Algorithmic Pipeline Research Articles

Residual Pyramidal GAN (RP-GAN) for crack detection and prediction of crack growth in engineered cementitious composites

Crack detection has recently gained credence in the field of engineering automation. Engineered cementitious composites (ECC) are a durable, and environmentally friendly construction material and a suitable replacement for conventional concrete. As such, there is a need to automate the process of detecting the unique ductile crack formation and growth in ECC. This research focuses on detecting ECC cracks, accurately mapping the crack morphology, and predicting the crack growth patterns of this advanced material. We design and present a novel generative adversarial algorithm named Residual Pyramidal Generative Adversarial Network (RP-GAN), capable of ECC crack detection. RP-GAN is an enhanced and improved version of the typical Pix2Pix GAN for paired image-to-image translation. It is an extensive network that could be used for macro and micro-segmentation purposes in addition to domain translation tasks. In this research, we first investigate the ability of RP-GAN to detect ECC micro-cracks and map the dimensions of each crack. Secondly, we compared RP-GAN’s crack detection results with those of known state-of-the-art architectures. We then developed an algorithmic pipeline for live crack tracking using an experimental framework. Finally, we evaluate the possibility of detecting ECC crack growth and propagation using sequential images by re-engineering RP-GAN and fusing Convolutional-LSTM into its architecture. The mean intersection over union (IOU) test results of RP-GAN for ECC crack detection equaled 81.28%, and 77.76% for ECC crack propagation prediction.

Dynamic reordering and inspection for the multi-item Inventory Record Inaccuracy problem

Inventory Record Inaccuracy (IRI) is a significant challenge in inventory management, caused by discrepancies between actual stock and inventory records due to factors such as spoilage, theft, and obsolescence. Despite extensive academic focus on robust ordering and inventory inspection planning, current literature either considers these in isolation, or focuses on stylized single-item scenarios, not addressing the full dynamics of reordering and inspection in warehouses with many SKUs. We consider such a large warehouse with several heterogeneous items that are subject to shrinkage and limited inspection opportunities. Our approach jointly optimizes reordering and inspection decisions in a dynamic decision framework, incorporating both inspection and travel times. We introduce an algorithmic pipeline for dynamic reordering and inspection that combines a neural network to decide on replenishments and a mixed-integer program to determine an optimal inspection subset. Our results show that ignoring shrinkage can increase costs by up to 95.3% and inspection already becomes viable at a relatively low level of IRI. Our proposed policy saves 8.3% in total costs compared to a static reordering and inspection benchmark, and 21.8% compared to a no-inspection policy.

UAV-based field watermelon detection and counting using YOLOv8s with image panorama stitching and overlap partitioning

Accurate watermelon yield estimation is crucial to the agricultural value chain, as it guides the allocation of agricultural resources as well as facilitates inventory and logistics planning. The conventional method of watermelon yield estimation relies heavily on manual labor, which is both time-consuming and labor-intensive. To address this, this work proposes an algorithmic pipeline that utilizes unmanned aerial vehicle (UAV) videos for detection and counting of watermelons. This pipeline uses You Only Look Once version 8 s (YOLOv8s) with panorama stitching and overlap partitioning, which facilitates the overall number estimation of watermelons in field. The watermelon detection model, based on YOLOv8s and obtained using transfer learning, achieved a detection accuracy of 99.20 %, demonstrating its potential for application in yield estimation. The panorama stitching and overlap partitioning based detection and counting method uses panoramic images as input and effectively mitigates the duplications compared with the video tracking based detection and counting method. The counting accuracy reached over 96.61 %, proving a promising application for yield estimation. The high accuracy demonstrates the feasibility of applying this method for overall yield estimation in large watermelon fields.

Segmentation of MHD modes using Fourier transform, wavelets and computer vision algorithms

Abstract Magnetohydrodynamic (MHD) activity in fusion devices is typically analyzed by examining time-frequency spectrograms obtained from various diagnostics. MHD modes often co-exist with various types of noise and complex patterns generated by other events like pellet injection or active diagnostics. Traditionally, identifying MHD modes has been a manual task, making it labor-intensive. To overcome this issue, this study proposes the use of computer vision (CV) algorithms for noise removal and automatic feature extraction. First, the automatic detection of straight-line patterns is achieved by applying the Hough transform. Then, the discrete wavelet transform is proposed to break down spectrograms into sub-images of different scales, removing broadband noise and pellet injection signatures. The multiscale decomposition is subsequently extended to multiple directions using either 2D Fourier transforms or curvelets, achieving a high signal-to-noise ratio in spectrograms and eliminating undesired frequency sweeps of toroidal Alfvén eigenmodes antenna. Once MHD activity is successfully enhanced, a pipeline of algorithms for ridge detection, thresholding and labeling perform a segmentation of the image, automatically labeling individual modes. This study demonstrates the effectiveness of CV algorithms for the identification of MHD modes. The use of such algorithms may potentially help in the analysis process and the creation of large databases of modes.

