Pub Date : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205770
Shuhao Liang
In-Circuit-Test (ICT) is an inevitable process for sustaining the quality of Printed Circuit Boards (PCBs) in the assembly and fabrication process. Applying automation to reduce labor and preventing errors in ICT has been studying by academics and industries for decades. Here we demonstrate a robot centric ICT testing system that integrates the peripheral equipment, also including the shop flow control system (SPCS). The graphic programming software – LabVIEW exploits to integrate robot arm, in-circuit test machine, PLC, HMI, and barcode reader. Communication among the facilities and error handling are the main challenges in the automated ICT system development. Heterogeneous communication protocols and third-party devices with unique syntax have caused some programming difficulties. The challenge of error handling is that it might be on hardware, software, or communication. Moreover, these errors may have occurred at 5-6 different facilities with chain effects. The robot arm dominates the main control sequence from the test start to finish. As a result, a steady automated ICT test system with real-time status monitoring has presented. That assists field personnel in eliminating problems quickly and promotes overall production line operation efficiency.
{"title":"A paradigm of automatic ICT testing system development in practice","authors":"Shuhao Liang","doi":"10.1109/ARIS50834.2020.9205770","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205770","url":null,"abstract":"In-Circuit-Test (ICT) is an inevitable process for sustaining the quality of Printed Circuit Boards (PCBs) in the assembly and fabrication process. Applying automation to reduce labor and preventing errors in ICT has been studying by academics and industries for decades. Here we demonstrate a robot centric ICT testing system that integrates the peripheral equipment, also including the shop flow control system (SPCS). The graphic programming software – LabVIEW exploits to integrate robot arm, in-circuit test machine, PLC, HMI, and barcode reader. Communication among the facilities and error handling are the main challenges in the automated ICT system development. Heterogeneous communication protocols and third-party devices with unique syntax have caused some programming difficulties. The challenge of error handling is that it might be on hardware, software, or communication. Moreover, these errors may have occurred at 5-6 different facilities with chain effects. The robot arm dominates the main control sequence from the test start to finish. As a result, a steady automated ICT test system with real-time status monitoring has presented. That assists field personnel in eliminating problems quickly and promotes overall production line operation efficiency.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123848489","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205791
J. Yang, Ui-Kai Chen, Kai-Chu Chang, Ying-Jen Chen
This paper proposes a robotic grasp detection technique by integrating you only look once (YOLO) deep neural network (DNN) and a grasp detection DNN. In this world, there are many people who cannot move their own bodies. The reason may be an accident or physical deterioration. So we need to invest more human resources to assist their lives. With new technological advances, robots are gradually able to perfectly replicate human movements. Hence, we intend to design a remote-control fetching robot. The system combines internet of things (IoT) technology, and users can use intelligent devices to control this robot with robotic arm to get the items they want. This paper focus on detecting the grasp of robotic arm by integrating YOLO and grasp detection DNNs. At first, YOLO V-v3 is applied to achieve object detection. Then a robotic grasp detection DNN is proposed to detect the robotic grasp. After that, the point cloud information of this object is utilized to calculate the normal vector of the grasp position such that the robotic arm can fetch the target along the normal vector. Finally, experiment results are given to show the practicality of the proposed robotic grasp detection Technique.
