control parameters can be obtained by the condition of system stability, which is derived in terms of linear matrix inequalities. A typical case is used to illustrate the validity and robustness of our approach.

Image recognition is one of the key technologies for worker’s helmet detection using an unmanned aerial vehicle (UAV). By analyzing the image feature extraction method for workers’ helmet detection based on convolutional neural network (CNN), a double-channel convolutional neural network (DCNN) model is proposed to improve the traditional image processing methods. On the basis of AlexNet model, the image features of the worker can be extracted using two independent CNNs, and the essential image features can be better reflected considering the abstraction degree of the features. Combining a traditional machine learning method and random forest (RF), an intelligent recognition algorithm based on DCNN and RF is proposed for workers’ helmet detection. The experimental results show that deep learning (DL) is closely related to the traditional machine learning methods. Moreover, adding a DL module to the traditional machine learning framework can improve the recognition accuracy.

Laser propulsion is expected to be the next-generation propulsion mechanism. In particular, metal-free water cannon realizes propulsion without a metallic target. In this study, we develop a numerical simulation code using the C-CUP (CIP and Combined, Unified Procedure) method to simulate a laser-induced bubble and a metal-free water cannon. We successfully reproduced the qualitative behavior of spouting water in a three-dimensional space when the metal-free water cannon is irradiated by laser. Furthermore, the calculated results for the time development of displacement of the metal-free water cannon agree qualitatively with the experimental results. We simulate the behavior of the laser-induced bubble and discovered that the bubble inhales the water once spouted out, and the target moves backward owing to the pressure difference generated by the bubble expansion as well as collapsing and inhaling actions. Furthermore, the laser-induced bubble repeats the expansion and collapse, and the target moves forward while it oscillates.

Vibration suppression in servo systems is significant in high-precision motion control. This paper describes a vibration-suppression method based on input shaping and adaptive model-following control. First, a zero vibration input shaper is used to suppress the vibration caused by an elastic load to obtain an ideal position output. Then, a configuration that combines input shaping with model-following control is developed to suppress the vibration caused by changes of system parameters. Finally, analyzing the percentage residual vibration reveals that it is effective to employ the sum of squared position error as a criterion. Additionally, a golden-section search is used to adjust the parameters of a compensator in an online fashion to adapt to the changes in the vibration frequency. A comparison with other input shaper methods shows the effectiveness and superiority of the developed method.

Hybrid A^* algorithm has been widely used in mobile robots to obtain paths that are collision-free and drivable. However, the outputs of hybrid A^* algorithm always contain unnecessary steering actions and are close to the obstacles. In this paper, the artificial potential field (APF) concept is applied to optimize the paths generated by the hybrid A^* algorithm. The generated path not only satisfies the non-holonomic constraints of the vehicle, but also is smooth and keeps a comfortable distance to the obstacle at the same time. Through the robot operating system (ROS) platform, the path planning experiments are carried out based on the hybrid A^* algorithm and the improved hybrid A^* algorithm, respectively. In the experiments, the results show that the improved hybrid A^* algorithm greatly reduces the number of steering actions and the maximum curvature of the paths in many different common scenarios. The paths generated by the improved algorithm nearly do not have unnecessary steering or sharp turning before the obstacles, which are safer and smoother than the paths generated by the hybrid A^* algorithm for the autonomous ground vehicle.

This study shows that a generalized fuzzy c-means (gFCM) clustering algorithm, which covers both standard and exponential fuzzy c-means clustering, can be constructed if a given fuzzified function, its derivative, and its inverse derivative can be calculated. Furthermore, our results show that the fuzzy classification function for gFCM exhibits a behavior similar to that of both standard and exponential fuzzy c-means clustering.

In this paper, a novel method is presented to improve the speed-sensorless control performance of an interior permanent magnet synchronous motor using a nonsingular fast terminal sliding-mode observer and fractional-order software phase-locked loop. The interior permanent magnet synchronous motor system is first described. Next, a nonsingular fast terminal sliding mode observer is constructed to estimate the d-q-axis back electromotive force. The speed and position of the rotor are then accurately tracked using a fractional-order software phase-locked loop. The effectiveness and feasibility are verified through a simulation in MATLAB/Simulink. The results show an excellent performance despite a fluctuation in speed and torque ripple.

In this paper, a Hadoop-based big data system for auto body precision is established. The system unifies the elements that affect auto body precision into a big data platform, which is more efficient than traditional management methods. Using big data analysis, we devised algorithms to improve the efficiency and accuracy of body precision monitoring. Furthermore, we developed techniques to analyze complex dimension deviation problems using a correlation analysis method, principal component analysis (PCA), and improved PCA method. We further established failure modes and devised monitoring and diagnosis models based on time series analysis.

