Pub Date : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649923
Yong-Geon Kim, Min-Woo Na, Jae-Bok Song
When performing robotic assembly, a task should be conducted through force-based control such as impedance control. Using impedance control, it is possible to control the contact force by appropriately adjusting the impedance parameters. However, the impedance parameters should be set by the user because it is difficult to accurately recognize the dynamics of the contact environment, which takes a lot of time because it should be performed whenever the assembly task changes. Moreover, the parameters may not be optimal because it depends on the experience and skill level of the user. To this end, a reinforcement learning-based impedance parameter tuning method is proposed in this study. Since this method uses only the physics-based robotic simulation on the virtual environment, there is no risk of damaging the robots or parts and learning time can be significantly reduced. The proposed method was verified by assembling an HDMI connector with a tolerance of 0.03 mm. Impedance parameters were learned in the virtual environment and transferred to the real environment. Finally, it was confirmed that parameter tuning for impedance without the aid of the user is possible by using the proposed method.
{"title":"Reinforcement Learning-based Sim-to-Real Impedance Parameter Tuning for Robotic Assembly","authors":"Yong-Geon Kim, Min-Woo Na, Jae-Bok Song","doi":"10.23919/ICCAS52745.2021.9649923","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649923","url":null,"abstract":"When performing robotic assembly, a task should be conducted through force-based control such as impedance control. Using impedance control, it is possible to control the contact force by appropriately adjusting the impedance parameters. However, the impedance parameters should be set by the user because it is difficult to accurately recognize the dynamics of the contact environment, which takes a lot of time because it should be performed whenever the assembly task changes. Moreover, the parameters may not be optimal because it depends on the experience and skill level of the user. To this end, a reinforcement learning-based impedance parameter tuning method is proposed in this study. Since this method uses only the physics-based robotic simulation on the virtual environment, there is no risk of damaging the robots or parts and learning time can be significantly reduced. The proposed method was verified by assembling an HDMI connector with a tolerance of 0.03 mm. Impedance parameters were learned in the virtual environment and transferred to the real environment. Finally, it was confirmed that parameter tuning for impedance without the aid of the user is possible by using the proposed method.","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":"131766765","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/ICCAS52745.2021.9649822
Eunho Lee, Youngbae Hwang
Many methods have been proposed to address the real noise, they suffer from restoring the edge regions appropriately. Because most convolutional neural network-based denoising methods capture noise characteristics through pixel loss that only detects contaminated pixels, high frequency components cannot be considered. This causes blurs and artifacts on edge regions which has the high frequency component. In this paper, we apply an edge loss function to the dual adversarial network to deal with this issue. Using the edge loss and the pixel loss together, the network has been improved to restore not only the actual intensity but also the edges effectively.
{"title":"Enhanced Dual Adversarial Network for Real Image Noise Removal and Generation using Edge Loss Function","authors":"Eunho Lee, Youngbae Hwang","doi":"10.23919/ICCAS52745.2021.9649822","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649822","url":null,"abstract":"Many methods have been proposed to address the real noise, they suffer from restoring the edge regions appropriately. Because most convolutional neural network-based denoising methods capture noise characteristics through pixel loss that only detects contaminated pixels, high frequency components cannot be considered. This causes blurs and artifacts on edge regions which has the high frequency component. In this paper, we apply an edge loss function to the dual adversarial network to deal with this issue. Using the edge loss and the pixel loss together, the network has been improved to restore not only the actual intensity but also the edges effectively.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":" 10","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132123710","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/ICCAS52745.2021.9649839
Kanghoon Lee, Kyuree Ahn, Jinkyoo Park
The Unmanned Surface Vehicles (USVs), which operate without a person at the surface, are used in various naval defense missions. Various missions can be conducted efficiently when a swarm of USVs are operated at the same time. However, it is challenging to establish a decentralised control strategy for all USVs. In addition, the strategy must consider various external factors, such as the ocean topography and the number of enemy forces. These difficulties necessitate a scalable and transferable decision-making module. This study proposes an algorithm to derive the decentralised and cooperative control strategy for the USV swarm using graph centric multi-agent reinforcement learning (MARL). The model first expresses the mission situation using a graph considering the various sensor ranges. Each USV agent encodes observed information into localized embedding and then derives coordinated action through communication with the surrounding agent. To derive a cooperative policy, we trained each agent's policy to maximize the team reward. Using the modified prey-predator environment of OpenAI gym, we have analyzed the effect of each component of the proposed model (state embedding, communication, and team reward). The ablation study shows that the proposed model could derive a scalable and transferable control policy of USVs, consistently achieving the highest win ratio.
