Jyun-Yu Jhang

With the continuing development of autonomous driving systems, the drivable area segmentation problem has become an indispensable part of self-driving cars. The drivable area segmentation technology introduces many features to self-driving car technology, such as providing information about the surrounding environment, assisting decision-making mechanisms, selecting appropriate driving paths, and avoiding obstacles. In addition, accurate segmentation of drivable areas is crucial for improving self-driving navigation and avoiding obstacles. To enable effective identification of drivable areas on the basis of environmental information, this study designed a drivable area segmentation network named DASNet. The proposed DASNet utilizes depthwise separable convolution as a basis/platform for feature extraction to enable features to be efficiently extracted to reduce both the computational load and required network parameters. Additionally, the proposed DASNet enhances the inference speed of the network while maintaining high accuracy. In order to reduce point cloud density without compromising essential information, we perform sampling and fusion on the point clouds in both Cartesian and spherical coordinate spaces during data preprocessing. The fused point cloud serves as the input to DASNet, while the output is the drivable area map. Finally, the proposed DASNet is ported to a field programmable gate array to achieve real-time drivable area detection. This study employs the publicly available KITTI dataset and proposed wooded environment dataset for experimental evaluation. By evaluating the model on two distinct datasets, we aim to demonstrate the capabilities across three types of urban road scenes and more densely wooded environments. The experimental results indicate that DASNet achieved an F1-score of 0.9449 and an inference speed of 9.32 ms on the KITTI dataset. Furthermore, the proposed DASNet was applied to unmanned vehicles operating in a wooded environment, achieving an F1-score of 98.64%.

Convolutional neural networks typically employ convolutional layers for feature extraction and pooling layers for dimensionality reduction. However, conventional pooling methods often lead to a loss of critical feature information, particularly in images with diverse content, such as vehicle images. This study proposes a novel approach to address these problems: a convolutional neural network with type-2 fuzzy-based pooling (CNN-T2FP). This innovative pooling method utilizes type-2 fuzzy membership functions to effectively manage local imprecision in feature maps. Compared with type-1 fuzzy pooling, which only addresses uncertainty to a certain extent, type-2 fuzzy pooling exhibits improved adaptability to different image contents. The experimental results of this study revealed that the CNN-T2FP achieved average accuracies of 92.14% and 93.34% on two datasets, surpassing the performance of existing pooling techniques. In addition, t-distributed stochastic neighbor embedding plots and feature visualization maps further highlighted the potential of type-2 fuzzy-based pooling to overcome the limitations of conventional pooling methods and enhance the performance of convolutional neural networks in image analysis tasks.

Deep learning technology has extensive application in the classification and recognition of medical images. However, several challenges persist in such application, such as the need for acquiring large-scale labeled data, configuring network parameters, and handling excessive network parameters. To address these challenges, in this study, we developed an adversarial-learning-based Taguchi convolutional fuzzy neural classifier (AL-TCFNC) for classifying malignant and benign lung tumors displayed in computed tomography images. In the framework of the developed AL-TCFNC, a fuzzy neural classifier replaces a conventional fully connected network, thereby reducing the number of network parameters and the training duration. To reduce experimental cost and training time, the Taguchi method was used. This method helps to identify the optimal combination of model parameters through a small number of experiments. The transfer learning of models across databases often results in subpar performance because of the paucity of labeled samples. To resolve this problem, we used a combination of maximum mean discrepancy and cross-entropy for adversarial learning with the proposed model. Two data sets, namely the SPIE-AAPM Lung CT Challenge data set and LIDC-IDRI Lung Imaging Research data set, were used to validate the AL-TCFNC model. When the AL-TCFNC model was used for transfer learning, it exhibited an accuracy rate of 89.55% and outperformed other deep learning models in terms of classification performance.

In the field of machining, product quality must meet customer specifications. In general, surface roughness is an essential indicator of machining quality. Low surface roughness correlates with increased fatigue strength and corrosion resistance. However, the main factor that affects surface roughness is the selection of the machining parameters. When different parameters are combined, the resulting machining quality varies. Therefore, to achieve the desired machining quality, appropriate machining parameters must be selected. In this study, an ultrasonic-assisted machining system (UAMS) was designed to help users determine the machining parameters and machine SiC materials. To establish a prediction model for surface roughness, a novel network mapping fusion (NMF) convolutional neuro-fuzzy network (CNFN) model was used in the designed UAMS. The differential evolution algorithm was then used to search for optimized machining parameters. To explain the prediction model, which can help analyze the factors that have the greatest influence on surface roughness, a Shapley additive explanations method is proposed. The proposed NMF–CNFN model was more accurate than were the other deep learning models and exhibited a MAPE of 1.98%. When optimized machining parameters were selected, the desired surface roughness was obtained, thereby confirming the effectiveness and accuracy of the proposed UAMS. Moreover, the proposed model was implemented in a field-programmable gate array (FPGA) to reduce its power consumption and increase its computational performance. Experimental results indicated that the computational speed of the FPGA was 99.64% and 99.16% higher than those of the CPU and GPU, respectively.