To solve uncertainties in industrial processes, interval kernel principal component analysis (IKPCA) has been proposed based on symbolic data analysis. However, it is experimentally discovered that the performance of IKPCA is worse than that of other algorithms. To improve the IKPCA algorithm, interval ensemble kernel principal component analysis (IEKPCA) is proposed. By optimizing the width parameters of the Gaussian kernel function, IEKPCA yields better performances. Ensemble learning is incorporated in the IEKPCA algorithm to build submodels with different width parameters. However, the multiple submodels will yield a large number of results, which will complicate the algorithm. To simplify the algorithm, a Bayesian decision is used to convert the result into fault probability. The final result is obtained via a weighting strategy. To verify the method, IEKPCA is applied to the Tennessee Eastman (TE) process. The false alarm rate, fault detection rate, accuracy, and other indicators used in the IEKPCA are compared with those of other algorithms. The results show that the IEKPCA improves the accuracy of uncertain nonlinear process monitoring.

An ameliorated Frenet trajectory optimization (AFTO) method based on artificial emotion (AE) and an equilibrium optimizer (EO) is proposed for the local trajectory planning of an unmanned ground vehicle (UGV). An artificial emotional potential field (AEPF) model is established to simulate AE. To realize a humanoid driving mode with emotional intelligence, AE is introduced into the Frenet trajectory optimization (FTO) method to determine the optimal trajectory. Based on the optimal discrete goal state of the FTO method, a first-sampling-then-optimization (FSTO) framework combining the FTO method with the EO is designed to obtain the optimal trajectory in a continuous goal state space. With different AEPF levels corresponding to different types of obstacles, simulation results show that the AEPF effectively adjusts the trajectory into different levels of safe distance between the UGV and obstacles corresponding to the humanoid driving mode. From the results of 30 independent experiments based on the AEPF, the FSTO framework in the AFTO method is effective for optimizing the trajectory of the FTO method at a lower cost. Moreover, the effectiveness of the proposed method for different types of roads is verified on a straight road and a curved road with obstacles in simulation. The improvement based on emotional intelligence and trajectory optimization in the AFTO method provides a humanoid driving mode for the UGV in the continuous goal state space.

The randomness of path generation and slow convergence to the optimal path are two major problems in the current rapidly exploring random tree (RRT) path planning algorithm. Herein, a novel reinforcement-learning-based hybrid bidirectional rapidly exploring random tree (H-BRRT) is presented to solve these problems. To model the random exploration process, a target gravitational strategy is introduced. Reinforcement learning is applied to the improved target gravitational strategy using two operations: random exploration and target gravitational exploration. The algorithm is controlled to switch operations adaptively according to the accumulated performance. It not only improves the search efficiency, but also shortens the generated path after the proposed strategy is applied to a bidirectional rapidly exploring random tree (BRRT). In addition, to solve the problem of the traditional RRT continuously falling into the local optimum, an improved exploration strategy with collision weight is applied to the BRRT. Experimental results implemented in a robot operating system indicate that the proposed H-BRRT significantly outperforms alternative approaches such as the RRT and BRRT. The proposed algorithm enhances the capability of identifying unknown spaces and avoiding local optima.

An improved fully convolutional network based on post-processing with global variance (GV) equalization and noise-aware training (PN-FCN) for speech enhancement model is proposed. It aims at reducing the complexity of the speech improvement system, and it solves overly smooth speech signal spectrogram problem and poor generalization capability. The PN-FCN is fed with the noisy speech samples augmented with an estimate of the noise. In this way, the PN-FCN uses additional online noise information to better predict the clean speech. Besides, PN-FCN uses the global variance information, which improve the subjective score in a voice conversion task. Finally, the proposed framework adopts FCN, and the number of parameters is one-seventh of deep neural network (DNN). Results of experiments on the Valentini-Botinhaos dataset demonstrate that the proposed framework achieves improvements in both denoising effect and model training speed.

Influenza outbreaks can be effectively prevented if further outbreaks are predicted as early as possible. This article proposes an autoregressive integrated moving average (ARIMA) model and a Holt-Winters exponential smoothing (HWES) model to analyze tweet data for predicting influenza outbreaks and to visualize the number of flu-infection-related tweets with heat maps. First, textual influenza data for Australia from June 2015 to June 2017 are collected through the Twitter Application Programming Interface (API). Next, the ARIMA and HWES models are applied to predict the difference between the flu tweets and confirmations from the Centers for Disease Control and Prevention. Finally, a visualized heat map based on influenza topics validates the modeling analysis in two different time zones. The results show that the average relative error of the ARIMA (HWES) model is 7.25% (11.29%) for the one-week flu forecast.