The upper limb exoskeleton has the advantages of high durability, low labor intensity, and repeatability, and has broad application prospects in stroke rehabilitation. Aiming at the incompatibility of the upper limb exoskeleton robotic with the human, an upper limb exoskeleton rehabilitation robot (ULERR) was designed. Firstly, according to the human anatomy, the joint configuration of human upper limbs is analyzed. The ULERR is designed for the rehabilitation training of patients with hemiplegia in the middle and late stages caused by stroke. Secondly, it is established the kinematics and dynamics model of the exoskeleton and completed the analysis of dynamic simulation. Finally, the rehabilitation robot prototype was tested by a 3D dynamic capture system to measure the range of motion (ROM) of the upper limb joints with the rehabilitation robot. Finally, the results of simulation and experimental concluded that joint motion of the robot is stable, the degrees of freedom (DoFs) of robot is conform to human motion, the designed robot is reasonable, and the robot is suitable for rehabilitation training requirements.
{"title":"Design and analysis of an upper limb exoskeleton robot for stroke rehabilitation","authors":"Shuang Li, Zhanli Wang, Zaixiang Pang, Zhifeng Duan, Moyao Gao","doi":"10.1109/RCAR54675.2022.9872238","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872238","url":null,"abstract":"The upper limb exoskeleton has the advantages of high durability, low labor intensity, and repeatability, and has broad application prospects in stroke rehabilitation. Aiming at the incompatibility of the upper limb exoskeleton robotic with the human, an upper limb exoskeleton rehabilitation robot (ULERR) was designed. Firstly, according to the human anatomy, the joint configuration of human upper limbs is analyzed. The ULERR is designed for the rehabilitation training of patients with hemiplegia in the middle and late stages caused by stroke. Secondly, it is established the kinematics and dynamics model of the exoskeleton and completed the analysis of dynamic simulation. Finally, the rehabilitation robot prototype was tested by a 3D dynamic capture system to measure the range of motion (ROM) of the upper limb joints with the rehabilitation robot. Finally, the results of simulation and experimental concluded that joint motion of the robot is stable, the degrees of freedom (DoFs) of robot is conform to human motion, the designed robot is reasonable, and the robot is suitable for rehabilitation training requirements.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115009409","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
As a high-throughput and highly flexible technique, optically induced dielectrophoresis (ODEP) is one of the most promising micromanipulation techniques applied for biomedical studies. However, most ODEP-based manipulation methods have not been explored deeply in terms of accurate control under unstructured environments with multiple interference. This paper reports a dynamic control framework for automatically transporting single particle to goal position in a complex environment with an optically induced dielectrophoresis platform. The POMDP-based path planner periodically provides the optimal motion strategy based on the real-time environmental information and current position of the particle to avoid collisions with randomly moving obstacles. The optimal motion strategies are smoothly expanded to short-distance trajectories, which are dynamically followed by the target particle with proxy-based sliding mode control (PSMC) closed-loop controller. Experimental results indicated that compared with traditional controllers such as PID, our control method possesses higher accuracy and stability in path following. In addition, the performance of the path planner was demonstrated by transporting a NIH/3T3 cell to the desired position within a relatively crowded environment.
{"title":"Dynamic Control Framework for Automated Particle Transport Based on Optically Induced Dielectrophoresis","authors":"Jiaxin Liu, Huaping Wang, Qing Shi, Xinyi Dong, Kaijun Lin, Tao Sun, Qiang Huang, Toshio Fukuda","doi":"10.1109/RCAR54675.2022.9872252","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872252","url":null,"abstract":"As a high-throughput and highly flexible technique, optically induced dielectrophoresis (ODEP) is one of the most promising micromanipulation techniques applied for biomedical studies. However, most ODEP-based manipulation methods have not been explored deeply in terms of accurate control under unstructured environments with multiple interference. This paper reports a dynamic control framework for automatically transporting single particle to goal position in a complex environment with an optically induced dielectrophoresis platform. The POMDP-based path planner periodically provides the optimal motion strategy based on the real-time environmental information and current position of the particle to avoid collisions with randomly moving obstacles. The optimal motion strategies are smoothly expanded to short-distance trajectories, which are dynamically followed by the target particle with proxy-based sliding mode control (PSMC) closed-loop controller. Experimental results indicated that compared with traditional controllers such as PID, our control method possesses higher accuracy and stability in path following. In addition, the performance of the path planner was demonstrated by transporting a NIH/3T3 cell to the desired position within a relatively crowded environment.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117236661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-17DOI: 10.1109/RCAR54675.2022.9872257
Yuzhen Chen, Ying Yang, Jiajin Lv
In the field of autonomous driving, recognition and segmentation of road point clouds is an important task for the automatic production of 3D high-precision maps. To address the problems of lack of large-scale and complex road scene datasets for the instance segmentation, and the poor applicability of algorithms under large scenes, this paper produces a brand new and large-scale road instance segmentation dataset. Meanwhile, this paper proposes a brand new solution for semantic segmentation and clustering bias prediction, based on an improved Pointnet++ network, which is used together with the clustering algorithm of DBSCAN to conduct the instance segmentation. Thorough experiments indicate that the semantic segmentation accuracy of the proposed method reaches 0.982 on our produced road instance segmentation datasets, meanwhile the average accuracy and recall of the three classes of instance segmentation reach 0.853 and 0.784, respectively. Moreover, the bias network branch proposed in this paper can further improve the effectiveness of clustering, and the precision of our algorithm was improved by 15.1% and the recall rate was improved by 16.2%. It can be concluded that our produced dataset can support the large-scale road instance segmentation and our proposed algorithm can better adapt to the instance segmentation under large-scale and complex road scenarios.
