Pub Date : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649860
J. Y. Kim, J. Cho, J. H. Kwon
Globally, the use of mobile robots is increasing in various fields such as delivery, logistics, and medical care, and accordingly, the importance of technology to efficiently utilize the number of autonomous mobile robots (AMR) rather than just one is also growing. Because AMR has high autonomy compared to automated guided vehicles (AGV), it has high usability but requires more difficult technology. Therefore, in this paper, we suggest a method to solve the deadlock situation and reduce task time in a condition where multiple robots work simultaneously in a narrow space. For this, a normalized grid area was assumed, and the Mixed Integer Programming (MIP) model was designed to solve the problem by generating the optimal path for each robot in the deadlock situation. In addition, to verify the designed model, experiments were conducted in an environment very similar to the real world using the Robot Operating System (ROS) and GAZEBO simulation, and finally verified the performance of the proposed method.
{"title":"Multi-Robot Traffic Management using MIP Path Negotiation Scheduler","authors":"J. Y. Kim, J. Cho, J. H. Kwon","doi":"10.23919/ICCAS52745.2021.9649860","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649860","url":null,"abstract":"Globally, the use of mobile robots is increasing in various fields such as delivery, logistics, and medical care, and accordingly, the importance of technology to efficiently utilize the number of autonomous mobile robots (AMR) rather than just one is also growing. Because AMR has high autonomy compared to automated guided vehicles (AGV), it has high usability but requires more difficult technology. Therefore, in this paper, we suggest a method to solve the deadlock situation and reduce task time in a condition where multiple robots work simultaneously in a narrow space. For this, a normalized grid area was assumed, and the Mixed Integer Programming (MIP) model was designed to solve the problem by generating the optimal path for each robot in the deadlock situation. In addition, to verify the designed model, experiments were conducted in an environment very similar to the real world using the Robot Operating System (ROS) and GAZEBO simulation, and finally verified the performance of the proposed method.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"312 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133759179","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649958
M. Ozaki, Teruo Yamaguchi
In recent years, there are more and more opportunities for robots to be active in society. It is necessary to develop a system that processes information that enters the robot's vision faster and more accurately. In the development of visual sensors, the calculation speed and accuracy of the measurement are issues that should be improved. Parallel trade off of the processing was used as a means of speeding up the calculation. In this research, we have developed a system that can measure optical flow with an embedded microprocessor, to detect obstacles in real time and accurately detect its own movement speed on a small robot. We have conducted two experiments, one in which the camera was fixed and a moving subject was photographed, and the other in which the camera was moved and photographed. Experimented results show that was possible to reduce the calculation time by about one-third for real-time processing, and it was possible to detect the difference in the speed of the camera itself.
{"title":"High-speed embedded optical flow measurement system for real-time use","authors":"M. Ozaki, Teruo Yamaguchi","doi":"10.23919/ICCAS52745.2021.9649958","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649958","url":null,"abstract":"In recent years, there are more and more opportunities for robots to be active in society. It is necessary to develop a system that processes information that enters the robot's vision faster and more accurately. In the development of visual sensors, the calculation speed and accuracy of the measurement are issues that should be improved. Parallel trade off of the processing was used as a means of speeding up the calculation. In this research, we have developed a system that can measure optical flow with an embedded microprocessor, to detect obstacles in real time and accurately detect its own movement speed on a small robot. We have conducted two experiments, one in which the camera was fixed and a moving subject was photographed, and the other in which the camera was moved and photographed. Experimented results show that was possible to reduce the calculation time by about one-third for real-time processing, and it was possible to detect the difference in the speed of the camera itself.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"127 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124222757","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649858
M. Salman, H. Khan, S. Abbasi, Min Choel Lee
The accurate dynamic model of the robot manipulator is essential for better control performance. Dynamic modelling refers to deriving such equations that explicitly describe the relationship between force and motion in a system. There are several methods to solve for the dynamics of the system. The main concern while doing dynamics is to check its reliability and verification. In this paper, Lagrange Euler (L-E) method is used to dynamically model a five-degrees-of-freedom (DOF) robot manipulator and verified through 3D model of robot in Simulink, simscape multibody toolbox. Followed by the implementation of a proportional-integral-derivative (PID) control with a low pass filter ‘N’ (PIDN) on the derived system. The low pass filter enhances the control performance by eliminating the effect of derivative term's fluctuations (noise). Furthermore, Ziegler Nichols's closed-loop tuning method is used to optimally tune the PID control. The derived model and control algorithm simulations have been performed in MATLAB. The simulations results illustrated that the derived model is controllable which indicates the correctness of the derived model.