Deep Learning Phenotyping of Tricuspid Regurgitation for Automated High Throughput Assessment of Transthoracic Echocardiography.

Diagnosis of tricuspid regurgitation (TR) requires careful expert evaluation. This study developed an automated deep learning pipeline for assessing TR from transthoracic echocardiography. An automated deep learning workflow was developed using 47,312 studies (2,079,898 videos) from Cedars-Sinai Medical Center (CSMC) between 2011 and 2021. The pipeline was tested on a temporally distinct test set of 2,462 studies (108,138 videos) obtained in 2022 at CSMC and a geographically distinct cohort of 5,549 studies (278,377 videos) from Stanford Healthcare (SHC). In the CSMC test dataset, the view classifier demonstrated an AUC of 1.000 (0.999 - 1.000) and identified at least one A4C video with colour Doppler across the tricuspid valve in 2,410 of 2,462 studies with a sensitivity of 0.975 (0.968-0.982) and a specificity of 1.000 (1.00-1.000). In the CSMC test cohort, moderate-or-severe TR was detected with an AUC of 0.928 (0.913 - 0.943) and severe TR was detected with an AUC of 0.956 (0.940 - 0.969). In the SHC cohort, the view classifier correctly identified at least one TR colour Doppler video in 5,268 of the 5,549 studies, resulting in an AUC of 0.999 (0.998 - 0.999), a sensitivity of 0.949 (0.944 - 0.955) and specificity of 0.999 (0.999 - 0.999). The AI model detected moderate-or-severe TR with an AUC of 0.951 (0.938 - 0.962) and severe TR with an AUC of 0.980 (0.966 - 0.988). We developed an automated pipeline to identify clinically significant TR with excellent performance. This approach carries potential for automated TR detection and stratification for surveillance and screening. Can an automated deep learning model assess tricuspid regurgitation severity from echocardiography? We developed and validated an automated tricuspid regurgitation detection algorithm pipeline across two healthcare systems with high volume echocardiography labs. The algorithm correctly identifies apical-4-chamber view videos with colour Doppler across the tricuspid valve and grades clinically significant TR with strong agreement to expert clinical readers. A deep learning pipeline could automate TR screening, facilitating reproducible accurate assessment of TR severity, allowing rapid triage or re-review and expand access in low-resource or primary care settings.

Real-time multi-camera 3D human pose estimation at the edge for industrial applications

There is an increasing interest in exploiting human pose estimation (HPE) software in human–machine interaction systems. Nevertheless, adopting such a computer vision application in real industrial scenarios is challenging. To overcome occlusion limitations, it requires multiple cameras, which in turn require multiple, distributed, and synchronized HPE software nodes running on resource-constrained edge devices. We address this challenge by presenting a real-time distributed 3D HPE platform, which consists of a set of 3D HPE software nodes on edge devices (i.e., one per camera) to redundantly extrapolate the human pose from different points of view. A centralized aggregator collects the pose information through a shared communication network and merges them, in real time, through a pipeline of filtering, clustering and association algorithms. It addresses network communication issues (e.g., delay and bandwidth variability) through a two-levels synchronization, and supports both single and multi-person pose estimation. We present the evaluation results with a real case of study (i.e., HPE for human–machine interaction in an intelligent manufacturing line), in which the platform accuracy and scalability are compared with state-of-the-art approaches and with a marker-based infra-red motion capture system.