{"title":"Large Scale Road Datasets and Point-Offset Network for 3D Instance Segmentation","authors":"Yuzhen Chen, Ying Yang, Jiajin Lv","doi":"10.1109/RCAR54675.2022.9872257","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872257","url":null,"abstract":"In the field of autonomous driving, recognition and segmentation of road point clouds is an important task for the automatic production of 3D high-precision maps. To address the problems of lack of large-scale and complex road scene datasets for the instance segmentation, and the poor applicability of algorithms under large scenes, this paper produces a brand new and large-scale road instance segmentation dataset. Meanwhile, this paper proposes a brand new solution for semantic segmentation and clustering bias prediction, based on an improved Pointnet++ network, which is used together with the clustering algorithm of DBSCAN to conduct the instance segmentation. Thorough experiments indicate that the semantic segmentation accuracy of the proposed method reaches 0.982 on our produced road instance segmentation datasets, meanwhile the average accuracy and recall of the three classes of instance segmentation reach 0.853 and 0.784, respectively. Moreover, the bias network branch proposed in this paper can further improve the effectiveness of clustering, and the precision of our algorithm was improved by 15.1% and the recall rate was improved by 16.2%. It can be concluded that our produced dataset can support the large-scale road instance segmentation and our proposed algorithm can better adapt to the instance segmentation under large-scale and complex road scenarios.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125886627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Deep reinforcement learning algorithms can enable agents to learn policies for complex tasks without expert knowledge. However, the learned policies are typically specialized to one specific task and can not generalize to new tasks. While meta-reinforcement learning (meta-RL) algorithms can enable agents to solve new tasks based on prior experience, most of them build on on-policy reinforcement learning algorithms which require large amounts of samples during meta-training and do not consider task-specific features across different tasks and thus make it very difficult to train an agent with high performance. To address these challenges, in this paper, we propose an off-policy meta-RL algorithm abbreviated as CRL (Celebrating Robustness Learning) that disentangles task-specific policy parameters by an adapter network to shared low-level parameters, learns a probabilistic latent space to extract universal information across different tasks and perform temporal-extended exploration. Our approach outperforms baseline methods both in sample efficiency and asymptotic performance on several meta-RL benchmarks.