{"title":"End-to-End control of USV swarm using graph centric Multi-Agent Reinforcement Learning","authors":"Kanghoon Lee, Kyuree Ahn, Jinkyoo Park","doi":"10.23919/ICCAS52745.2021.9649839","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649839","url":null,"abstract":"The Unmanned Surface Vehicles (USVs), which operate without a person at the surface, are used in various naval defense missions. Various missions can be conducted efficiently when a swarm of USVs are operated at the same time. However, it is challenging to establish a decentralised control strategy for all USVs. In addition, the strategy must consider various external factors, such as the ocean topography and the number of enemy forces. These difficulties necessitate a scalable and transferable decision-making module. This study proposes an algorithm to derive the decentralised and cooperative control strategy for the USV swarm using graph centric multi-agent reinforcement learning (MARL). The model first expresses the mission situation using a graph considering the various sensor ranges. Each USV agent encodes observed information into localized embedding and then derives coordinated action through communication with the surrounding agent. To derive a cooperative policy, we trained each agent's policy to maximize the team reward. Using the modified prey-predator environment of OpenAI gym, we have analyzed the effect of each component of the proposed model (state embedding, communication, and team reward). The ablation study shows that the proposed model could derive a scalable and transferable control policy of USVs, consistently achieving the highest win ratio.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"21 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":"134356226","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/ICCAS52745.2021.9649922
Ki-Wook Jung, Chang-Hun Lee, Jun-Seong Lee, Sunghyuck Im, Keejoo Lee, Marco Sagliano, David Seelbinder
This paper aims to propose a new guidance algorithm for a rocket with aerodynamics control for launch operations, based on the concept of the instantaneous impact point (IIP). In this study, the rocket with aerodynamics control is considered with the purpose of reducing dispersion of the impact point after separation of the rocket for safety reasons. Since a very limited aerodynamic maneuverability is typically allowed for the rocket due to the structural limit, a guidance algorithm producing a huge acceleration demand is not desirable. Based on this aspect, the proposed guidance algorithm is derived directly from the underlying principle of the guidance process: forming the collision geometry towards a target point. To be more specific, the collision-ballistic-trajectory where the instantaneous impact point becomes the target point, and the corresponding heading error are first determined using a rapid ballistic trajectory prediction technique. Here, the trajectory prediction method is based on the partial closed-form solutions of the ballistic trajectory equations considering aerodynamic drag and gravity. And then, the proposed guidance algorithm works to nullify the heading error in a finite time, governed by the optimal error dynamics. The key feature of the proposed guidance algorithm lies in its simple implementation and exact collision geometry nature. Hence, the proposed method allows achieving the collision course with minimal guidance command, and it is a desirable property for the guidance algorithm of the rocket with the aerodynamics control. Finally, numerical simulations are conducted to demonstrate the effectiveness of the proposed guidance algorithms.
