Pub Date : 2021-03-07DOI: 10.1109/ICM46511.2021.9385698
Yuichi Kizu, T. Kai, H. Ikeda
This paper proposes a method for suppressing the vibration of machine bases by using a counter mass driving (CMD) system for fast and accurate positioning. In horizontal positioning systems, servo motors are widely used to drive a moving mass, but its reaction force causes the machine base on which the motors are mounted to vibrate. The CMD system, which drives an additional mass for vibration suppression, is well-known for its effectiveness. However, for practical uses, it is hard to design the performances of the system, e.g., positioning time, max torque, and range of motion. To suppress vibration and satisfy user requirements for performance, we devise a method for designing a vibration control filter analytically on the basis of a command shaping strategy. In experiments, we confirmed that the proposed method can suppress vibration and make the positioning time shorter by tuning the design parameters of the filter.
{"title":"Horizontal Counter Control Method for Suppressing Vibration of Machine Base","authors":"Yuichi Kizu, T. Kai, H. Ikeda","doi":"10.1109/ICM46511.2021.9385698","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385698","url":null,"abstract":"This paper proposes a method for suppressing the vibration of machine bases by using a counter mass driving (CMD) system for fast and accurate positioning. In horizontal positioning systems, servo motors are widely used to drive a moving mass, but its reaction force causes the machine base on which the motors are mounted to vibrate. The CMD system, which drives an additional mass for vibration suppression, is well-known for its effectiveness. However, for practical uses, it is hard to design the performances of the system, e.g., positioning time, max torque, and range of motion. To suppress vibration and satisfy user requirements for performance, we devise a method for designing a vibration control filter analytically on the basis of a command shaping strategy. In experiments, we confirmed that the proposed method can suppress vibration and make the positioning time shorter by tuning the design parameters of the filter.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"40 5","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133391772","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385662
Koya Hashizume, Kikuko Miyata, S. Hara
Flyby imaging has attracted attention as a method for small body exploration and the requirement for an accurate target tracking system becomes higher for the mission quality. A relative trajectory that is either designed or estimated offline contains uncertainties, and they cause errors in the targeting profile generated for imaging missions. Online relative trajectory parameter estimation is required to obtain high-quality imaging, and visual-based tracking systems are popular for estimating the relative trajectory. In addition to the uncertainties in the measured or estimated relative parameters of trajectories, the high relative velocity between the spacecraft and the asteroid, as well as the actuator driving characteristics have a negative impact on asteroid tracking. To overcome the problems, this paper proposes a control system with two degrees of freedom, which includes a vision-based feedback controller and feedforward controller. The feedforward controller contains two compensators: the first utilizes a real-time vision-based relative trajectory estimator and predicts the future relative trajectory from the results of the estimation, and the other compensates for the actuator characteristics. The applicability of the proposed control system is discussed by using a case study based on numerical simulation. The results show the effectiveness of the proposed concept.