本文提出了一种将you only look once (YOLO)深度神经网络(DNN)与抓取检测深度神经网络相结合的机器人抓取检测技术。在这个世界上，有很多人无法移动自己的身体。原因可能是意外事故或身体恶化。所以我们需要投入更多的人力资源来帮助他们的生活。随着新技术的进步，机器人逐渐能够完美地复制人类的动作。因此，我们打算设计一个遥控抓取机器人。该系统结合了物联网(IoT)技术，用户可以使用智能设备控制机器人的机械臂，以获得他们想要的物品。本文将YOLO和抓取检测dnn相结合，对机械臂抓取进行检测。首先使用YOLO V-v3实现目标检测。然后提出了一种机器人抓取检测深度神经网络来检测机器人抓取。然后利用该目标的点云信息计算抓取位置的法向量，使机械臂沿着法向量获取目标。最后给出了实验结果，验证了所提机器人抓取检测技术的实用性。
{"title":"A Novel Robotic Grasp Detection Technique by Integrating YOLO and Grasp Detection Deep Neural Networks *","authors":"J. Yang, Ui-Kai Chen, Kai-Chu Chang, Ying-Jen Chen","doi":"10.1109/ARIS50834.2020.9205791","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205791","url":null,"abstract":"This paper proposes a robotic grasp detection technique by integrating you only look once (YOLO) deep neural network (DNN) and a grasp detection DNN. In this world, there are many people who cannot move their own bodies. The reason may be an accident or physical deterioration. So we need to invest more human resources to assist their lives. With new technological advances, robots are gradually able to perfectly replicate human movements. Hence, we intend to design a remote-control fetching robot. The system combines internet of things (IoT) technology, and users can use intelligent devices to control this robot with robotic arm to get the items they want. This paper focus on detecting the grasp of robotic arm by integrating YOLO and grasp detection DNNs. At first, YOLO V-v3 is applied to achieve object detection. Then a robotic grasp detection DNN is proposed to detect the robotic grasp. After that, the point cloud information of this object is utilized to calculate the normal vector of the grasp position such that the robotic arm can fetch the target along the normal vector. Finally, experiment results are given to show the practicality of the proposed robotic grasp detection Technique.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124000129","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205776
W. Lai
The proposed continuous-time sigma-delta $(SigmaDelta)$ modulator employing nonreturn-to-zero (NRZ) digital-to-analog converter (DAC) and pulse shaping to achieve the performance of reducing the impact of clock jitter noise reduction is presented. The proposed modulator comprises a third order RC operational-amplifier-based loop filter, 4-bit internal quantizer operating at 160 MHz and three DACs. The NRZ DAC with quantizer excess loop delay compensation is set to be half the sampling period of the quantizer. The $SigmaDelta$ modulator dissipates 10.1 mW at 1.2 V supply voltage is implemented in the TSMC 0.18 um CMOS technology for robotic light communication and intelligent sensor fusion. Measured results illustrate that the $SigmaDelta$ modulator achieves 66.9 dB SNR, a peak 62 dB SNDR and 10.3 ENOB over a 10 MHz band at an over-sampling ratio (OSR) of 8. Including pads, the chip dimension is $0.363mm^{2}.$
提出了一种采用非归零(NRZ)数模转换器(DAC)和脉冲整形的连续时间sigma-delta $(SigmaDelta)$调制器，以达到降低时钟抖动降噪影响的性能。所提出的调制器包括一个基于三阶RC运算放大器的环路滤波器、工作频率为160 MHz的4位内部量化器和三个dac。采用量化器补偿多余环路延迟的NRZ DAC设置为量化器采样周期的一半。$SigmaDelta$调制器在1.2 V电源电压下耗电10.1 mW，采用台积电0.18 um CMOS技术，用于机器人光通信和智能传感器融合。测量结果表明，$SigmaDelta$调制器在过采样比(OSR)为8的情况下，在10 MHz频段内实现了66.9 dB信噪比，峰值62 dB SNDR和10.3 ENOB。包括衬垫在内，芯片尺寸为 $0.363mm^{2}.$
{"title":"Design of Continuous-Time Sigma-Delta Modulator with Noise Reduction for Robotic Light Communication and Sensing","authors":"W. Lai","doi":"10.1109/ARIS50834.2020.9205776","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205776","url":null,"abstract":"The proposed continuous-time sigma-delta $(SigmaDelta)$ modulator employing nonreturn-to-zero (NRZ) digital-to-analog converter (DAC) and pulse shaping to achieve the performance of reducing the impact of clock jitter noise reduction is presented. The proposed modulator comprises a third order RC operational-amplifier-based loop filter, 4-bit internal quantizer operating at 160 MHz and three DACs. The NRZ DAC with quantizer excess loop delay compensation is set to be half the sampling period of the quantizer. The $SigmaDelta$ modulator dissipates 10.1 mW at 1.2 V supply voltage is implemented in the TSMC 0.18 um CMOS technology for robotic light communication and intelligent sensor fusion. Measured results illustrate that the $SigmaDelta$ modulator achieves 66.9 dB SNR, a peak 62 dB SNDR and 10.3 ENOB over a 10 MHz band at an over-sampling ratio (OSR) of 8. Including pads, the chip dimension is $0.363mm^{2}.