{"title":"Celebrating Robustness in Efficient Off-Policy Meta-Reinforcement Learning","authors":"Ziyi Liu, Zongyuan Li, Qianqian Cao, Yuan Wan, Xian Guo","doi":"10.1109/RCAR54675.2022.9872291","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872291","url":null,"abstract":"Deep reinforcement learning algorithms can enable agents to learn policies for complex tasks without expert knowledge. However, the learned policies are typically specialized to one specific task and can not generalize to new tasks. While meta-reinforcement learning (meta-RL) algorithms can enable agents to solve new tasks based on prior experience, most of them build on on-policy reinforcement learning algorithms which require large amounts of samples during meta-training and do not consider task-specific features across different tasks and thus make it very difficult to train an agent with high performance. To address these challenges, in this paper, we propose an off-policy meta-RL algorithm abbreviated as CRL (Celebrating Robustness Learning) that disentangles task-specific policy parameters by an adapter network to shared low-level parameters, learns a probabilistic latent space to extract universal information across different tasks and perform temporal-extended exploration. Our approach outperforms baseline methods both in sample efficiency and asymptotic performance on several meta-RL benchmarks.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126138809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Magnetically controlled microrobots for drug delivery and noninvasive treatment have great potential applications in the biomedical field in the future. The construction of the magnetic actuation system is an important step to realize the automated control of micro/nano swimmers. However, the construction of a magnetic actuation system still faces challenges; for example, the magnetic field cannot be turned off immediately, the distribution of the magnetic field in the workspace is not uniform, the working space is limited and the feedback is inconvenient. In view of the above problems, a design method based on an eight-axis electromagnetic coil magnetic control system is introduced in this paper, which can compositely actuate the microrobot and ensure movement with five degrees of freedom. In addition, the overall size of the system can be reduced as much as possible under the condition that the magnetic field in the workspace is sufficiently uniform and the magnetic field intensity is sufficiently large. Finally, in the experimental part, the magnetic field uniformity is verified by magnetic field simulation and measurement, and then the path following of the square trajectory is realized with the $75 mu mathrm{m}$ helical microswimmer as the operating object.
磁控微型给药机器人在生物医学领域具有广阔的应用前景。磁致动系统的构建是实现微纳游泳器自动控制的重要步骤。然而，磁致动系统的构建仍然面临着挑战;例如，磁场不能立即关闭，磁场在工作空间中的分布不均匀，工作空间有限，反馈不方便。针对上述问题，本文介绍了一种基于八轴电磁线圈磁控制系统的设计方法，该系统可以复合驱动微型机器人，保证其五自由度运动。此外，在工作空间磁场足够均匀、磁场强度足够大的条件下，可以尽可能地减小系统的整体尺寸。最后，在实验部分，通过磁场仿真和测量验证了磁场的均匀性，然后以$75 mu mathm {m}$螺旋微游泳器为操作对象，实现了方形轨迹的路径跟随。
{"title":"Design of a Miniaturized Magnetic Actuation System for Motion Control of Micro/Nano Swimming Robots","authors":"Liwen Sun, Huaping Wang, Qing Shi, Siyu Guo, Zhiqiao Gao, Tao Sun, Qiang Huang, Toshio Fukuda","doi":"10.1109/RCAR54675.2022.9872234","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872234","url":null,"abstract":"Magnetically controlled microrobots for drug delivery and noninvasive treatment have great potential applications in the biomedical field in the future. The construction of the magnetic actuation system is an important step to realize the automated control of micro/nano swimmers. However, the construction of a magnetic actuation system still faces challenges; for example, the magnetic field cannot be turned off immediately, the distribution of the magnetic field in the workspace is not uniform, the working space is limited and the feedback is inconvenient. In view of the above problems, a design method based on an eight-axis electromagnetic coil magnetic control system is introduced in this paper, which can compositely actuate the microrobot and ensure movement with five degrees of freedom. In addition, the overall size of the system can be reduced as much as possible under the condition that the magnetic field in the workspace is sufficiently uniform and the magnetic field intensity is sufficiently large. Finally, in the experimental part, the magnetic field uniformity is verified by magnetic field simulation and measurement, and then the path following of the square trajectory is realized with the $75 mu mathrm{m}$ helical microswimmer as the operating object.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114133660","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-17DOI: 10.1109/RCAR54675.2022.9872298
Shuai Zhang, Bo Ouyang, Xian He, Xin Yuan, Shanlin Yang
Nurses recognize facial expressions or eye motions to monitor a patient’s condition in the intensive care unit (ICU), for example, pain, agitation, and delirium. However, there are no instruments that can record the facial expression or eye motion accurately like an ECG monitor. To tackle this issue, we develop a face tracking strategy using a 7-DOF robot arm with a camera mounted on the end-effector. First, we constrain the linear and angular velocities of head motion intention to ellipsoids which are determined by the patient’s head pose and the geometry of hospital beds, named head motion intention ellipsoids (HMIEs). Moreover, we defined manipulability ellipsoids (MEs) of the 7-DOF robot arm based on Jacobian matrix, which is adjusted in the null space during the tracking. We calculate the optimal configuration of the camera with the feedback of the head configuration while minimizing the difference between HMIEs and MEs. Simulation experimental results verified that the proposed face tracking strategy outperforms the visual servoing control strategy only based on the pseudo-inverse of the Jacobian matrix.
