Average Computation Time Research Articles

An Online Data-Driven Evolutionary Algorithm–Based Optimal Design of Urban Stormwater-Drainage Systems

To reduce flood risk in urban areas, an optimal design of drainage networks for urban areas is essential for flooding control and drainage management of the urban stormwater drainage system (USDS). A conventional design is generally used for drainage networks, resulting in high computational costs and limited flooding reduction effect. In this study, based on an on-line data-driven evolutionary algorithm coupled with the storm water management model (SWMM), a novel approach for USDS drainage network optimization design was developed. A case in Xi’an City, China, was then selected for practical implementation, where the performances of the local planning scheme, the particle swarm optimization algorithm (PSO) and the proposed approach were compared. Results confirmed that our proposed methodological approach is feasibility and highly efficiency, leading to a 32% reduction in total flooding from that resulting from the local planning scheme. In addition, the average computational time was reduced by 57%, while the flooding control effect was better, compared to PSO algorithm optimization. These results suggest that our optimization design approach is reliable and applicable, and can benefit and assist designers in practice.

Real-Time Data-Driven Force and Torque Modeling on a 2-D Halbach Array by a Symmetric Coil Considering End Effect

Halbach arrays have been widely implemented in magnetic levitation systems. They concentrate the magnetic field over one side, and their magnetic flux density in the central region can be modeled. This article proposes a novel lookup table-free data-driven method to model the force and torque on a 2-D Halbach array when the magnet array is above a stationary coil. One critical contribution of this article is the rapid and accurate force and torque modeling when the coil is located near the edge of the 2-D magnet array. First, the area under the 2-D Halbach array is divided into a central region, edge regions, and corner regions. Then, nonlinear force and torque components are identified and modeled with a modified third-order Fourier series for force and torque modeling. The proposed model is implemented on an industrial server, and the average computation time is measured to be <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$2.3 ~\mu \text{s}$ </tex-math></inline-formula> . This model is universal for symmetric coils, verified through experimental measurements with circular and square coils of different sizes. The measurement result agrees well with the proposed modeling method, showing that the proposed method is suitable for real-time control of the ironless planar motor.

A Method Noise-Based Convolutional Neural Network Technique for CT Image Denoising

Medical imaging is a complex process that capitulates images created by X-rays, ultrasound imaging, angiography, etc. During the imaging process, it also captures image noise during image acquisition, some of which are extremely corrosive, creating a disturbance that results in image degradation. The proposed work addresses the challenge to eliminate the corrosive Gaussian additive white noise from computed tomography (CT) images while preserving the fine details. The proposed approach is synthesized by amalgamating the concept of method noise with a deep learning-based framework of a convolutional neural network (CNN). The corrupted images are obtained by explicit addition of Gaussian additive white noise at multiple noise variance levels (σ = 10, 15, 20, 25). The denoised images obtained are then evaluated according to their visual quality and quantitative metrics, such as peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM). These metrics for denoised CT images are then compared with their respective values for the reference CT image. The average PSNR value of the proposed method is 25.82, the average SSIM value is 0.85, and the average computational time is 2.8760. To better understand the proposed approach’s effectiveness, an intensity profile of denoised and original medical images is plotted and compared. To further test the performance of the proposed methodology, the results obtained are also compared with that of other non-traditional methods. The critical analysis of the results shows the commendable efficiency of the proposed methodology in denoising the medical CT images corrupted by Gaussian noise. This approach can be utilized in multiple pragmatic areas of application in the field of medical image processing.

Enhanced EKF and SVSF for state of charge estimation of Li‐ion battery in electric vehicle using a fuzzy parameters model

The precision of equivalent circuit model (ECM)-based state of charge (SoC) estimation methods is vulnerable to the variation of the battery parameters, due to several internal and external factors. In this regard, this study proposes a fuzzy logic method for the approximate estimation of the ECM parameters at different temperatures and SoC levels. The fuzzy inference system is designed to handle the non-linear deviation of the battery parameters from their reference values. On this basis, the extended Kalman filter and smooth variable structure filter are used to estimate the SoC. The two algorithms with fuzzy parameters (FP), namely FP-EKF and FP-SVSF, are tested on a 20 Ah Nickel Manganese Cobalt cell with maximum voltage of 4.2 V. The results show that the maximum root mean square error (RMSE) of the estimated SoC is kept within 1.51% with the FP-EKF and 0.68% with the FP-SVSF. Moreover, the reduction of the maximum absolute error may reach 0.34% with the FP-EKF, and 0.82% with the FP-SVSF, compared to the same algorithms without the proposed FP method. The executable codes are implemented on a low-cost controller, and the average computational time is obtained as 215 μs, which confirms the real-time practicality of the proposed method.