In this paper, a real-time navigation control system based on lidar sensing is proposed for use in unknown environments. The proposed system comprises a behavioral controller for controlling an autonomous Ackerman robot for obstacle avoidance in the absence of global map information when moving toward a goal. The adopted obstacle avoidance method is selected by a wall-following fuzzy controller. The input parameter of this controller is the distance between the robot and the wall, which is determined by the lidar sensor, and the output parameter of the controller is the steering angle of the robot for it to reach the destination without collision. To prevent the robot from entering an endless loop, an endless loop escape mechanism is added to the proposed system. The simulation and experimental results of this study indicate that the proposed navigation control system can effectively assist an Ackerman robot to complete the navigation task successfully in unknown environments.

In this study, we proposed a type 2 fuzzy neural controller (FNC) based on Bayesian dynamic group particle swarm optimization (BDGPSO) for the navigation control of Ackermann steering vehicles. The type 2 FNC has a five-layer network architecture. The advantages of the proposed BDGPSO algorithm are that it prevents PSO from falling into a local optimum, and it uses the Bayesian algorithm to determine the optimal inertia weight and learning factor combination. To evaluate the proposed type 2 FNC on the basis of the distance information returned by lidar, we used a new fitness function from the reinforcement-learning strategy. In an unknown testing environment, the fitness value and time required for the proposed type 2 FNC with angular and linear velocity outputs were 0.973510 and 25.555 s, respectively. These values indicated that the proposed approach outperforms other wall-following control methods. In addition, the proposed controller could successfully control the navigation of an Ackerman steering vehicle in an unknown actual environment.

With the advancement of science and technology, the development and application of unmanned mobile vehicles (UMVs) have emerged as topics of crucial concern in the global industry. The development goals and directions of UMVs vary according to their industrial uses, which include navigation, autonomous driving, and environmental recognition; these uses have become the priority development goals of researchers in various fields. UMVs employ sensors to collect environmental data for environmental analysis and path planning. However, the analysis function of a single sensor is generally affected by natural environmental factors, resulting in poor identification results. Therefore, this study introduces fusion technology that employs heterogeneous sensors in the Ackerman UMV, leveraging the advantages of each sensor to enhance accuracy and stability in environmental detection and identification. This study proposes a fusion technique involving heterogeneous imaging and LiDAR (laser imaging, detection, and ranging) sensors in an Ackerman UMV. A camera is used to obtain real-time images, and YOLOv4-tiny and simple online real-time tracking are then employed to detect the location of objects and conduct object classification and object tracking. LiDAR is simultaneously used to obtain real-time distance information of detected objects. An inertial measurement unit is used to gather odometry information to determine the position of the Ackerman UMV. Static maps are created using simultaneous localization and mapping. When the user commands the Ackerman UMV to move to the target point, the vehicle control center composed of the robot operating system activates the navigation function through the navigation control module. The Ackerman UMV can reach the destination and instantly identify obstacles and pedestrians when in motion.

With the rapid pace of urbanization, the number of vehicles traveling between cities has increased significantly. Consequently, many traffic-related problems have emerged, such as traffic jams and excessive numbers and types of vehicles. To solve traffic problems, road data collection is important. Therefore, in this paper, we develop an intelligent traffic-monitoring system based on you only look once (YOLO) and a convolutional fuzzy neural network (CFNN), which record traffic volume, and vehicle type information from the road. In this system, YOLO is first used to detect vehicles and is combined with a vehicle-counting method to calculate traffic flow. Then, two effective models (CFNN and Vector-CFNN) and a network mapping fusion method are proposed for vehicle classification. In our experiments, the proposed method achieved an accuracy of 90.45% on the Beijing Institute of Technology public dataset. On the GRAM-RTM data set, the mean average precision and F-measure (F1) of the proposed YOLO-CFNN and YOLO-VCFNN vehicle classification methods are 99%, superior to those of other methods. On actual roads in Taiwan, the proposed YOLO-CFNN and YOLO-VCFNN methods not only have a high F1 score for vehicle classification but also have outstanding accuracy in vehicle counting. In addition, the proposed system can maintain a detection speed of more than 30 frames per second in the AGX embedded platform. Therefore, the proposed intelligent traffic monitoring system is suitable for real-time vehicle classification and counting in the actual environment.

When a machine tool is used for a long time, its bearing experiences wear and failure due to heat and vibration, resulting in damage to the machine tool. In order to make the machine tool stable for processing, this paper proposes a smart bearing diagnosis system (SBDS), which uses a gradient-weighted class activation mapping (Grad-CAM)-based convolutional neuro-fuzzy network (GC-CNFN) to detect the bearing status of the machine tool. The developed GC-CNFN is composed of a convolutional layer and neuro-fuzzy network. The convolutional layer can automatically extract vibration signal features, which are then classified using the neuro-fuzzy network. Moreover, Grad-CAM is used to analyze the attention of the diagnosis model. To verify the performance of bearing fault classification, the 1D CNN (ODCNN) and improved 1D LeNet-5 (I1DLeNet) were adopted to compare with the proposed GC-CNFN. Experimental results showed that the proposed GC-CNFN required fewer parameters (20K), had a shorter average calculation time (117.7 s), and had a higher prediction accuracy (99.88%) in bearing fault classification. The proposed SBDS can not only accurately classify bearing faults, but also help users understand the current status of the machine tool.