{"title":"Face Tracking Strategy Based on Manipulability of a 7-DOF Robot Arm and Head Motion Intention Ellipsoids","authors":"Shuai Zhang, Bo Ouyang, Xian He, Xin Yuan, Shanlin Yang","doi":"10.1109/RCAR54675.2022.9872298","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872298","url":null,"abstract":"Nurses recognize facial expressions or eye motions to monitor a patient’s condition in the intensive care unit (ICU), for example, pain, agitation, and delirium. However, there are no instruments that can record the facial expression or eye motion accurately like an ECG monitor. To tackle this issue, we develop a face tracking strategy using a 7-DOF robot arm with a camera mounted on the end-effector. First, we constrain the linear and angular velocities of head motion intention to ellipsoids which are determined by the patient’s head pose and the geometry of hospital beds, named head motion intention ellipsoids (HMIEs). Moreover, we defined manipulability ellipsoids (MEs) of the 7-DOF robot arm based on Jacobian matrix, which is adjusted in the null space during the tracking. We calculate the optimal configuration of the camera with the feedback of the head configuration while minimizing the difference between HMIEs and MEs. Simulation experimental results verified that the proposed face tracking strategy outperforms the visual servoing control strategy only based on the pseudo-inverse of the Jacobian matrix.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127891765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-17DOI: 10.1109/RCAR54675.2022.9872295
Zuming He, Ruili Dong, Yonghong Tan
In this paper, in order to improve the unexpected dynamic performance and input hysteresis of MEMS electromagnetic scanning micromirror (MEMS-ESM), a BP neural network PID control (BP-PID) scheme is adopted. Firstly, the BP-PID controller is designed, and since then PID parameters is self-adjusted by tracking error. Finally, the results of experiment show that the BP-PID control can improve the unexpected dynamic performance of the MEMS-ESM, with faster response speed and smaller tracking error.
{"title":"BP Neural Network PID Control Scheme for Electromagnetic Scanning Micromirror*","authors":"Zuming He, Ruili Dong, Yonghong Tan","doi":"10.1109/RCAR54675.2022.9872295","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872295","url":null,"abstract":"In this paper, in order to improve the unexpected dynamic performance and input hysteresis of MEMS electromagnetic scanning micromirror (MEMS-ESM), a BP neural network PID control (BP-PID) scheme is adopted. Firstly, the BP-PID controller is designed, and since then PID parameters is self-adjusted by tracking error. Finally, the results of experiment show that the BP-PID control can improve the unexpected dynamic performance of the MEMS-ESM, with faster response speed and smaller tracking error.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121742187","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-17DOI: 10.1109/RCAR54675.2022.9872260
Shiping Zuo, Jianfeng Li, Mingjie Dong, Guangsheng Li, Ran Jiao, Guotong Li
Foot-ankle deformity is one of the common complaints in orthopaedic surgery. The external fixator has been selected as the medical apparatus to help with gradual correction, and the configuration may affect the final correction results. It is meaningful to design novel parallel external fixator with deformity-targeting property. Taking the main foot-ankle deformities with four corrective degree-of-freedom (c-DOF) as research object, a 4-DOF parallel configuration is proposed in this paper. Considering several applicable conditions, lower-mobility kinematic struts are selected to provide desired constraints. Then, inverse kinematic model and Jacobian matrix are derived. Finally, after the structural design, clinical case is simulated to prove the applicability of the parallel external fixator.