Fast Deformable Image Registration for Real-Time Target Tracking During Radiation Therapy Using Cine MRI and Deep Learning

We developed a deep learning (DL) model for fast deformable image registration using 2-dimensional sagittal cine magnetic resonance imaging (MRI) acquired during radiation therapy and evaluated its potential for real-time target tracking compared with conventional image registration methods. Our DL model uses a pair of cine MRI images as input and provides a motion vector field (MVF) as output. The MVF is then applied to align the input images. A retrospective study was conducted to train and evaluate our model using cine MRI data from patients undergoing treatment for abdominal and thoracic tumors. For each treatment fraction, MR-linear accelerator delivery log files, tracking videos, and cine image files were analyzed. Individual MRI frames were temporally sampled to construct a large set of image registration pairs used to evaluate multiple methods. The DL model was optimized using 5-fold cross validation, and model outputs (transformed images and MVFs) using test set images were saved for comparison with 3 conventional registration methods (affine, b-spline, and demons). Evaluation metrics were 3-fold: (1) registration error, (2) MVF stability (both spatial and temporal), and (3) average computation time. We analyzed >21 hours of cine MRI (>629,000 frames) acquired during 86 treatment fractions from 21 patients. In a test set of 10,320 image registration pairs, DL registration outperformed conventional methods in both registration error (affine, b-spline, demons, DL; root mean square error: 0.067, 0.040, 0.036, 0.032; paired t test demons vs DL: t[20]=4.2, P < .001) and computation time per frame (51, 1150, 4583, 8 ms). Among deformable methods, spatial stability of resulting MVFs was comparable; however, the DL model had significantly improved temporal consistency. DL-based image registration can leverage large-scale MR cine data sets to outperform conventional registration methods and is a promising solution for real-time deformable motion estimation in radiation therapy.

Klasifikasi Jenis Mangga Apel Menggunakan Metode K-Means Klustering

Introduction: The rapid development of technology is pushing it further, making people more comfortable in many fields, including industry. Mango can be processed into various types of food. Using sweets, various processed products, and various mangoes. One example of the current impact of technology on the industrial sector is the potential for systems to self-study (automatically) like humans. This is a process known as an artificial neural network. Purpose: Knowing and analyzing the classification of mango species using the K-Means Clustering method. Methods: This study uses the K-means clustering method, in which the system is built by applying an artificial neural network to the modeling and extraction of RGB values and standard RGB matrices, circumference, area, length, width, shape, and slenderness. Results: Based on the experimental results, the computation time for the Mango image required for the feature extraction process for each dataset is on average 0.85 seconds, and the computational time for training data on the test data is an average of 0.006 seconds. Conclusion: In this study, it can be concluded that the average computation time of mango image for each dataset is 0.856 seconds. The use of 1 hidden layer is more effective with the highest accuracy and the fastest time than using 2 hidden layers.

Real-Time Internal Motion Monitoring during Radiation Therapy Using Cine MRI and Deep Learning