Searching for local features in primordial power spectrum using genetic algorithms

ABSTRACT We present a novel methodology for exploring local features directly in the primordial power spectrum using a genetic algorithm pipeline coupled with a Boltzmann solver and Cosmic Microwave Background data (CMB). After testing the robustness of our pipeline using mock data, we apply it to the latest CMB data, including Planck 2018 and CamSpec PR4. Our model-independent approach provides an analytical reconstruction of the power spectra that best fits the data, with the unsupervised machine learning algorithm exploring a functional space built off simple ‘grammar’ functions. We find significant improvements upon the simple power-law behaviour, by Δχ2 ≲ −21, consistently with more traditional model-based approaches. These best-fits always address both the low-ℓ anomaly in the TT spectrum and the residual high-ℓ oscillations in the TT, TE, and EE spectra. The proposed pipeline provides an adaptable tool for exploring features in the primordial power spectrum in a model-independent way, providing valuable hints to theorists for constructing viable inflationary models that are consistent with the current and upcoming CMB surveys.

Artificial Intelligence-Based, Wavelet-Aided Prediction of Long-Term Outdoor Performance of Perovskite Solar Cells.

The commercial development of perovskite solar cells (PSCs) has been significantly delayed by the constraint of performing time-consuming degradation studies under real outdoor conditions. These are necessary steps to determine the device lifetime, an area where PSCs traditionally suffer. In this work, we demonstrate that the outdoor degradation behavior of PSCs can be predicted by employing accelerated indoor stability analyses. The prediction was possible using a swift and accurate pipeline of machine learning algorithms and mathematical decompositions. By training the algorithms with different indoor stability data sets, we can determine the most relevant stress factors, thereby shedding light on the outdoor degradation pathways. Our methodology is not specific to PSCs and can be extended to other PV technologies where degradation and its mechanisms are crucial elements of their widespread adoption.

Determining individual glomerular proteinuria and periglomerular infiltration in a cleared murine kidney by a 3-dimensional fast marching algorithm

Three-dimensional imaging has advanced basic research and clinical medicine. However, limited resolution and imperfections of real-world 3D image material often preclude algorithmic image analysis. Here, we present a methodological framework for such imaging and analysis for functional and spatial relations in experimental nephritis. First, optical tissue clearing protocols were optimized to preserve fluorescence signals for light-sheet-fluorescence-microscopy and compensated attenuation effects using adjustable 3D correction fields. Next, we adapted the Fast — Marching — Algorithm to conduct backtracking in 3D environments and developed a tool to determine local concentrations of extractable objects. As a proof-of-concept application, we used this framework to determine in a glomerulonephritis model the individual proteinuria and periglomerular immune cell infiltration for all glomeruli of half a mouse kidney. A correlation between these parameters surprisingly did not support the intuitional assumption that the most inflamed glomeruli are the most proteinuric. Instead, the spatial density of adjacent glomeruli positively correlated with the proteinuria of a given glomerulus. Since proteinuric glomeruli appear clustered, this suggests that the exact location of a kidney biopsy may affect the observed severity of glomerular damage. Thus, our algorithmic pipeline described here allows analysis of various parameters of various organs composed of functional subunits, such as the kidney, and can theoretically be adapted to processing other image modalities.

BIMBot for Autonomous Laser Scanning in Built Environments

Accurate and complete 3D point clouds are essential in creating as-built building information modeling (BIM) models, although there are challenges in automating the process for 3D point cloud creation in complex environments. In this paper, an autonomous scanning system named BIMBot is introduced, which integrates advanced light detection and ranging (LiDAR) technology with robotics to create 3D point clouds. Using our specially developed algorithmic pipeline for point cloud processing, iterative registration refinement, and next best view (NBV) calculation, this system facilitates an efficient, accurate, and fully autonomous scanning process. The BIMBot’s performance was validated using a case study in a campus laboratory, featuring complex structural and mechanical, electrical, and plumbing (MEP) elements. The experimental results showed that the autonomous scanning system produced 3D point cloud mappings in fewer scans than the manual method while maintaining comparable detail and accuracy, demonstrating its potential for wider application in complex built environments.