{"title":"A 4-DOF Parallel External fixator for Foot-Ankle Deformity Correction","authors":"Shiping Zuo, Jianfeng Li, Mingjie Dong, Guangsheng Li, Ran Jiao, Guotong Li","doi":"10.1109/RCAR54675.2022.9872260","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872260","url":null,"abstract":"Foot-ankle deformity is one of the common complaints in orthopaedic surgery. The external fixator has been selected as the medical apparatus to help with gradual correction, and the configuration may affect the final correction results. It is meaningful to design novel parallel external fixator with deformity-targeting property. Taking the main foot-ankle deformities with four corrective degree-of-freedom (c-DOF) as research object, a 4-DOF parallel configuration is proposed in this paper. Considering several applicable conditions, lower-mobility kinematic struts are selected to provide desired constraints. Then, inverse kinematic model and Jacobian matrix are derived. Finally, after the structural design, clinical case is simulated to prove the applicability of the parallel external fixator.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132207256","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-17DOI: 10.1109/RCAR54675.2022.9872265
Cai Chen, Fulai Peng, Yue Sun, Danyang Lv, Ningling Zhang, Xingwei Wang, Lin Wang
The sudden epileptic seizures may not only cause accidental injuries to the patient, but also lead to psychological trauma. It is crucial to predict the onset of a seizure before it occurs. Although the current researches could achieve relatively high prediction performance, there still remain some challenges in the practical scenes, such as class-imbalance problem between pre-ictal and inter-ictal samples, manual hyperparameter tuning problem, etc. This paper proposes a feature-enhancing strategy combining automatic machine learning method to solve these problems. Firstly, the EEG signals are divided into preictal and interictal stages, and then separated into five sub-bands by the pre-processing stage. Then, the features are extracted from the preprocessed signals, followed by feature smoothing and feature augmentation process, which we employ conditional tabular generative adversarial network (CTGAN) to generate the preictal samples. Finally, the processed features are fed into the automatic machine learning (Auto-ML) for seizure prediction. The CHB-MIT EEG dataset is used in this study to evaluate the performance of our proposed method. The combination CTGAN and K-nearest neighbors (KNN), logistic regression (LR), Naive Bayes (NB) classifier and multilayer perceptron (MLP) achieved an average precision of 0.97, 0.94, 0.87 and 0.95, respectively. Auto-ML combined with CTGAN outperforms traditional machine learning models in seizure prediction, with an average accuracy of 99%. Results show that feature augmentation strategy and automatic machine learning can improve the epileptic seizures prediction performance.
{"title":"Epileptic Seizure Prediction Based on EEG by Auto-Machine Learning","authors":"Cai Chen, Fulai Peng, Yue Sun, Danyang Lv, Ningling Zhang, Xingwei Wang, Lin Wang","doi":"10.1109/RCAR54675.2022.9872265","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872265","url":null,"abstract":"The sudden epileptic seizures may not only cause accidental injuries to the patient, but also lead to psychological trauma. It is crucial to predict the onset of a seizure before it occurs. Although the current researches could achieve relatively high prediction performance, there still remain some challenges in the practical scenes, such as class-imbalance problem between pre-ictal and inter-ictal samples, manual hyperparameter tuning problem, etc. This paper proposes a feature-enhancing strategy combining automatic machine learning method to solve these problems. Firstly, the EEG signals are divided into preictal and interictal stages, and then separated into five sub-bands by the pre-processing stage. Then, the features are extracted from the preprocessed signals, followed by feature smoothing and feature augmentation process, which we employ conditional tabular generative adversarial network (CTGAN) to generate the preictal samples. Finally, the processed features are fed into the automatic machine learning (Auto-ML) for seizure prediction. The CHB-MIT EEG dataset is used in this study to evaluate the performance of our proposed method. The combination CTGAN and K-nearest neighbors (KNN), logistic regression (LR), Naive Bayes (NB) classifier and multilayer perceptron (MLP) achieved an average precision of 0.97, 0.94, 0.87 and 0.95, respectively. Auto-ML combined with CTGAN outperforms traditional machine learning models in seizure prediction, with an average accuracy of 99%. Results show that feature augmentation strategy and automatic machine learning can improve the epileptic seizures prediction performance.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133483927","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-17DOI: 10.1109/RCAR54675.2022.9872249
Yaozhong Cao, Dalei Song, Zhan Wang, Yu Wang
This paper aims to achieve the hovering control of an underwater vehicle with uncertain parameters based on cascade PID. First, the inner loops for each channels are designed respectively. Then, the outer loops including depth controller, horizontal position controller and yaw controller are designed. The vector transformation method are used to force the vehicle to approach the target point along approximately straight line in the horizontal plane. Simulation results demonstrate the stability and validity of the proposed method.
{"title":"Hovering Control of an Underwater Vehicle","authors":"Yaozhong Cao, Dalei Song, Zhan Wang, Yu Wang","doi":"10.1109/RCAR54675.2022.9872249","DOIUrl":"https://doi.org/10.1109/RCAR54675.2022.9872249","url":null,"abstract":"This paper aims to achieve the hovering control of an underwater vehicle with uncertain parameters based on cascade PID. First, the inner loops for each channels are designed respectively. Then, the outer loops including depth controller, horizontal position controller and yaw controller are designed. The vector transformation method are used to force the vehicle to approach the target point along approximately straight line in the horizontal plane. Simulation results demonstrate the stability and validity of the proposed method.","PeriodicalId":304963,"journal":{"name":"2022 IEEE International Conference on Real-time Computing and Robotics (RCAR)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133526493","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：