<h3>Purpose/Objective(s)</h3> Fast and accurate internal motion estimation is fundamental to enabling real-time MRI-guided radiation therapy (RT). We hypothesized that a deep learning (DL) model for fast deformable image registration using 2D sagittal cine MRI is feasible. Herein, we report performance of a DL-based image registration model as compared to conventional image registration methods. <h3>Materials/Methods</h3> Our DL model uses a pair of cine MRI images acquired using an MR-LINAC as inputs and outputs a motion vector field (MVF) which is applied to align the images. Cine delivery data for patients (pts) undergoing treatment for abdominal and thoracic tumors were retrospectively analyzed. Cine MRI scans were sampled, constructing a large set of image registration pairs capturing pt respiratory motion to train the DL model. Data were partitioned by pt using five-fold cross validation. Model outputs (transformed images and MVFs) on test set images were saved for comparison with three conventional registration methods (affine, b-spline, and demons). Performance was quantified via average registration error and computation time. Paired t-tests determined whether differences in performance were statistically significant. <h3>Results</h3> We analyzed >21 hours of cine MRI (>629,000 frames) acquired during 86 treatment fractions from 21 pts. In a test set of 10,320 image registration pairs, DL registration outperformed conventional methods in both registration error (affine, b-spline, demons vs. DL; RMSE: 0.067, 0.040, 0.036 vs. 0.032; paired t-test demons vs DL: t(20) = 4.2, p<0.001) and computation time per frame (51 ms, 1150 ms, 4583 ms vs. 8 ms). <h3>Conclusion</h3> DL-based image registration can leverage large-scale cine MRI datasets to learn to predict pt internal motion fields using a single reference and a single moving frame as inputs. Low registration error and fast computation speed indicate that DL-based registration is a promising solution for real-time deformable motion tracking in RT. Further inter-institutional validation is underway.

Deep Learning-Based Water-Fat Separation from Dual-Echo Chemical Shift-Encoded Imaging.

Conventional water–fat separation approaches suffer long computational times and are prone to water/fat swaps. To solve these problems, we propose a deep learning-based dual-echo water–fat separation method. With IRB approval, raw data from 68 pediatric clinically indicated dual echo scans were analyzed, corresponding to 19382 contrast-enhanced images. A densely connected hierarchical convolutional network was constructed, in which dual-echo images and corresponding echo times were used as input and water/fat images obtained using the projected power method were regarded as references. Models were trained and tested using knee images with 8-fold cross validation and validated on out-of-distribution data from the ankle, foot, and arm. Using the proposed method, the average computational time for a volumetric dataset with ~400 slices was reduced from 10 min to under one minute. High fidelity was achieved (correlation coefficient of 0.9969, error of 0.0381, SSIM of 0.9740, pSNR of 58.6876) and water/fat swaps were mitigated. I is of particular interest that metal artifacts were substantially reduced, even when the training set contained no images with metallic implants. Using the models trained with only contrast-enhanced images, water/fat images were predicted from non-contrast-enhanced images with high fidelity. The proposed water–fat separation method has been demonstrated to be fast, robust, and has the added capability to compensate for metal artifacts.

Building height calculation for an urban area based on street view images and deep learning

AbstractThe building heights of an urban area are useful for space analysis, urban planning, and city management. To this end, a novel method for building height calculation for an urban area is proposed based on street view images and a deep learning model, that is, mask region‐based convolutional neural network (Mask R‐CNN). First, a spider of street view maps was developed, and an optimization model for observation locations was designed based on a genetic algorithm, by which the street view images of all buildings can be obtained with the minimum number of downloads. Subsequently, a deep learning workflow was designed based on the Mask R‐CNN to detect buildings from the panorama images. Finally, an accurate height calculation model considering repeated detection of buildings was developed by mapping between detected buildings and actual buildings. Case studies indicate that the mean error of height calculation is 0.78 m, which achieves high precision for calculating building heights in urban areas, while the average calculation time is 4.57 s per building, which indicates that the proposed method is efficient for the application in urban areas.

Assessing mixed-integer-based heat pump modeling approaches for model predictive control applications in buildings

Model predictive control can reduce heating systems’ operating costs and energy consumption. This especially applies to heat pumps, whose operation efficiency highly depends on the heating system’s source and sink temperatures. In literature, process models for heat pumps mostly introduce the coefficient of performance as a parameter thus assuming constant supply temperatures. However, supply temperature adjustment is crucial to overcharge buffer storages and hence shift energy to more favorable times, which is the basis of model predictive control concepts. We close this gap by developing two different air-source heat pump modeling approaches using the supply temperature as control variable: a piecewise linear model based on simulation results and a quadratic modeling approach. Both methods are benchmarked with a simplified linear model representing the state of research. The simplified linear model underestimates the cost of storage charging as it neglects the supply temperature’s influence on the coefficient of performance and results in 8.7% higher operating costs and 12.1% higher energy demand than the piecewise linear approach. However, the simplified linear model yields the lowest average computation time compared to the piecewise and quadratic approaches. The quadratic approach results in both lower operating costs (-1.9%) and energy demand (-1.8%), as well as lower computation times than the piecewise approach, consequently representing the best trade-off between performance and computational effort. We recommend future work to apply the method to a ground- or water-source heat pump and to a coupled building energy system to investigate the transferability of our findings.