{"title":"An Instantaneous Impact Point Guidance for Rocket with Aerodynamics Control","authors":"Ki-Wook Jung, Chang-Hun Lee, Jun-Seong Lee, Sunghyuck Im, Keejoo Lee, Marco Sagliano, David Seelbinder","doi":"10.23919/ICCAS52745.2021.9649922","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649922","url":null,"abstract":"This paper aims to propose a new guidance algorithm for a rocket with aerodynamics control for launch operations, based on the concept of the instantaneous impact point (IIP). In this study, the rocket with aerodynamics control is considered with the purpose of reducing dispersion of the impact point after separation of the rocket for safety reasons. Since a very limited aerodynamic maneuverability is typically allowed for the rocket due to the structural limit, a guidance algorithm producing a huge acceleration demand is not desirable. Based on this aspect, the proposed guidance algorithm is derived directly from the underlying principle of the guidance process: forming the collision geometry towards a target point. To be more specific, the collision-ballistic-trajectory where the instantaneous impact point becomes the target point, and the corresponding heading error are first determined using a rapid ballistic trajectory prediction technique. Here, the trajectory prediction method is based on the partial closed-form solutions of the ballistic trajectory equations considering aerodynamic drag and gravity. And then, the proposed guidance algorithm works to nullify the heading error in a finite time, governed by the optimal error dynamics. The key feature of the proposed guidance algorithm lies in its simple implementation and exact collision geometry nature. Hence, the proposed method allows achieving the collision course with minimal guidance command, and it is a desirable property for the guidance algorithm of the rocket with the aerodynamics control. Finally, numerical simulations are conducted to demonstrate the effectiveness of the proposed guidance algorithms.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"62 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":"134444185","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/ICCAS52745.2021.9649902
Yuxiang Zhang, Xiaoling Liang, S. Ge, B. Gao, Tong-heng Lee
Guaranteed safety and performance under various circumstances remain technically critical and practically challenging for the wide deployment of autonomous vehicles. For such safety-critical systems, it will certainly be a requirement that safe performance should be ensured even during the reinforcement learning period in the presence of system uncertainty. To address this issue, a Barrier Lyapunov Function-based safe reinforcement learning algorithm (BLF-SRL) is proposed here for the formulated nonlinear system in strict-feedback form. This approach appropriately arranges the Barrier Lyapunov Function item into the optimized backstepping control method to constrain the state-variables in the designed safety region during learning when unknown bounded system uncertainty exists. More specifically, the overall system control is optimized with the optimized backstepping technique under the framework of Actor-Critic, which optimizes the virtual control in every backstepping subsystem. Wherein, the optimal virtual control is decomposed into Barrier Lyapunov Function items; and also with an adaptive item to be learned with deep neural networks, which achieves safe exploration during the learning process. Eventually, the principle of Bellman optimality is satisfied through iteratively updating the independently approximated actor and critic to solve the Hamilton-Jacobi-Bellman equation in adaptive dynamic programming. More notably, the variance of control performance under uncertainty is also reduced with the proposed method. The effectiveness of the proposed method is verified with motion control problems for autonomous vehicles through appropriate comparison simulations.
{"title":"Barrier Lyapunov Function-Based Safe Reinforcement Learning Algorithm for Autonomous Vehicles with System Uncertainty","authors":"Yuxiang Zhang, Xiaoling Liang, S. Ge, B. Gao, Tong-heng Lee","doi":"10.23919/ICCAS52745.2021.9649902","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649902","url":null,"abstract":"Guaranteed safety and performance under various circumstances remain technically critical and practically challenging for the wide deployment of autonomous vehicles. For such safety-critical systems, it will certainly be a requirement that safe performance should be ensured even during the reinforcement learning period in the presence of system uncertainty. To address this issue, a Barrier Lyapunov Function-based safe reinforcement learning algorithm (BLF-SRL) is proposed here for the formulated nonlinear system in strict-feedback form. This approach appropriately arranges the Barrier Lyapunov Function item into the optimized backstepping control method to constrain the state-variables in the designed safety region during learning when unknown bounded system uncertainty exists. More specifically, the overall system control is optimized with the optimized backstepping technique under the framework of Actor-Critic, which optimizes the virtual control in every backstepping subsystem. Wherein, the optimal virtual control is decomposed into Barrier Lyapunov Function items; and also with an adaptive item to be learned with deep neural networks, which achieves safe exploration during the learning process. Eventually, the principle of Bellman optimality is satisfied through iteratively updating the independently approximated actor and critic to solve the Hamilton-Jacobi-Bellman equation in adaptive dynamic programming. More notably, the variance of control performance under uncertainty is also reduced with the proposed method. The effectiveness of the proposed method is verified with motion control problems for autonomous vehicles through appropriate comparison simulations.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"39 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":"115208446","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/ICCAS52745.2021.9649965
J. W. Lee, H. J. Yoon, H. S. Woo
This paper presents a prosthetic leg using a two-way hydraulic cylinder. Depending on the walking pattern of people and the walking environment required, it is necessary to adjust the prosthetic leg according to the conditions. The two-way hydraulic cylinder can adjust the tension and compression force separately, and therefore it can be fine-tuned according to the walking conditions. The two-way hydraulic cylinder is actively controlled through the stepping motor so that the human with the developed prosthetic leg can walk similar to a temporarily able-bodied person.