$","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116888653","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205782
M. Elsisi
The steering control of the autonomous vehicles represents an avital issue in the vehicular system. The model predictive control was proved as an effective controller. However, the representation of the model predictive control (MPC) by a large prediction horizon and control horizon requires a large number of parameters and it is complicated. Discrete-time Laguerre functions can cope with this issue to represent the MPC with few parameters. Whilst, the Laguerre functions require a proper tuning for its parameters in order to provide a good response with MPC. This paper introduces a new design method to tune the parameters of the MPC with the Laguerre function by a new artificial intelligence (AI) technique named social ski driver algorithm (SSDA). The proposed MPC based on the SSDA is applied to adjust the steering angle of an autonomous vehicle including vision dynamics. Further test scenarios are created to ensure the effectiveness of the proposed control to cope with the variations of road curvatures.
{"title":"Model Predictive Control with Laguerre Function based on Social Ski Driver Algorithm for Autonomous Vehicle","authors":"M. Elsisi","doi":"10.1109/ARIS50834.2020.9205782","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205782","url":null,"abstract":"The steering control of the autonomous vehicles represents an avital issue in the vehicular system. The model predictive control was proved as an effective controller. However, the representation of the model predictive control (MPC) by a large prediction horizon and control horizon requires a large number of parameters and it is complicated. Discrete-time Laguerre functions can cope with this issue to represent the MPC with few parameters. Whilst, the Laguerre functions require a proper tuning for its parameters in order to provide a good response with MPC. This paper introduces a new design method to tune the parameters of the MPC with the Laguerre function by a new artificial intelligence (AI) technique named social ski driver algorithm (SSDA). The proposed MPC based on the SSDA is applied to adjust the steering angle of an autonomous vehicle including vision dynamics. Further test scenarios are created to ensure the effectiveness of the proposed control to cope with the variations of road curvatures.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"242 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115507544","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205771
Hsin-Hsiung Huang, Juing-Huei Su, Chyi-Shyong Lee, Hsuan-Hao Li
In this paper, we implement the camera-based approach to identify the image and take the information to guide the minimization movement of the robot with the given starting and target points. The advantage of the paper are as follows, First, the camera is used to guide the robot by the wireless transmission. Second, the inner product-based approach is applied to calculated the distance between two given points. Third, we minimize the error between the calculated distance and the physical distance. Hence, the approach is accurately to guide the robot to move the target by the camera. Experimental results show that the camera-based approach can accurately guide the robot to the target with the minimization movement, which leads the power saving of the battery.