Development of an American Sign Language Recognition System using Canny Edge and Histogram of Oriented Gradient

Sign language is used by people who have hearing and speaking difficulties, but not understood by many without these difficulties. Therefore, sign language recognition systems are developed to aid communication between hearing impaired people and others. This paper developed a static American Sign Language Recognition (ASLR) system using canny-edge and histogram of oriented gradient (HOG) for feature extraction with K-Nearest Neighbour (K-NN) as classifier. The sign language image datasets used consist of English alphabets from both Massey University and Kaggle, and numbers (0-9) from Massey University. Median filter was used to remove noise after images were converted to grayscale. Otsu algorithm was used for segmentation while edges in the images were preserved using canny edge detection technique with HOG parameters tuning to obtain feature vectors. The extracted features were used by K-NN for classification. An average recognition accuracy and computational testing time of 97.6% and 0.39s respectively were obtained based on experiments with the Massey University dataset. Similarly, an average recognition accuracy and computational testing time of 99.0% and 0.43s respectively were obtained based on experiments with the Kaggle dataset. The developed system successfully recognized static English alphabets and numbers and outperformed some existing systems.

Data processing method for simultaneous estimation of temperature and emissivity in multispectral thermometry.

The data processing in multispectral thermometry remains a huge challenge due to the unknown emissivity. In this article, a novel data processing model of multispectral thermometer is established by adding new constraints of emissivity on the basis of object function. The new two algorithms for model optimizing, Sequential Randomized Coordinate Shrinking (SRCS) and Multiple-Population Genetic (MPG), are introduced. The temperature and emissivity of two samples are calculated by MPG algorithm to prove the validity of the MPG algorithm in practical application. The experiments reveal that the relative error of temperature is within 0.4% with the average calculation time of 0.36 s. The method proposed in this article can realize the simultaneous estimation of temperature and emissivity without emissivity assumption model, which is expected to be applied to real-time measurement of temperature in industrial fields.

Research on a Random Route-Planning Method Based on the Fusion of the A* Algorithm and Dynamic Window Method

Path planning is a hot topic at present. Considering the global and local path planning of mobile robot is one of the challenging research topics. The objective of this paper is to create a rasterized environment that optimizes the planning of multiple paths and solves barrier avoidance issues. Combining the A* algorithm with the dynamic window method, a robo-assisted random barrier avoidance method is used to resolve the issues caused by collisions and path failures. Improving the A* algorithm requires analyzing and optimizing its evaluation function to increase search efficiency. The redundant point removal strategy is then presented. The dynamic window method is utilized for local planning between each pair of adjacent nodes. This method guarantees that random obstacles are avoided in real-time based on the globally optimal path. The experiment demonstrates that the enhanced A* algorithm reduces the average path length and computation time when compared to the traditional A* algorithm. After fusing the dynamic window method, the local path is corrected using the global path, and the resolution for random barrier avoidance is visualized.

Optimization of mesh generation for geometric accuracy, robustness, and efficiency of biomechanical-model-based deformable image registration.

Successful generation of biomechanical-model-based deformable image registration (BM-DIR) relies on user-defined parameters that dictate surface mesh quality. The trial-and-error process to determine the optimal parameters can be labor-intensive and hinder DIR efficiency and clinical workflow. To identify optimal parameters in surface mesh generation as boundary conditions for a BM-DIR in longitudinal liver and lung CT images to facilitate streamlined image registration processes. Contrast-enhanced CT images of 29 colorectal liver cancer patients and end-exhale four-dimensional CT images of 26 locally advanced non-small cell lung cancer patients were collected. Different combinations of parameters that determine the triangle mesh quality (voxel side length and triangle edge length) were investigated. The quality of DIRs generated using these parameters was evaluated with metrics for geometric accuracy, robustness, and efficiency. Metrics for geometric accuracy included target registration error (TRE) of internal vessel bifurcations, dice similar coefficient (DSC), mean distance to agreement (MDA), Hausdorff distance (HD) for organ contours, and number of vertices in the triangle mesh. American Association of Physicists in Medicine Task Group 132 was used to ensure parameters met TRE, DSC, MDA recommendations before the comparison among the parameters. Robustness was evaluated as the success rate of DIR generation, and efficiency was evaluated as the total time to generate boundary conditions and compute finite element analysis. Voxel side length of 0.2cm and triangle edge length of 3 were found to be the optimal parameters for both liver and lung, with success rate of 1.00 and 0.98 and average DIR computation time of 100 and 143s, respectively. For this combination, the average TRE, DSC, MDA, and HD were 0.38-0.40, 0.96-0.97, 0.09-0.12, and 0.87-1.17mm, respectively. The optimal parameters were found for the analyzed patients. The decision-making process described in this study serves as a recommendation for BM-DIR algorithms to be used for liver and lung. These parameters can facilitate consistence in the evaluation of published studies and more widespread utilization of BM-DIR in clinical practice.