{"title":"Design and Fabrication of a Robotic Knee-Type Prosthetic Leg with a Two-Way Hydraulic Cylinder","authors":"J. W. Lee, H. J. Yoon, H. S. Woo","doi":"10.23919/ICCAS52745.2021.9649965","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649965","url":null,"abstract":"This paper presents a prosthetic leg using a two-way hydraulic cylinder. Depending on the walking pattern of people and the walking environment required, it is necessary to adjust the prosthetic leg according to the conditions. The two-way hydraulic cylinder can adjust the tension and compression force separately, and therefore it can be fine-tuned according to the walking conditions. The two-way hydraulic cylinder is actively controlled through the stepping motor so that the human with the developed prosthetic leg can walk similar to a temporarily able-bodied person.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"30 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":"114235784","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/ICCAS52745.2021.9649844
Hyeon-Woo Na, P. Park
This paper analyzes the stability of sampled-data multi-agent systems with a weighted consensus protocol by the use of looped-functional and free matrix based integral inequality. In the existing stability analysis of the multi-agent system, the typical Lyapunov-functional was used, but a less conservative solution can be obtained by using the looped-functional which is developed for the single-agent system. In addition, when analyzing the stability using Lyapunov-functional, integral inequality is used to obtain the upper bound of the integral term. A larger maximum sampling interval can be obtained by using the free matrix based integral inequality which is developed in time-delay system recently. Therefore, in this paper, the Lyapunov-functional including the looped-functional was constructed, the stability condition was relaxed using the free matrix based integral inequality, and the system was confirmed to be stable at the larger sampling interval compared to the existing literature through experimental examples.
{"title":"Application of free matrix based integral inequality: sampled-data multi-agent system","authors":"Hyeon-Woo Na, P. Park","doi":"10.23919/ICCAS52745.2021.9649844","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649844","url":null,"abstract":"This paper analyzes the stability of sampled-data multi-agent systems with a weighted consensus protocol by the use of looped-functional and free matrix based integral inequality. In the existing stability analysis of the multi-agent system, the typical Lyapunov-functional was used, but a less conservative solution can be obtained by using the looped-functional which is developed for the single-agent system. In addition, when analyzing the stability using Lyapunov-functional, integral inequality is used to obtain the upper bound of the integral term. A larger maximum sampling interval can be obtained by using the free matrix based integral inequality which is developed in time-delay system recently. Therefore, in this paper, the Lyapunov-functional including the looped-functional was constructed, the stability condition was relaxed using the free matrix based integral inequality, and the system was confirmed to be stable at the larger sampling interval compared to the existing literature through experimental examples.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"6 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":"114806198","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/ICCAS52745.2021.9649912
Jasper Z. Tan, A. Dasgupta, Arjun Agrawal, S. Srigrarom
A novel control & software architecture using ROS C++ is introduced for object interception by a UAV with a mounted depth camera and no external aid. Existing work in trajectory prediction focused on the use of off-board tools like motion capture rooms to intercept thrown objects. The present study designs the UAV architecture to be completely on-board capable of object interception with the use of a depth camera and point cloud processing. The architecture uses an iterative trajectory prediction algorithm for non-propelled objects like a ping-pong ball. A variety of path planning approaches to object interception and their corresponding scenarios are discussed, evaluated & simulated in Gazebo. The successful simulations exemplify the potential of using the proposed architecture for the onboard autonomy of UAVs intercepting objects.