{"title":"Vision-Based Rapid Target Tracking Method for Trajectories Estimation and Actuator Parameter Uncertainties for Asteroid Flyby Problem","authors":"Koya Hashizume, Kikuko Miyata, S. Hara","doi":"10.1109/ICM46511.2021.9385662","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385662","url":null,"abstract":"Flyby imaging has attracted attention as a method for small body exploration and the requirement for an accurate target tracking system becomes higher for the mission quality. A relative trajectory that is either designed or estimated offline contains uncertainties, and they cause errors in the targeting profile generated for imaging missions. Online relative trajectory parameter estimation is required to obtain high-quality imaging, and visual-based tracking systems are popular for estimating the relative trajectory. In addition to the uncertainties in the measured or estimated relative parameters of trajectories, the high relative velocity between the spacecraft and the asteroid, as well as the actuator driving characteristics have a negative impact on asteroid tracking. To overcome the problems, this paper proposes a control system with two degrees of freedom, which includes a vision-based feedback controller and feedforward controller. The feedforward controller contains two compensators: the first utilizes a real-time vision-based relative trajectory estimator and predicts the future relative trajectory from the results of the estimation, and the other compensates for the actuator characteristics. The applicability of the proposed control system is discussed by using a case study based on numerical simulation. The results show the effectiveness of the proposed concept.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114008917","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385693
Mohammad Reza Chalak Qazani, Houshyar Asadi, Mohammed Al-Ashmori, Shady M. K. Mohamed, C. Lim, S. Nahavandi
Motion signals can be reproduced using a simulation-based motion platform (SBMP) and virtual reality. In this respect, time series prediction of the driving motion scenarios can enhance the quality of the regenerated motion signals with the motion cueing algorithm (MCA). Specifically, the MCA is employed to regenerate the motion signals for a SBMP with respect to the workspace limitations. The use of the feedforward neural network (NN) produces inaccurate predictions pertaining to the driving motion scenarios. In this paper, an interval type-2 fuzzy neural network (FNN) is proposed to predict the driving motion scenarios. As type-1 FNN is not able to represent the uncertainty in information, a Type-2 Quantum (T2Q) FNN is used to handle the undefined indexes with consideration of uncertain jump positions. The T2QFNN model can identify the overlaps between classes and adjust the fuzzy parameters automatically, including fuzzy rules as a linear combination of the exogenous input variables. The simulation results indicate that T2QFNN is able to yield lower prediction error and shorter learning time as compared with those from the feedforward NN and type-1 FNN models.
{"title":"Time Series Prediction of Driving Motion Scenarios Using Fuzzy Neural Networks: * Motion Signal Prediction Using FNNs","authors":"Mohammad Reza Chalak Qazani, Houshyar Asadi, Mohammed Al-Ashmori, Shady M. K. Mohamed, C. Lim, S. Nahavandi","doi":"10.1109/ICM46511.2021.9385693","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385693","url":null,"abstract":"Motion signals can be reproduced using a simulation-based motion platform (SBMP) and virtual reality. In this respect, time series prediction of the driving motion scenarios can enhance the quality of the regenerated motion signals with the motion cueing algorithm (MCA). Specifically, the MCA is employed to regenerate the motion signals for a SBMP with respect to the workspace limitations. The use of the feedforward neural network (NN) produces inaccurate predictions pertaining to the driving motion scenarios. In this paper, an interval type-2 fuzzy neural network (FNN) is proposed to predict the driving motion scenarios. As type-1 FNN is not able to represent the uncertainty in information, a Type-2 Quantum (T2Q) FNN is used to handle the undefined indexes with consideration of uncertain jump positions. The T2QFNN model can identify the overlaps between classes and adjust the fuzzy parameters automatically, including fuzzy rules as a linear combination of the exogenous input variables. The simulation results indicate that T2QFNN is able to yield lower prediction error and shorter learning time as compared with those from the feedforward NN and type-1 FNN models.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116487118","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385612
G. Vinco, P. Braun, L. Zaccarian
Ahstract-We illustrate the design and development of a modular hardware/software system for multiple unicycle-like mobile robots localized via a set of camera modules. We describe the architecture and calibration of the hardware/software setup and then discuss two continuous-discrete observation laws for the distributed estimation of the robot positions. We show that a suitable model exploiting the onboard IMU measurements of the robots, enables obtaining an estimation error that is a cascade of two linear systems, for which we can show global exponential convergence to zero. The results are illustrated by our experimental tests.