Distribution network reconfiguration using time-varying acceleration coefficient assisted binary particle swarm optimization

The particle swarm optimization (PSO) algorithm is widely used to solve a variety of complicated engineering problems. However, PSO may suffer from an effective balance between local and global search ability in the solution search process. This study proposes a new acceleration coefficient for the PSO algorithm to overcome this issue. The proposed coefficient is implemented on the distribution network reconfiguration (DNR) problem to reduce power loss. The lowest power loss is obtained while problem constraints (maintain radial structure, voltage limits, and power flow balance) are satisfied with the proposed method. The validity of the proposed acceleration coefficient-based binary particle swarm optimization (BPSO) in handling the DNR problem is examined through simulation studies on IEEE 33-bus, P&G 69-bus, and 84-bus Taiwan Power Company (TPC) practical distribution networks. Furthermore, the DNR problem is evaluated regarding energy cost and environmental issues. Finally, the average computational times of the different acceleration coefficient-based PSO methods are compared. The solution speed of the proposed acceleration coefficient-based method is faster than the other methods in the DNR problem.

Optimizing the path of seedling transplanting with multi-end effectors by using an improved greedy annealing algorithm

Automated transplanters perform the repetitive task of replugging bad or missing cells with healthy seedlings in greenhouses. The work efficiency of transplanters can be improved by optimizing the transplanting paths of manipulators. In this study, an improved greedy simulated annealing algorithm (IGSA) based on MTSP model was proposed for path optimization. Aimed at improving the computation speed of the algorithm, a new improved 2-Opt algorithm based on the metropolis criterion was utilized. Moreover, the greedy algorithm (GRA) and the greedy genetic algorithm (GGA) were modified to adapt to the multi-end effectors’ replugging operation. The performance of IGSA and those of the GGA, GRA, and common sequence method (CSM) in path planning for replugging transplanting were compared in terms of their optimization effects and computation times. Then, manipulators with different numbers of end effectors (four, five, and six end effectors) were used to verify the increased effect of the path planning. Under the verification test conditions, the comparative average optimization ratios of GRA, GGA, and IGSA compared to the CSM were 4.87 %, 10.64 %, and 20.07 %, respectively. The performance ranking of the methods with short average paths was in the order of IGSA, GGA, GRA, and CSM. Compared with the average path shortening ratio of the four end effectors, those of the five and six end effectors were 11.44 % and 18.68 %, respectively. The average computation times of CSM, GRA, GGA, and IGSA were 0.002, 0.007, 6.94, and 3.49 s in MATLAB (R2019a), respectively. All of them can meet the real-time operation requirements except for GGA. In contrast to GGA, GRA, and CSM, the proposed IGGA performed effectively in the path planning of the seedlings’ replug transplanting owing to its comprehensive performance derived from its path optimization ratio and low cost in terms of computation time.

A Region Selection Method for Real-time Local Correlation Tracking of Solar Full-disk Magnetographs

Hundreds of images with the same polarization state are first registered to compensate for the jitters during an observation and then integrated to realize the needed spatial resolution and sensitivity for solar magnetic field measurement. Due to the feature dependent properties of the correlation tracker technique, an effective method to select the feature region is critical for low-resolution full-disk solar filtergrams, especially those with less significant features when the Sun is quiet. In this paper, we propose a region extraction method based on a Hessian matrix and information entropy constraints for local correlation tracking (CT) to get linear displacement between different images. The method is composed of three steps: (1) extract feature points with the Hessian matrix, (2) select good feature points with scale spaces and thresholds, and (3) locate the feature region with the two-dimensional information entropy constraints. Both the simulated and observational experiments demonstrated that our region selection method can efficiently detect the linear displacement and improve the quality of a ground-based full-disk solar magnetogram. The local CT with the selected regions can obtain displacement detection results as good as the global CT and at the same time significantly reduce the average calculation time.