{"title":"Trajectory Prediction & Path Planning for an Object Intercepting UAV with a Mounted Depth Camera","authors":"Jasper Z. Tan, A. Dasgupta, Arjun Agrawal, S. Srigrarom","doi":"10.23919/ICCAS52745.2021.9649912","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649912","url":null,"abstract":"A novel control & software architecture using ROS C++ is introduced for object interception by a UAV with a mounted depth camera and no external aid. Existing work in trajectory prediction focused on the use of off-board tools like motion capture rooms to intercept thrown objects. The present study designs the UAV architecture to be completely on-board capable of object interception with the use of a depth camera and point cloud processing. The architecture uses an iterative trajectory prediction algorithm for non-propelled objects like a ping-pong ball. A variety of path planning approaches to object interception and their corresponding scenarios are discussed, evaluated & simulated in Gazebo. The successful simulations exemplify the potential of using the proposed architecture for the onboard autonomy of UAVs intercepting objects.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"3 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":"114557829","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/ICCAS52745.2021.9649968
Wooseok Oh, Hwiyeon Yoo, Timothy Ha, Songhwai Oh
The vision-based 3D reconstruction methods have various advantages and can be used in various applications such as navigation. Although various vision-based methods are being studied, it is difficult to reconstruct many parts at once with a general camera because of a small FOV. To solve this problem, we propose a coarse but lightweight reconstruction method using a camera with a unique structure called a compound eye with various advantages such as large FOV. In the process, we devise a network that performs depth estimation on a compound eye structure to obtain a depth image containing 3D information from an RGB image. We tested our methods by collecting data using a compound eye camera implemented in a Gazebo simulation and simulation scenes we created. As a result, our 3D reconstruction method using the data we collected and the confidence score from our depth estimation result, can capture the environment with a high recall of 97.51 %.
{"title":"Vision-Based 3D Reconstruction Using a Compound Eye Camera","authors":"Wooseok Oh, Hwiyeon Yoo, Timothy Ha, Songhwai Oh","doi":"10.23919/ICCAS52745.2021.9649968","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649968","url":null,"abstract":"The vision-based 3D reconstruction methods have various advantages and can be used in various applications such as navigation. Although various vision-based methods are being studied, it is difficult to reconstruct many parts at once with a general camera because of a small FOV. To solve this problem, we propose a coarse but lightweight reconstruction method using a camera with a unique structure called a compound eye with various advantages such as large FOV. In the process, we devise a network that performs depth estimation on a compound eye structure to obtain a depth image containing 3D information from an RGB image. We tested our methods by collecting data using a compound eye camera implemented in a Gazebo simulation and simulation scenes we created. As a result, our 3D reconstruction method using the data we collected and the confidence score from our depth estimation result, can capture the environment with a high recall of 97.51 %.","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":"116679141","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/ICCAS52745.2021.9649893
Nguyen Duc Toan, Kim Gon-Woo
In recent years, reinforcement learning has attracted researchers' attention with the AlphaGo event. Especially in autonomous mobile robots, the reinforcement learning approach can be applied to the mapless navigation problem. The Robot can complete the set tasks well and works well in different environments without maps and ready-made path plans. However, for reinforcement learning in general and mapless navigation based on reinforcement learning in particular, exploitation and exploration balance are issues that need to be carefully considered. Specifically, the fact that the agent (Robot) can discover and execute actions in a particular working environment plays a significant role in improving the performance of the reinforcement learning problem. By creating some noise during the convolutional neural network training, the above problem can be solved by some popular approaches today. With outstanding advantages compared to other approaches, the Boltzmann policy approach has been used in our problem. It helps the Robot explore more thoroughly in complex environments, and the policy is also more optimized.
{"title":"Environment Exploration for Mapless Navigation based on Deep Reinforcement Learning","authors":"Nguyen Duc Toan, Kim Gon-Woo","doi":"10.23919/ICCAS52745.2021.9649893","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649893","url":null,"abstract":"In recent years, reinforcement learning has attracted researchers' attention with the AlphaGo event. Especially in autonomous mobile robots, the reinforcement learning approach can be applied to the mapless navigation problem. The Robot can complete the set tasks well and works well in different environments without maps and ready-made path plans. However, for reinforcement learning in general and mapless navigation based on reinforcement learning in particular, exploitation and exploration balance are issues that need to be carefully considered. Specifically, the fact that the agent (Robot) can discover and execute actions in a particular working environment plays a significant role in improving the performance of the reinforcement learning problem. By creating some noise during the convolutional neural network training, the above problem can be solved by some popular approaches today. With outstanding advantages compared to other approaches, the Boltzmann policy approach has been used in our problem. It helps the Robot explore more thoroughly in complex environments, and the policy is also more optimized.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"4 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":"117105691","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：