{"title":"Implementation of the Camera-based Approach to Guide the Robot with Minimization Movements","authors":"Hsin-Hsiung Huang, Juing-Huei Su, Chyi-Shyong Lee, Hsuan-Hao Li","doi":"10.1109/ARIS50834.2020.9205771","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205771","url":null,"abstract":"In this paper, we implement the camera-based approach to identify the image and take the information to guide the minimization movement of the robot with the given starting and target points. The advantage of the paper are as follows, First, the camera is used to guide the robot by the wireless transmission. Second, the inner product-based approach is applied to calculated the distance between two given points. Third, we minimize the error between the calculated distance and the physical distance. Hence, the approach is accurately to guide the robot to move the target by the camera. Experimental results show that the camera-based approach can accurately guide the robot to the target with the minimization movement, which leads the power saving of the battery.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127124981","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205785
W. Lai
The article proposes wide tuning voltage-controlled oscillator (VCO) with adopting 4-bit switched capacitor array (SCA). The SCA with a cross-coupled switching pair, varactors and LC circuit at a low supply voltage of 1 V was fabricated in the 0.18-$mu$ m 1P6M CMOS technology for intelligent sensor fusion. Measured results illustrate that at the voltage source of 1 V, the SCA VCO is tunable from 4.47 GHz to 5.95 GHz, corresponding to 28.7 %. The phase noise is -115.8dBc/Hz at 1 MHz offset from 5.8 GHz, the tuning range is 1880 MHz, the FOM is -182.7dBc/Hz, the power consumption is 7.0 mW and the chipset dimension is $0.817times 0.599mm^{2}$
{"title":"Design of 1V CMOS 5.8 GHz VCO with Switched Capacitor Array Tuning for Intelligent Sensor Fusion","authors":"W. Lai","doi":"10.1109/ARIS50834.2020.9205785","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205785","url":null,"abstract":"The article proposes wide tuning voltage-controlled oscillator (VCO) with adopting 4-bit switched capacitor array (SCA). The SCA with a cross-coupled switching pair, varactors and LC circuit at a low supply voltage of 1 V was fabricated in the 0.18-$mu$ m 1P6M CMOS technology for intelligent sensor fusion. Measured results illustrate that at the voltage source of 1 V, the SCA VCO is tunable from 4.47 GHz to 5.95 GHz, corresponding to 28.7 %. The phase noise is -115.8dBc/Hz at 1 MHz offset from 5.8 GHz, the tuning range is 1880 MHz, the FOM is -182.7dBc/Hz, the power consumption is 7.0 mW and the chipset dimension is $0.817times 0.599mm^{2}$","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131635540","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}
Protecting soldiers in the long march of platoon formation is crucial mission in the military operation. Autonomous ground mobile robots can be deployed to carry out such kind of the mission. The primary task is maintaining robot position automatically based on the movement of soldier in a platoon formation. The GPS is employed to determine the soldier current pose. This information is used as an input to create a dynamic convex hull around a platoon. The Proportional-Integral (PI) controller is applied to control each robot so it can move in a desired trajectory. The fuzzy logic control (FLC) is involved to tune the gain of PI controller to optimize the performance. Three protective robots and nine soldiers are used to evaluate the algorithm in simulation. The proposed algorithm will provide a platoon soldiers with optimal protection encirclement and enhance their safety. The simulation results show good performance using the proposed controller.
{"title":"Formation Control for Mobile Robot using Fuzzy - PI Controller","authors":"Min-Fan Ricky Lee, H.P.M Willybrordus, Sukamto, Sharfiden Hassen, Asep Nugroho","doi":"10.1109/ARIS50834.2020.9205789","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205789","url":null,"abstract":"Protecting soldiers in the long march of platoon formation is crucial mission in the military operation. Autonomous ground mobile robots can be deployed to carry out such kind of the mission. The primary task is maintaining robot position automatically based on the movement of soldier in a platoon formation. The GPS is employed to determine the soldier current pose. This information is used as an input to create a dynamic convex hull around a platoon. The Proportional-Integral (PI) controller is applied to control each robot so it can move in a desired trajectory. The fuzzy logic control (FLC) is involved to tune the gain of PI controller to optimize the performance. Three protective robots and nine soldiers are used to evaluate the algorithm in simulation. The proposed algorithm will provide a platoon soldiers with optimal protection encirclement and enhance their safety. The simulation results show good performance using the proposed controller.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"494 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123352341","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205774
Chia-Chin Wang, H. Samani
In this paper the usage of Transfer Learning method for object detection in underwater environment is experienced and evaluated. Deep learning method of YOLO is utilized for detection of different types of fish underwater. A ROV equipped with camera is employed for video streaming underwater and the data has been analyzed on the main computer Our experimental results confirmed improvement in the mAP by 4% using transfer learning.