{"title":"Dynamics Analysis and Control of 5 DOF Robot Manipulator","authors":"M. Salman, H. Khan, S. Abbasi, Min Choel Lee","doi":"10.23919/ICCAS52745.2021.9649858","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649858","url":null,"abstract":"The accurate dynamic model of the robot manipulator is essential for better control performance. Dynamic modelling refers to deriving such equations that explicitly describe the relationship between force and motion in a system. There are several methods to solve for the dynamics of the system. The main concern while doing dynamics is to check its reliability and verification. In this paper, Lagrange Euler (L-E) method is used to dynamically model a five-degrees-of-freedom (DOF) robot manipulator and verified through 3D model of robot in Simulink, simscape multibody toolbox. Followed by the implementation of a proportional-integral-derivative (PID) control with a low pass filter ‘N’ (PIDN) on the derived system. The low pass filter enhances the control performance by eliminating the effect of derivative term's fluctuations (noise). Furthermore, Ziegler Nichols's closed-loop tuning method is used to optimally tune the PID control. The derived model and control algorithm simulations have been performed in MATLAB. The simulations results illustrated that the derived model is controllable which indicates the correctness of the derived model.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"91 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124240484","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9650005
K. T. Nguyen, Gwan-Fyo Go, M. Hoang, Jincheol Ha, Jong-Oh Park, Chang-sei Kim
This paper presents an optimized design and development of the electromagnetic actuation (EMA) system, which can generate a strong electromagnetic field to control micro/nanomagnetic objects. The EMA system consists of nine electromagnetic sources, so called Ennead Electromagnetic Actuation (EnEMA) system. The system configuration is designed based on our proposed optimization algorithm considering the design constraints given by the user, and the optimal configuration has the most isotropic and the strongest generated field among all available candidates considering the same design constraints. The EnEMA system can control the magnetic object in 5-DoFs without singularity, which is capable of multi-task control of the micro/nanomagnetic objects. The maximum achievable magnetic field and gradient field of the system are 174 mT and 5 T/m. A prototype of the system was constructed, and calibrated, which was then used to perform precise control of magnetic particle in three-dimensional space. The system is ready for in-vivo in small animal such as rats for targeted drug delivery.