Generative models for color normalization in digital pathology and dermatology: Advancing the learning paradigm

Color medical images introduce an additional confounding factor compared to conventional grayscale medical images: color variability. This variability can lead to inconsistent evaluation by clinicians and the misinterpretation or suboptimal learning process of automatic quantitative algorithms. To mitigate the potential negative consequences of color variability, several color normalization strategies have been developed, proving to be effective in standardizing image appearance. In this paper, we present a novel paradigm for color normalization using generative adversarial networks (GANs). Our method focuses on standardizing images in the field of digital pathology (stain normalization) and dermatology (color constancy), where high color variability is consistently observed. Specifically, we formulate the color normalization task as an image-to-image translation problem, ensuring a pixel-to-pixel correspondence between the original and normalized images. Our approach outperforms existing state-of-the-art methods in both the digital pathology and dermatology fields. Extensive validation using public datasets demonstrate the effectiveness of our color normalization results on entirely external test sets. Our framework exhibits strong generalization capability on unseen data, making it suitable for inclusion in the pipeline of automatic quantitative algorithms to reduce color variability and improve segmentation and/or classification performance. Lastly, we provide the source code of our models to encourage open science.

Speech-Based Digital Biomarkers for Alzheimer's Research.

Digital biomarkers are of growing interest in the field of Alzheimer's Disease(AD) research. Digital biomarker data arising from digital health tools holds various potential benefits: more objective and more accurate assessment of patients' symptoms and remote collection of signals in real-world scenarios but also multimodal variance for prediction models of individual disease progression. Speech can be collected at minimal patient burden and provides rich data for assessing multiple aspects of AD pathology including cognition. However, the operations around collecting, preparing, and validly interpreting speech data within the context of clinical research on AD remains complex and sometimes challenging. Through a dedicated pipeline of speech collection tools, preprocessing steps and algorithms, precise qualification and quantification of an AD patient's pathology can be achieved from their speech. The aim of this chapter is to describe the methods that are needed to create speech collection scenarios that result in valuable speech-based digital biomarkers for clinical research.

Design, Prototyping, and Evaluation of A New Machine Vision-Based Automated Sweetpotato Grading and Sorting System

HighlightsA computer algorithm pipeline with YOLOv8 was developed for grading sweetpotatoes for size and surface defects.Pneumatically powered sorting mechanisms were prototyped and evaluated for sweetpotato sorting.The sorting mechanisms achieved 96.9% accuracy and 92.5% repeatability at a conveyor speed of 12 cm s-1.The integrated system achieved online, automated grading and sorting accuracies of 93.8% and 92.7%, respectively.Abstract. Automated quality grading and sorting of fresh market sweetpotato roots have not been achieved at packing facilities, which consequently requires significant labor to manually grade the products and sort them by hand. Given an increasing shortage of labor and rising labor costs, there is a pressing need to develop automated grading and sorting technology for sweetpotato packers to stay competitive and profitable. This study reports on the design, prototyping, and evaluation of an innovative machine vision-based, automated sweetpotato grading and sorting system empowered by state-of-the-art artificial intelligence. The developed prototype consists of a custom-designed roller conveyor-based machine vision unit that performs size and surface defect detection and grading of sweetpotatoes traveling and rotating on the conveyor and pneumatically powered sorting mechanisms for automated sorting of the sweetpotatoes into three grades after they exit from the vision area. A computer algorithm pipeline was developed to segment and track individual sweetpotato instances by YOLOv8l in real-time and perform quality grading based on multiple views of samples for size (length and width) and surface defects, and then activate the sorting mechanisms to selectively segregate graded sweetpotatoes online. Experiments were conducted on the performance of the sorting mechanisms, the machine vision unit, and the fully integrated system at varied conveyor speeds. The sorter without vision integration achieved an accuracy of 96.9% and repeatability of 92.5% at a conveyor speed of 12 cm s-1. The machine vision system yielded an overall accuracy of 95.8% and 93.8% in grading three-class sweetpotatoes for surface defects alone and for both size and surface defects, respectively, at the conveyor speed of 12 cm s-1. An overall sorting accuracy of 92.7% was achieved by the integrated system at the conveyor speed of 12 cm s-1. Grading and sorting accuracies tended to decrease as the conveyor speed increased. A video demo of automated sweetpotato grading and sorting processes is available at https://youtu.be/3QEp9wFTiu0. With more dedicated efforts on hardware and software improvements, the developed grading and sorting machine prototype promises to evolve into a practical implementation at commercial sweetpotato packing facilities. Keywords: Artificial intelligence, Automated grading and sorting, Machine vision, Sweetpotato, System integration.

A Deep Learning Driven Algorithmic Pipeline for Autonomous Navigation in Row-Based Crops

A Deep Learning Driven Algorithmic Pipeline for Autonomous Navigation in Row-Based Crops

Using Computer Vision and Artificial Intelligence to Track the Healing of Severe Burns.