{"title":"Object Detection using Transfer Learning for Underwater Robot","authors":"Chia-Chin Wang, H. Samani","doi":"10.1109/ARIS50834.2020.9205774","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205774","url":null,"abstract":"In this paper the usage of Transfer Learning method for object detection in underwater environment is experienced and evaluated. Deep learning method of YOLO is utilized for detection of different types of fish underwater. A ROV equipped with camera is employed for video streaming underwater and the data has been analyzed on the main computer Our experimental results confirmed improvement in the mAP by 4% using transfer learning.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123631173","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205772
Min-Fan Ricky Lee, Tzu-Wei Chien
The fatal injury rate for the construction industry is higher than the average for all industries. Recently, researchers have shown an increased interest in occupational safety in the construction industry. However, all the current methods using conventional machine learning with stationary cameras suffer from some severe limitations, perceptual aliasing (e.g., different places/objects can appear identical), occlusion (e.g., place/object appearance changes between visits), seasonal / illumination changes, significant viewpoint changes, etc. This paper proposes a perception module using end-to-end deep-learning and visual SLAM (Simultaneous Localization and Mapping) for an effective and efficient object recognition and navigation using a differential-drive mobile robot. Various deep-learning frameworks and visual navigation strategies with evaluation metrics are implemented and validated for the selection of the best model. The deep-learning model's predictions are evaluated via the metrics (model speed, accuracy, complexity, precision, recall, P-R curve, F1 score). The YOLOv3 shows the best trade-off among all algorithms, 57.9% mean average precision (mAP), in real-world settings, and can process 45 frames per second (FPS) on NVIDIA Jetson TX2 which makes it suitable for real-time detection, as well as a right candidate for deploying the neural network on a mobile robot. The evaluation metrics used for the comparison of laser SLAM are Root Mean Square Error (RMSE). The Google Cartographer SLAM shows the lowest RMSE and acceptable processing time. The experimental results demonstrate that the perception module can meet the requirements of head protection criteria in Occupational Safety and Health Administration (OSHA) standards for construction. To be more precise, this module can effectively detect construction worker's non-hardhat-use in different construction site conditions and can facilitate improved safety inspection and supervision.
建筑业的致命伤害率高于所有行业的平均水平。最近，研究人员对建筑行业的职业安全表现出越来越大的兴趣。然而，目前所有使用固定相机的传统机器学习方法都存在一些严重的局限性，如感知混叠(例如，不同的地方/物体可能看起来相同)、遮挡(例如，两次访问之间的地方/物体外观变化)、季节/光照变化、重大的视点变化等。本文提出了一种基于端到端深度学习和视觉SLAM (Simultaneous Localization and Mapping)的感知模块，用于差分驱动移动机器人的有效和高效的目标识别和导航。实现并验证了各种深度学习框架和带有评估指标的视觉导航策略，以选择最佳模型。深度学习模型的预测通过指标(模型速度、准确性、复杂性、精度、召回率、P-R曲线、F1分数)进行评估。YOLOv3在所有算法中表现出最好的折衷，在现实环境中平均精度(mAP)为57.9%，并且可以在NVIDIA Jetson TX2上每秒处理45帧(FPS)，这使得它适合于实时检测，并且是在移动机器人上部署神经网络的合适候选人。比较激光SLAM的评价指标为均方根误差(RMSE)。b谷歌Cartographer SLAM显示最低RMSE和可接受的处理时间。实验结果表明，感知模块能够满足OSHA (Occupational Safety and Health Administration，职业安全与健康管理局)建筑标准中头部保护标准的要求。更准确地说，该模块可以有效地检测建筑工人在不同施工现场条件下的不戴安全帽情况，便于改进安全检查和监督。
{"title":"Intelligent Robot for Worker Safety Surveillance: Deep Learning Perception and Visual Navigation","authors":"Min-Fan Ricky Lee, Tzu-Wei Chien","doi":"10.1109/ARIS50834.2020.9205772","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205772","url":null,"abstract":"The fatal injury rate for the construction industry is higher than the average for all industries. Recently, researchers have shown an increased interest in occupational safety in the construction industry. However, all the current methods using conventional machine learning with stationary cameras suffer from some severe limitations, perceptual aliasing (e.g., different places/objects can appear identical), occlusion (e.g., place/object appearance changes between visits), seasonal / illumination changes, significant viewpoint changes, etc. This paper proposes a perception module using