Medium- to long-term nickel price forecasting using LSTM and GRU networks

Recently, nickel is a critical metal for manufacturing stainless steel, rechargeable electric vehicle batteries, and alloys utilized in the state-of-the-art technologies. The use of more environmentally friendly electric vehicles has become widespread and brought tackling climate change to forefront, especially for reducing greenhouse gas emissions. Therefore, the demand for rechargeable batteries that power electric vehicles and the need for the nickel in the production of these batteries will increase as well. In addition to those, nickel prices significantly impact mine investment decisions, mine planning, economic development of nickel companies, and countries that depend on nickel resources. However, there is uncertainty about how the nickel price will trend in the future, and the solution to this problem attracts the attention of researchers. For forecasting nickel price, this paper proposes recurrent neural networks-based on long short-term memory (LSTM) and gated recurrent unit (GRU) networks, classified as deep learning algorithms. Mean absolute percentage error (MAPE) was used as the performance measure to compute the accuracy of the proposed techniques. As a result, it has been determined that the LSTM and GRU networks are very useful and successful in forecasting the nickel price variations owing to having average MAPE values of 7.060% and 6.986%, respectively. Furthermore, it has been observed that GRU networks surpassed the LSTM networks by 33% in terms of average computational time.

Social Distancing Detection in Real Time using Deep Learning and IoT

Abstract: Social Distancing is the practice of maintaining a minimum amount of physical distance from your surrounding people or avoiding direct contact with people or objects in public places so as to minimize the exposure or transmission of diseases like COVID-19. This can also be coupled with wearing of face masks and regular sanitization of hands. The government has made social distancing norms compulsory and to be followed, and penalties would be imposed if violated. But there are a lot of examples of social distancing violations in public places due to negligence of people. There are approaches using bluetooth and mobile phones but it requires an app to be installed. This study provides a method for determining whether the social distancing rule has been broken by combining machine learning models and object detecting techniques. Appropriate actions would be taken in case of any violations detected. In this paper, a comparison is done on all the object detection techniques like Faster R-CNN, YOLO v3, SSD Mobile Net etc. based on accuracy, mean average precision, computational time and ease of integration. Based on this, SSD Mobile Net performed well on accuracy and had a faster computational time. This model can also be deployed on end devices like Raspberry Pi.

Morphological Dynamics-Based Anomaly Detection Towards In Situ Layer-Wise Certification for Directed Energy Deposition Processes

Abstract The process uncertainty induced quality issue remains the major challenge that hinders the wider adoption of additive manufacturing (AM) technology. The defects occurred significantly compromise structural integrity and mechanical properties of fabricated parts. Therefore, there is an urgent need in fast, yet reliable AM component certification. Most finite element analysis related methods characterize defects based on the thermomechanical relationships, which are computationally inefficient and cannot capture process uncertainty. In addition, there is a growing trend in data-driven approaches on characterizing the empirical relationships between thermal history and anomaly occurrences, which focus on modeling an individual image basis to identify local defects. Despite their effectiveness in local anomaly detection, these methods are quite cumbersome when applied to layer-wise anomaly detection. This paper proposes a novel in situ layer-wise anomaly detection method by analyzing the layer-by-layer morphological dynamics of melt pools and heat affected zones (HAZs). Specifically, the thermal images are first preprocessed based on the g-code to assure unified orientation. Subsequently, the melt pool and HAZ are segmented, and the global and morphological transition metrics are developed to characterize the morphological dynamics. New layer-wise features are extracted, and supervised machine learning methods are applied for layer-wise anomaly detection. The proposed method is validated using the directed energy deposition (DED) process, which demonstrates superior performance comparing with the benchmark methods. The average computational time is significantly shorter than the average build time, enabling in situ layer-wise certification and real-time process control.