{"title":"A modular architecture for mobile robots equipped with continuous-discrete observers","authors":"G. Vinco, P. Braun, L. Zaccarian","doi":"10.1109/ICM46511.2021.9385612","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385612","url":null,"abstract":"Ahstract-We illustrate the design and development of a modular hardware/software system for multiple unicycle-like mobile robots localized via a set of camera modules. We describe the architecture and calibration of the hardware/software setup and then discuss two continuous-discrete observation laws for the distributed estimation of the robot positions. We show that a suitable model exploiting the onboard IMU measurements of the robots, enables obtaining an estimation error that is a cascade of two linear systems, for which we can show global exponential convergence to zero. The results are illustrated by our experimental tests.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116819189","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385672
Mingxing Yuan, Litong Lyu, Xin Liu
Achieving high control performance in the presence of unavoidable disturbances/uncertainties has always been a major objective in modern control design. These disturbances/uncertainties may cause the controlled system, designed on the nominal model, to be unstable or have a much degraded performance. To deal with the disturbances/uncertainties, disturbance-observer-base-controls construct observers to estimate the disturbance and then compensate their influences. Among them, active disturbance rejection control (ADRC) takes all disturbances and uncertainties as total disturbances and then constructs an extended state observer to estimate the these disturbances. The simplicity and independence on precise model make ADRC one of the popular control algorithms in recent years. From another prospect, the mathematically rigorous adaptive robust control (ARC) is proposed, where the model uncertainties are classified into parametric uncertainties and uncertain nonlinearities and handled by integrating the working mechanisms of two of the main control research areas - robust adaptive control and nonlinear robust controls. The ARC theory has also been applied to the design of various intelligent and precision industrial mechatronic systems. In this paper, the disturbance rejection performance of ARC is studied with the comparison with ADRC. First-order nonlinear system with disturbance/uncertainties is studied as an example and simulations are conducted to support the comparative results.
{"title":"Disturbance Rejection Performance of Adaptive Robust Control","authors":"Mingxing Yuan, Litong Lyu, Xin Liu","doi":"10.1109/ICM46511.2021.9385672","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385672","url":null,"abstract":"Achieving high control performance in the presence of unavoidable disturbances/uncertainties has always been a major objective in modern control design. These disturbances/uncertainties may cause the controlled system, designed on the nominal model, to be unstable or have a much degraded performance. To deal with the disturbances/uncertainties, disturbance-observer-base-controls construct observers to estimate the disturbance and then compensate their influences. Among them, active disturbance rejection control (ADRC) takes all disturbances and uncertainties as total disturbances and then constructs an extended state observer to estimate the these disturbances. The simplicity and independence on precise model make ADRC one of the popular control algorithms in recent years. From another prospect, the mathematically rigorous adaptive robust control (ARC) is proposed, where the model uncertainties are classified into parametric uncertainties and uncertain nonlinearities and handled by integrating the working mechanisms of two of the main control research areas - robust adaptive control and nonlinear robust controls. The ARC theory has also been applied to the design of various intelligent and precision industrial mechatronic systems. In this paper, the disturbance rejection performance of ARC is studied with the comparison with ADRC. First-order nonlinear system with disturbance/uncertainties is studied as an example and simulations are conducted to support the comparative results.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125449862","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385674
E. Sariyildiz
This paper analyses and synthesises the Disturbance Observer (DOb) based motion control systems in the discrete-time domain. By employing Bode Integral Theorem, it is shown that continuous-time analysis methods fall-short in explaining the dynamic behaviours of the DOb-based robust motion controllers implemented by computers and microcontrollers. For example, continuous-time analysis methods cannot explain why the robust stability and performance of the digital motion controller deteriorate as the bandwidth of the DOb increases. Therefore, unexpected dynamic responses (e.g., poor stability and performance, and high-sensitivity to disturbances and noise) may be observed when the parameters of the digital robust motion controller are tuned by using continuous-time synthesis methods in practice. This paper also analytically derives the robust stability and performance constraints of the DOb-based motion controllers in the discrete-time domain. The proposed design constraints allow one to systematically synthesise a high-performance digital robust motion controller. The validity of the proposed analysis and synthesis methods are verified by simulations.