Burn care management includes assessing the severity of burns accurately, especially distinguishing superficial partial-thickness burns from deep partial-thickness burns, in the context of providing definitive, downstream treatment. Moreover, the healing of the wound in the subacute care setting requires continuous tracking to avoid complications. Artificial intelligence (AI) and computer vision (CV) provide a unique opportunity to build low-cost and accessible tools to classify burn severity and track changes in wound parameters, both in the clinic by physicians and nurses and asynchronously in the remote setting by the patient themselves. Wound assessments can be achieved by AI-CV using the principles of image-guided therapy using high-quality 2D color images. Wound parameters can include wound 2D spatial dimension and the characterization of wound color changes, which demonstrates physiological changes such as the presentation of eschar/necrotic tissue, pustulence, granulation tissue, and scabbing. Here we present the development of AI-CV-based Skin Abnormality Tracking Algorithm pipeline. Additionally, we provide the results on a single localized burn tracked for a 6-week period in the clinic and an additional 2-week period of home monitoring.

Dual regularized policy updating and shiftpoint detection for automated deployment of reinforcement learning controllers on industrial mechatronic systems

We propose an algorithmic pipeline enabling deep reinforcement learning controllers to detect when a significant change in system characteristics has occurred and update the control policy accordingly to reattain performance. Reinforcement learning algorithms can learn a policy directly from input–output data and thus optimize for system-specific properties. Yet they face difficulties to adapt, after deployment, to varying operating conditions. Real-world industrial mechatronic systems however demand further levels of performance through adaptation while remaining safe. So far, methods that detect changes in environments exist but have never been studied and applied as a means to update control policies for time-varying systems. We benchmark several methods that detect significant changes in these systems, i.e. shiftpoint detection methods, and present a novel algorithm with a dual regularization architecture. This architecture exploits the prior policy while allowing sufficient flexibility to update for the safety-critical and time-varying system. We validate the method’s performance through benchhmarking and study the effect of its different components and targeted ablation studies on mechatronic systems, both in simulations and experimentally. Results show that our algorithmic pipeline allows for rapid shiftpoint detection, followed by a policy update that reaches expert performance after convergence.

The High Time Resolution Universe Pulsar survey – XVIII. The reprocessing of the HTRU-S Low Lat survey around the Galactic Centre using a Fast Folding Algorithm pipeline for accelerated pulsars

ABSTRACT The HTRU-S Low Latitude survey data within 1° of the Galactic Centre (GC) were searched for pulsars using the Fast Folding Algorithm (FFA). Unlike traditional Fast Fourier Transform (FFT) pipelines, the FFA optimally folds the data for all possible periods over a given range, which is particularly advantageous for pulsars with low-duty cycles. For the first time, a search over acceleration was included in the FFA to improve its sensitivity to binary pulsars. The steps in dispersion measure (DM) and acceleration were optimized, resulting in a reduction of the number of trials by 86 per cent. This was achieved over a search period range from 0.6 to 432-s, i.e. 10 per cent of the observation time (4320s), with a maximum DM of 4000 pc cm−3 and an acceleration range of ±128 m s−2. The search resulted in the re-detections of four known pulsars, including a pulsar that was missed in the previous FFT processing of this survey. This result indicates that the FFA pipeline is more sensitive than the FFT pipeline used in the previous processing of the survey within our parameter range. Additionally, we discovered a 1.89-s pulsar, PSR J1746-2829, with a large DM, located 0.5 from the GC. Follow-up observations revealed that this pulsar has a relatively flat spectrum (α = −0.9 ± 0.1) and has a period derivative of ∼1.3 × 10−12 s s−1, implying a surface magnetic field of ∼5.2 × 1013 G and a characteristic age of ∼23 000 yr. While the period, spectral index, and surface magnetic field strength are similar to many radio magnetars, other characteristics such as high linear polarization are absent.