{"title":"The EnEMA System-Ennead Electromagnetic Actuation System-Towards locomotion control for targeted drug delivery","authors":"K. T. Nguyen, Gwan-Fyo Go, M. Hoang, Jincheol Ha, Jong-Oh Park, Chang-sei Kim","doi":"10.23919/ICCAS52745.2021.9650005","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9650005","url":null,"abstract":"This paper presents an optimized design and development of the electromagnetic actuation (EMA) system, which can generate a strong electromagnetic field to control micro/nanomagnetic objects. The EMA system consists of nine electromagnetic sources, so called Ennead Electromagnetic Actuation (EnEMA) system. The system configuration is designed based on our proposed optimization algorithm considering the design constraints given by the user, and the optimal configuration has the most isotropic and the strongest generated field among all available candidates considering the same design constraints. The EnEMA system can control the magnetic object in 5-DoFs without singularity, which is capable of multi-task control of the micro/nanomagnetic objects. The maximum achievable magnetic field and gradient field of the system are 174 mT and 5 T/m. A prototype of the system was constructed, and calibrated, which was then used to perform precise control of magnetic particle in three-dimensional space. The system is ready for in-vivo in small animal such as rats for targeted drug delivery.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"170 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114734213","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649792
S. Manzoor, Eun-jin Kim, Gun-Gyo In, Tae-Yong Kuc
The performance evaluation of an AI network model is the important part for building an effective solution before its deployment in real-world on the robot. In our study, we have implemented YOLOv3-tiny and YOLOv4-tiny darknet based frameworks for performance evaluation of the elevator button recognition task and tested both variants on image and video datasets. The objective of our study is two-fold: First, to overcome the limitation of elevator buttons dataset by creating new dataset and increasing its quantity without compromising the quality; Second, to provide a comparative analysis through experimental results and the performance evaluation of both detectors using four machine learning metrics. The purpose of our work is to assist the researchers and developers in decision making of suitable detector selection for deployment in the elevator robot towards button recognition application. The results show that YOLOv4-tiny outperforms YOLOv3-tiny with an overall accuracy of 98.60% compared to 97.91% at 0.5 IoU.
{"title":"Performance Evaluation of YOLOv3 and YOLOv4 Detectors on Elevator Button Dataset for Mobile Robot","authors":"S. Manzoor, Eun-jin Kim, Gun-Gyo In, Tae-Yong Kuc","doi":"10.23919/ICCAS52745.2021.9649792","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649792","url":null,"abstract":"The performance evaluation of an AI network model is the important part for building an effective solution before its deployment in real-world on the robot. In our study, we have implemented YOLOv3-tiny and YOLOv4-tiny darknet based frameworks for performance evaluation of the elevator button recognition task and tested both variants on image and video datasets. The objective of our study is two-fold: First, to overcome the limitation of elevator buttons dataset by creating new dataset and increasing its quantity without compromising the quality; Second, to provide a comparative analysis through experimental results and the performance evaluation of both detectors using four machine learning metrics. The purpose of our work is to assist the researchers and developers in decision making of suitable detector selection for deployment in the elevator robot towards button recognition application. The results show that YOLOv4-tiny outperforms YOLOv3-tiny with an overall accuracy of 98.60% compared to 97.91% at 0.5 IoU.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"93 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116263581","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649772
J. Jeong, Sun Young Kim, B. Kwon
Vehicle-to-everything (V2X) communication is an essential technology for autonomous driving above Level 4. This paper presents preview control of automotive actives suspension systems to improve ride comfort using vehicle-to-vehicle (V2V) communication. Generally, while preview-controlled active suspension systems have even greater potential than feedback-controlled systems, their main challenge is obtaining preview information of the road profile ahead. A critical drawback of the “look-ahead” sensors is an increased risk of incorrect detection influenced by water, snow, and other soft obstacles on the road. In this work, a preview control algorithm for active suspension systems has been developed without information about the road elevation. The proposed controller is designed with the future disturbance information, vertical wheel acceleration transmitted from preceding vehicle via V2V communication. For active suspension control, an optimal preview controller is designed with the reduced linear half-car model. The performance of the proposed controller has been validated via computer simulation studies. It is shown that the vehicle ride comfort can be improved by the proposed preview control approach more than by the feedback method.
