Driving Profile Optimization Using a Deep Q-Network to Enhance Electric Vehicle Battery Life

In the COVID-19 era, automobiles with internal combustion engines are being replaced by eco-friendly vehicles. The demand for battery electric vehicles (BEVs) has increased explosively. Treatment of spent batteries has received much attention. Battery life can be extended via both efficient charging and driving. Consideration of the vehicles ahead when driving a BEV effectively prolongs battery life. Several studies have presented eco-friendly driving profiles for BEVs, the cited authors did not develop a BEV driving profile that considered battery life using reinforcement learning. Here, this paper presents a method of driving profile optimization that increases BEV battery life. This paper does not address how to regenerate spent batteries in an eco-friendly manner. The BEV driving profile is optimized employing a deep Q-network (a reinforcement learning method). This paper uses simulations to evaluate the effect of the driving profile on BEV battery life; these verified the applicability of our model. Finally, the speed profile optimization method was limited to improve energy efficiency or battery life in rapid speed change sections.

Pi-Score: An Estimation Strategy of the Class Prior in Positive-Unlabeled Learning for Electrical Insulator Defect Detection with Incomplete Annotations

Insulators in high-voltage power systems serve as brackets for overhead lines and prevent these wires from becoming grounded. Due to long-term exposure to a harsh environment, it is indispensable to apply periodic inspection for defective insulators, facilitating the timely overhaul of insulators. In the field of object detection, convolutional neural networks (CNNs) have been introduced and have achieved good performance. Therefore, various CNN-based detectors are applied to the insulator detection task. Because collecting ideal annotations is time-consuming and labor-intensive, research with imperfect annotations has drawn more attention. However, fewer insulator defect detection approaches account for this imperfect annotation problem. This paper focuses on a novel and challenging insulator defect detection scenario: a part of the insulators in datasets is unannotated and viewed as the background. We introduce positive-unlabeled (PU) learning to solve the problem of incomplete annotation for insulator defect detection. To further improve PU learning, a Pi-Score algorithm is proposed to estimate class prior, a crucial parameter in PU loss. Our designed framework is built on Faster R-CNN and incorporates the improved PU learning in the region proposal network. Experimental results on the Insulator Defect Image Dataset (IDID) demonstrate that the proposed framework achieved an average precision (AP) metric that is approximately 1%–2% higher than the positive–negative mainstream detectors with varying degrees of missing annotation. Meanwhile, the proposed framework obtained 0.33 and 0.47 higher AP metrics than the mainstream PU detectors with complete and half of IDID’s annotations.

Dual-Mode Pressure Sensor Integrated with Deep Learning Algorithm for Joint State Monitoring in Tennis Motion

The precise capture and identification of movement features are important for numerous scientific endeavors. In this work, we present a novel multimodal sensor, called the resistance/capacitance dual-mode (RCDM) sensor, which effectively differentiates between compression and stretchable strains during tennis motion; meanwhile, it can also accurately identify various joint movements. The proposed wearable device features a seamless design, comprising two separate components: a resistive part and a capacitive part. The resistive and capacitive components operate independently and utilize a resistance–capacitance mechanism to measure pressure and strain signals, respectively. The RCDM sensor demonstrates remarkable sensitivity to strains (GF = 7.84, 0%–140%) and exceptional linear sensitivity (S = 4.08 kPa−1) through capacitance. Utilizing machine learning algorithms, the sensor achieves a recognition rate of 97.21% in identifying various joint movement patterns. This advanced production method makes it feasible to manufacture the sensors on a large scale, offering tremendous potential for various applications, including tennis sports systems.

3D-Printed Electrocardiogram Dry Electrodes Using Four Commercially Available Polylactic Acid Conductive Filaments

Cardiovascular diseases (CVDs) are the leading cause of death worldwide. Therefore, the ability to monitor electrocardiogram (ECG) in the long term could save patients’ lives by allowing early diagnosis and intervention. Dry electrodes are the best option to accomplish this task. 3D printing is one of the popular techniques currently used to construct dry electrodes. This project designs, builds, tests, and compares 3D printed ECG electrodes using four commercially available electrically conductive polylactic acid-based fuzed deposition modeling filaments. Also, this project uses these printed electrodes to acquire ECG raw signals. Then it compares these signals by calculating their signal-to-noise ratio (SNR) and their ability to measure heart rate using the Pan–Tompkins algorithm. Also, the resistance of these printed electrodes is measured. In conclusion, all printed electrodes in this experiment show acceptable efficiency with SNR values equal to or larger than 18.89 dB, and these electrodes prove their ability to measure heart rate using the Pan–Tompkins algorithm. However, according to the results of this experiment, one commercially available brand (Proto-Pasta) is more favorable to use. Proto-Pasta has the highest SNR value (20.71 dB), the second lower resistance values (∼ from 600 Ω to 2 kΩ), the lowest 60 Hz powerline and low-frequency noise, the most accurate printed electrodes (the printed electrodes have less shape deformity compared to other brands), and no any noticeable physical malfunction (no knops break from the electrodes during the test process).

Research of GPCEMBP Glucose Prediction Algorithm Based on Continuous Glucose Monitoring

In view of the poor timeliness of dynamic blood glucose data and a delay of insulin effect for blood glucose control, and considering the nonlinearity and nonstationarity of the glucose data, a new blood Glucose Prediction algorithm combined Correlation coefficient-based complete ensemble empirical mode decomposition with adaptive noise and back propagation neural network (GPCEMBP) was proposed to increase the prediction time and improve the prediction accuracy. It refined the mode decomposition algorithm and integrated the correlation mode filter function to extract the characteristic intrinsic mode functions from the original signal. A new neural network prediction model was constructed by optimizing the number of hidden layer neurons, the number of hidden layers, activation functions, the number of inputs, structure, and other parameters. Finally, a predicted blood glucose value was calculated by phase space reconstruction technology. Through ablation and comparison experiments, it was demonstrated that the GPCEMBP algorithm had better prediction accuracy, convergence, and robustness in blood glucose prediction within 84 min. In addition, it has good adaptability to deal with different quality glucose data.

Machine Learning-Based Lung Cancer Detection Using Multiview Image Registration and Fusion

The exact lung cancer identification is a critical problem that has attracted the researchers’ attention. The practice of multiview single image and segmentation has been widely used for the last 2 years to improve the identification of lung cancer disease. The utilization of machine learning (ML) and deep learning (DL) techniques can significantly expedite the process of cancer detection and stage classification, enabling researchers to study a larger number of patients in a shorter time frame and at a reduced cost applying the image segmentation approach herein, the multiresolution rigid registration mechanism is applied to enhance the segmentation further. Techniques like principle component averaging and discrete wavelet transform are verified for the image fusion development. To review the performance of the suggested technique, the image database resource initiative-based lungs image database consortium is tested in this paper which includes 4,682 computed tomography scan images of 61 patients with nodules sizes from 3 to 30 mm. According to the study finding, the outperformed results of our model are obtained in terms of feature mutual information, and peak signal-to-noise ratio, which were recorded at 0.80 and 19.25, respectively. Moreover, the detection and stages of cancer (STG-1, STG-2, STG-3, and STG-4) of lung nodules are also assessed by using the ResNet-18 convolutional neural network classifier. With only 1.8 FP/scan, the achieved accuracy and sensitivity for detection are 98.2% and 96.4%, respectively. The study’s findings show that our proposed strategy outperforms existing models significantly. Therefore, the proposed models have the potential to be implemented in clinical settings to provide support to doctors in the early diagnosis of cancer, while minimizing the occurrence of false positives in scans.