end-to-end deep-learning and visual SLAM (Simultaneous Localization and Mapping) for an effective and efficient object recognition and navigation using a differential-drive mobile robot. Various deep-learning frameworks and visual navigation strategies with evaluation metrics are implemented and validated for the selection of the best model. The deep-learning model's predictions are evaluated via the metrics (model speed, accuracy, complexity, precision, recall, P-R curve, F1 score). The YOLOv3 shows the best trade-off among all algorithms, 57.9% mean average precision (mAP), in real-world settings, and can process 45 frames per second (FPS) on NVIDIA Jetson TX2 which makes it suitable for real-time detection, as well as a right candidate for deploying the neural network on a mobile robot. The evaluation metrics used for the comparison of laser SLAM are Root Mean Square Error (RMSE). The Google Cartographer SLAM shows the lowest RMSE and acceptable processing time. The experimental results demonstrate that the perception module can meet the requirements of head protection criteria in Occupational Safety and Health Administration (OSHA) standards for construction. To be more precise, this module can effectively detect construction worker's non-hardhat-use in different construction site conditions and can facilitate improved safety inspection and supervision.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121709562","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 : 2020-08-01DOI: 10.1109/ARIS50834.2020.9205790
Min-Fan Ricky Lee, S. K., A. J.
The landing is one of the most dangerous maneuvers in the entirety of the flight phase of an Unmanned Aerial Vehicle (UAVs). Sudden changes in the environment cause issues regarding the stability of the drone, which poses a difficult challenge in landing the UAV precisely. To better the safety of any UAVs flying in urban areas, UAVs should be landed carefully, in a GPS-denied or network-disconnected environment, by using vision and inertial data. This paper presents UAV safe landing system which comprises of three sub-systems for detection of designated landing sites and autonomous pose correction, landing site inspection and landing flight control. This paper deals with vision-based target detection and pose correction system in-depth. The airborne vision system is utilized to recognize certain markers on the landing site. The information from the onboard visual sensors and Inertial Measurement Unit (IMU) is utilized to control and land UAV in a perfect landing trajectory, on a precise location. A series of experiments have been outlined to test and optimize the proposed method using Parrot AR.Drone 2.0.
{"title":"Autonomous Pose Correction and Landing System for Unmanned Aerial Vehicles","authors":"Min-Fan Ricky Lee, S. K., A. J.","doi":"10.1109/ARIS50834.2020.9205790","DOIUrl":"https://doi.org/10.1109/ARIS50834.2020.9205790","url":null,"abstract":"The landing is one of the most dangerous maneuvers in the entirety of the flight phase of an Unmanned Aerial Vehicle (UAVs). Sudden changes in the environment cause issues regarding the stability of the drone, which poses a difficult challenge in landing the UAV precisely. To better the safety of any UAVs flying in urban areas, UAVs should be landed carefully, in a GPS-denied or network-disconnected environment, by using vision and inertial data. This paper presents UAV safe landing system which comprises of three sub-systems for detection of designated landing sites and autonomous pose correction, landing site inspection and landing flight control. This paper deals with vision-based target detection and pose correction system in-depth. The airborne vision system is utilized to recognize certain markers on the landing site. The information from the onboard visual sensors and Inertial Measurement Unit (IMU) is utilized to control and land UAV in a perfect landing trajectory, on a precise location. A series of experiments have been outlined to test and optimize the proposed method using Parrot AR.Drone 2.0.","PeriodicalId":423389,"journal":{"name":"2020 International Conference on Advanced Robotics and Intelligent Systems (ARIS)","volume":"127 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132054450","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：