{"title":"Preview Control of Automotive Active Suspension Systems to Improve Ride Comfort Using V2V Communication","authors":"J. Jeong, Sun Young Kim, B. Kwon","doi":"10.23919/ICCAS52745.2021.9649772","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649772","url":null,"abstract":"Vehicle-to-everything (V2X) communication is an essential technology for autonomous driving above Level 4. This paper presents preview control of automotive actives suspension systems to improve ride comfort using vehicle-to-vehicle (V2V) communication. Generally, while preview-controlled active suspension systems have even greater potential than feedback-controlled systems, their main challenge is obtaining preview information of the road profile ahead. A critical drawback of the “look-ahead” sensors is an increased risk of incorrect detection influenced by water, snow, and other soft obstacles on the road. In this work, a preview control algorithm for active suspension systems has been developed without information about the road elevation. The proposed controller is designed with the future disturbance information, vertical wheel acceleration transmitted from preceding vehicle via V2V communication. For active suspension control, an optimal preview controller is designed with the reduced linear half-car model. The performance of the proposed controller has been validated via computer simulation studies. It is shown that the vehicle ride comfort can be improved by the proposed preview control approach more than by the feedback method.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128235893","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649869
Oe Ryung Kang, Jung Hoon Kim
This paper is concerned with a new approach to sliding mode control of a robot manipulator with its computed torque treatment. In other words, a new control law is developed by combining the characteristics of the sliding mode control and the computed torque control. The theoretical effectiveness of the new control approach is validated by using the notion of input-to-state stability (ISS), in which the function corresponding to a sliding surface variable is shown to be bounded for bounded modeling errors occurred from the computed torque treatment. The practical effectiveness of the new control approach is also demonstrated through an experiment result for 6-link robotic manipulators.
{"title":"A New Approach to Sliding Mode Control of a Robot System with Its Computed Torque Treatment","authors":"Oe Ryung Kang, Jung Hoon Kim","doi":"10.23919/ICCAS52745.2021.9649869","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649869","url":null,"abstract":"This paper is concerned with a new approach to sliding mode control of a robot manipulator with its computed torque treatment. In other words, a new control law is developed by combining the characteristics of the sliding mode control and the computed torque control. The theoretical effectiveness of the new control approach is validated by using the notion of input-to-state stability (ISS), in which the function corresponding to a sliding surface variable is shown to be bounded for bounded modeling errors occurred from the computed torque treatment. The practical effectiveness of the new control approach is also demonstrated through an experiment result for 6-link robotic manipulators.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128426780","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649815
Daewon Park, H. Jeong, Nguyen Ngoc Phi, S. Hong
This paper presents a robust formation tracking control algorithm for multi-agent system in a finite time. For faster convergence and less steady state error than the existing formation tracking algorithm, we first propose a fast and robust formation tracking control algorithm by using the fast terminal sliding mode control. Then, a disturbance observer algorithm is added to formation tracking control law to handle external disturbances. The stability of the algorithm is verified by Lyapunov theory. The numerical simulation is proposed to show the effectiveness of proposed method using MATLAB environment.
{"title":"Formation Tracking Control Algorithm for Multi-agent Combined with Fast Terminal Sliding Mode Control And Disturbance Observer","authors":"Daewon Park, H. Jeong, Nguyen Ngoc Phi, S. Hong","doi":"10.23919/ICCAS52745.2021.9649815","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649815","url":null,"abstract":"This paper presents a robust formation tracking control algorithm for multi-agent system in a finite time. For faster convergence and less steady state error than the existing formation tracking algorithm, we first propose a fast and robust formation tracking control algorithm by using the fast terminal sliding mode control. Then, a disturbance observer algorithm is added to formation tracking control law to handle external disturbances. The stability of the algorithm is verified by Lyapunov theory. The numerical simulation is proposed to show the effectiveness of proposed method using MATLAB environment.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127005608","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649865
Debadrata Sarkar, S. Chakraborty, Aman Arora, Soumen Sen
Physically interactive applications like tissue sampling have become essential in present day minimally invasive endoscopy/colonoscopy applications. This calls for specialized steering heads with force exertion ability, still maintaining simplified actuation mechanisms having inherent softness and steerability. This article presents development of soft hyperelastic material (silicone) based pneumatically actuated steering head with improved performance achieved through O-ring reinforcements. The design attains omnidirectional bending operations as well as extension capability enabling needle insertion by the steering head. The ring reinforcement not only contains the bulging effect but also enhances extension ability and range of bending. Additionally, it improves stiffness, structural stability and force exertion capability. The presented work implements two layer design principle - the inner actuator is made from silicone of harder shore hardness followed by a thin outer layer of softer silicone, embedding the O-rings in between. The steering head has been designed iteratively through in depth Finite Element modeling and analyses after obtaining material models of silicones experimentally. The actuator has undergone detailed characterization through simulation for its workspace and blocked-tip force capabilities at various configurations. The characteristics have been experimentally validated on the developed prototype.