{"title":"A Guide to Design Disturbance Observer-based Motion Control Systems in Discrete-time Domain","authors":"E. Sariyildiz","doi":"10.1109/ICM46511.2021.9385674","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385674","url":null,"abstract":"This paper analyses and synthesises the Disturbance Observer (DOb) based motion control systems in the discrete-time domain. By employing Bode Integral Theorem, it is shown that continuous-time analysis methods fall-short in explaining the dynamic behaviours of the DOb-based robust motion controllers implemented by computers and microcontrollers. For example, continuous-time analysis methods cannot explain why the robust stability and performance of the digital motion controller deteriorate as the bandwidth of the DOb increases. Therefore, unexpected dynamic responses (e.g., poor stability and performance, and high-sensitivity to disturbances and noise) may be observed when the parameters of the digital robust motion controller are tuned by using continuous-time synthesis methods in practice. This paper also analytically derives the robust stability and performance constraints of the DOb-based motion controllers in the discrete-time domain. The proposed design constraints allow one to systematically synthesise a high-performance digital robust motion controller. The validity of the proposed analysis and synthesis methods are verified by simulations.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130201045","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}
Automated Driving for open-pit mines has become a trend in recent years, and one of the most challenging scenario is parking path planning based on the demand of the excavator. In path planning field, dynamic window approach (DWA) is widely used. However, such method negalects expected direction at the destination, which makes it difficult to be implemented for mining trucks parking. To deal with this problem, we proposed a fuzzy logic-based adaptive DWA by considering truck kinematics. Firstly, an improved DWA considering obstacle avoidance is designing by taking Ackermann steering constraint into account. In the approach, we design a novel objective function to regulate the truck direction. Secondly, the detail of fuzzy logic is presented to guarantee the robustness against different situations. Simulations performed with the fuzzy logic-based adaptive DWA show that the mining truck can drive to the destination in high direction accuracy in different scenarios.
{"title":"A Fuzzy Logic-Based Adaptive Dynamic Window Approach for Path Planning of Automated Driving Mining Truck","authors":"Yu Lei, Yafei Wang, Shaoteng Wu, X. Gu, Xiaojun Qin","doi":"10.1109/ICM46511.2021.9385634","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385634","url":null,"abstract":"Automated Driving for open-pit mines has become a trend in recent years, and one of the most challenging scenario is parking path planning based on the demand of the excavator. In path planning field, dynamic window approach (DWA) is widely used. However, such method negalects expected direction at the destination, which makes it difficult to be implemented for mining trucks parking. To deal with this problem, we proposed a fuzzy logic-based adaptive DWA by considering truck kinematics. Firstly, an improved DWA considering obstacle avoidance is designing by taking Ackermann steering constraint into account. In the approach, we design a novel objective function to regulate the truck direction. Secondly, the detail of fuzzy logic is presented to guarantee the robustness against different situations. Simulations performed with the fuzzy logic-based adaptive DWA show that the mining truck can drive to the destination in high direction accuracy in different scenarios.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127342365","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385696
R. Y. Hou
We observed that predictable driving behaviors can lead to a reduction of reaction time and safety improvement for auto-driving. In this study, we investigate the different safety alerts related to auto-driving. We design a pre-warning scheme to improve the predictability of driving behaviors for each autonomous vehicle. Based on the pre-warning scheme, we propose an algorithm to predict the safety status for each peer vehicle and extend the concept to a cluster of autonomous vehicles by proposing another algorithm to avoid risk accumulation during driving. We evaluate the effectiveness of proposed algorithms by using various simulations and case illustrations. The simulations showed that the results are promising.