Automated measurements of interscrew angles in vertebral body tethering patients with deep learning

BACKGROUND CONTEXTVertebral body tethering is the most popular nonfusion treatment for adolescent idiopathic scoliosis. The effect of the tether cord on the spine can be segmentally assessed by comparing the angle between two adjacent screws (interscrew angle) over time. Tether breakage has historically been assessed radiographically by a change in adjacent interscrew angle by greater than 5° between two sets of imaging. A threshold for growth modulation has not yet been established in the literature. These angle measurements are time consuming and prone to interobserver variability. PURPOSEThe purpose of this study was to develop an automated deep learning algorithm for measuring the interscrew angle following VBT surgery. STUDY DESIGN/SETTINGSingle institution analysis of medical images. PATIENT SAMPLEWe analyzed 229 standing or bending AP or PA radiographs from 100 patients who had undergone VBT at our institution. OUTCOME MEASURESPhysiologic Measures: An image processing algorithm was used to measure interscrew angles. METHODSA total of 229 standing or bending AP or PA radiographs from 100 VBT patients with vertebral body tethers were identified. Vertebral body screws were segmented by hand for all images and interscrew angles measured manually for 60 of the included images. A U-Net deep learning model was developed to automatically segment the vertebral body screws. Screw label maps were used to develop and tune an image processing algorithm which measures interscrew angles. Finally, the completed model and algorithm pipeline was tested on a 30-image test set. Dice score and absolute error were used to measure performance. RESULTSInter- and Intra-rater reliability for manual angle measurements were assessed with ICC and were both 0.99. The segmentation model Dice score against manually segmented ground truth across the 30-image test set was 0.96. The average interscrew angle absolute error between the algorithm and manually measured ground truth was 0.66° and ranged from 0° to 2.67° in non-overlapping screws (N=206). The primary modes of failure for the model were overlapping screws on a right thoracic/left lumbar construct with two screws in one vertebra and overexposed images. An algorithm step which determines whether an overlapping screw was present correctly identified all overlapping screws, with no false positives. CONCLUSIONWe developed and validated an algorithm which measures interscrew angles for radiographs of vertebral body tether patients with an accuracy of within 1° for the majority of interscrew angles. The algorithm can process five images per second on a standard computer, leading to substantial time savings. This algorithm may be used for rapid processing of large radiographic databases of tether patients and could enable more rigorous definitions of growth modulation and cord breakage to be established.

A Machine Learning Framework for Quantifying Chemical Segregation and Microstructural Features in Atom Probe Tomography Data.

Atom probe tomography (APT) is ideally suited to characterize and understand the interplay of segregation and microstructure in modern multi-component materials. Yet, the quantitative analysis typically relies on human expertise to define regions of interest. We introduce a computationally efficient, multi-stage machine learning strategy to identify compositionally distinct domains in a semi-automated way, and subsequently quantify their geometric and compositional characteristics. In our algorithmic pipeline, we first coarse-grain the APT data into voxels, collect the composition statistics, and decompose it via clustering in composition space. The composition classification then enables the real-space segmentation via a density-based clustering algorithm, thus revealing the microstructure at voxel resolution. Our approach is demonstrated for a Sm-(Co,Fe)-Zr-Cu alloy. The alloy exhibits two precipitate phases with a plate-like, but intertwined morphology. The primary segmentation is further refined to disentangle these geometrically complex precipitates into individual plate-like parts by an unsupervised approach based on principle component analysis, or a U-Net-based semantic segmentation trained on the former. Following the composition and geometric analysis, detailed composition distribution and segregation effects relative to the predominant plate-like geometry can be readily mapped from the point cloud, without resorting to the voxel compositions.

Touch-and-Heal

Affective touch plays an important role in human-robot interaction. However, it is challenging for robots to perceive various natural human tactile gestures accurately, and feedback human intentions properly. In this paper, we propose a data-driven affective computing system based on a biomimetic quadruped robot with large-format, high-density flexible pressure sensors, which can mimic the natural tactile interaction between humans and pet dogs. We collect 208-minute videos from 26 participates and construct a dataset of 1212 human gestures-dog actions interaction sequences. The dataset is manually annotated with an 81-tactile-gesture vocabulary and a 44-corresponding-dog-reaction vocabulary, which are constructed through literature, questionnaire, and video observation. Then, we propose a deep learning algorithm pipeline with a gesture classification algorithm based on ResNet and an action prediction algorithm based on Transformer, which achieve the classification accuracy of 99.1% and the 1-gram BLEU score of 0.87 respectively. Finally, we conduct a field study to evaluate the emotion regulation effects through tactile affective interaction, and compare it with voice interaction. The results show that our system with tactile interaction plays a significant role in alleviating user anxiety, stimulating user excitement and improving the acceptability of robotic dogs.