{"title":"Ring Reinforced Silicone based Steering Head for Endoscopy like Applications: FEM simulation, development and force characterization","authors":"Debadrata Sarkar, S. Chakraborty, Aman Arora, Soumen Sen","doi":"10.23919/ICCAS52745.2021.9649865","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649865","url":null,"abstract":"Physically interactive applications like tissue sampling have become essential in present day minimally invasive endoscopy/colonoscopy applications. This calls for specialized steering heads with force exertion ability, still maintaining simplified actuation mechanisms having inherent softness and steerability. This article presents development of soft hyperelastic material (silicone) based pneumatically actuated steering head with improved performance achieved through O-ring reinforcements. The design attains omnidirectional bending operations as well as extension capability enabling needle insertion by the steering head. The ring reinforcement not only contains the bulging effect but also enhances extension ability and range of bending. Additionally, it improves stiffness, structural stability and force exertion capability. The presented work implements two layer design principle - the inner actuator is made from silicone of harder shore hardness followed by a thin outer layer of softer silicone, embedding the O-rings in between. The steering head has been designed iteratively through in depth Finite Element modeling and analyses after obtaining material models of silicones experimentally. The actuator has undergone detailed characterization through simulation for its workspace and blocked-tip force capabilities at various configurations. The characteristics have been experimentally validated on the developed prototype.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"158 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127293666","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9650004
Jeong-Soo Kim, Kyungmin Lee, Hyeongkeun Lee, Hunmin Yang, Se-Yoon Oh
The existence of physical-world adversarial examples such as adversarial patches proves the vulnerability of real-world deep learning systems. Therefore, it is essential to develop efficient adversarial attack algorithms to identify potential risks and build a robust system. The patch-based physical adversarial attack has shown its effectiveness in attacking neural network-based object detectors. However, the generated patches are quite perceptible for humans, violating the fundamental assumption of adversarial examples. In this work, we present task-specific loss functions that can generate imperceptible adversarial patches based on camouflaged patterns. First, we propose a constrained optimization method with two camouflage assessment metrics to quantify camouflage performance. Then, we show the regularization with those metrics can help generate the adversarial patches based on camouflage patterns. Furthermore, we validate our methods with various experiments and show that we can generate natural-style camouflaged adversarial patches with comparable attack performance.
{"title":"Camouflaged Adversarial Attack on Object Detector","authors":"Jeong-Soo Kim, Kyungmin Lee, Hyeongkeun Lee, Hunmin Yang, Se-Yoon Oh","doi":"10.23919/ICCAS52745.2021.9650004","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9650004","url":null,"abstract":"The existence of physical-world adversarial examples such as adversarial patches proves the vulnerability of real-world deep learning systems. Therefore, it is essential to develop efficient adversarial attack algorithms to identify potential risks and build a robust system. The patch-based physical adversarial attack has shown its effectiveness in attacking neural network-based object detectors. However, the generated patches are quite perceptible for humans, violating the fundamental assumption of adversarial examples. In this work, we present task-specific loss functions that can generate imperceptible adversarial patches based on camouflaged patterns. First, we propose a constrained optimization method with two camouflage assessment metrics to quantify camouflage performance. Then, we show the regularization with those metrics can help generate the adversarial patches based on camouflage patterns. Furthermore, we validate our methods with various experiments and show that we can generate natural-style camouflaged adversarial patches with comparable attack performance.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130062402","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：