{"title":"Algorithm to Improve the Predictability for Auto-vehicles' Behaviors and Avoid Risk Accumulation during Driving","authors":"R. Y. Hou","doi":"10.1109/ICM46511.2021.9385696","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385696","url":null,"abstract":"We observed that predictable driving behaviors can lead to a reduction of reaction time and safety improvement for auto-driving. In this study, we investigate the different safety alerts related to auto-driving. We design a pre-warning scheme to improve the predictability of driving behaviors for each autonomous vehicle. Based on the pre-warning scheme, we propose an algorithm to predict the safety status for each peer vehicle and extend the concept to a cluster of autonomous vehicles by proposing another algorithm to avoid risk accumulation during driving. We evaluate the effectiveness of proposed algorithms by using various simulations and case illustrations. The simulations showed that the results are promising.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"127 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129822827","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385601
Aurélien Quelin, L. Petit, C. Prelle, Nicolas Damay
Embedding a wireless power source in a microrobot rises several constraints that can make the system non-functional: limited power, limited energy, or varying voltage or current. To avoid this situation, the robot should be able to work efficiently under varying voltage/current. We then propose a methodology that helps predicting if a given microrobot can meet these requirements. This methodology is applied on a microrobot that integrates a single two-dimensional digital electromagnetic actuator used for an impact-drive locomotion. We designed a prototype of the robot, qualitatively demonstrated its two-dimensional displacement and characterized it in one dimension. This characterization is used to demonstrate that the performance and the operating range of the microrobot are sufficiently adaptable to consider the integration of a wireless power supply.
{"title":"Experimental performance analysis of an electromagnetic impact-drive microrobot","authors":"Aurélien Quelin, L. Petit, C. Prelle, Nicolas Damay","doi":"10.1109/ICM46511.2021.9385601","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385601","url":null,"abstract":"Embedding a wireless power source in a microrobot rises several constraints that can make the system non-functional: limited power, limited energy, or varying voltage or current. To avoid this situation, the robot should be able to work efficiently under varying voltage/current. We then propose a methodology that helps predicting if a given microrobot can meet these requirements. This methodology is applied on a microrobot that integrates a single two-dimensional digital electromagnetic actuator used for an impact-drive locomotion. We designed a prototype of the robot, qualitatively demonstrated its two-dimensional displacement and characterized it in one dimension. This characterization is used to demonstrate that the performance and the operating range of the microrobot are sufficiently adaptable to consider the integration of a wireless power supply.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128900449","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 : 2021-03-07DOI: 10.1109/ICM46511.2021.9385686
Yasuhiro Osaka, Naoya Odajima, Y. Uchimura
Since the publication showed DQN based reinforcement learning methods exceeds human's score in Atari 2600 video games, various deep reinforcement learning have bee researched. This paper proposes a method to control bulldozer autonomously by learning the sediment leveling route using PPO that enables distributed deep reinforcement learning. The simulator was originally developed that enables to reproduce the behavior of small and uniform sediment. By incorporating an LSTM that processes the input state as time-series data into the agent network, more than 95% of the sediment in the target area on average was achieved. In addition, the generalization performance for unknown condition was evaluated, by giving unlearned conditions were given as initial setups.
{"title":"Route optimization for autonomous bulldozer by distributed deep reinforcement learning","authors":"Yasuhiro Osaka, Naoya Odajima, Y. Uchimura","doi":"10.1109/ICM46511.2021.9385686","DOIUrl":"https://doi.org/10.1109/ICM46511.2021.9385686","url":null,"abstract":"Since the publication showed DQN based reinforcement learning methods exceeds human's score in Atari 2600 video games, various deep reinforcement learning have bee researched. This paper proposes a method to control bulldozer autonomously by learning the sediment leveling route using PPO that enables distributed deep reinforcement learning. The simulator was originally developed that enables to reproduce the behavior of small and uniform sediment. By incorporating an LSTM that processes the input state as time-series data into the agent network, more than 95% of the sediment in the target area on average was achieved. In addition, the generalization performance for unknown condition was evaluated, by giving unlearned conditions were given as initial setups.","PeriodicalId":373423,"journal":{"name":"2021 IEEE International Conference on Mechatronics (ICM)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122931587","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：