In this study,relevant work on autonomy evaluation(AE)in recent years was comprehensively reviewed and classified from the perspective of task models,and a closed-loop task models based theoretical framework for AE wa...In this study,relevant work on autonomy evaluation(AE)in recent years was comprehensively reviewed and classified from the perspective of task models,and a closed-loop task models based theoretical framework for AE was developed.The main contributions of this study are as follows.1)A taxonomy for AE based on task models was introduced to classify current theories,methods and standards.2)The limitations of the current autonomous evaluation methods were addressed to provide a theoretical framework for quantitative evaluation based on task models,and evaluation metrics for each stage were proposed based on the AE theoretical framework.3)Qualitative analyses of the superiority of the proposed AE framework based on the closed-loop task models were conducted.This study attempts to provide a reference for researchers and engineers in the autonomous unmanned systems field and inspire future development of AE.展开更多
With the development of the Internet of Things and devices continuing to scale,using cloud computing resources to process data in real-time is challenging.Edge computing technologies can improve real-time performance ...With the development of the Internet of Things and devices continuing to scale,using cloud computing resources to process data in real-time is challenging.Edge computing technologies can improve real-time performance in processing data.By introducing the FPGA into the computing node and using the dynamic reconfigurability of the FPGA,the FPGA-based edge node can increase the edge node capability.In this paper,a task-based collaborative method for an FPGA-based edge computing system is proposed in order to meet the collaboration among FPGA-based edge nodes,edge nodes,and the cloud.The modeling of the task includes two parts,task information and task-dependent file.Task information is used to describe the running information and dependency infor-mation required for the task execution.Task-dependent file contains the configuration bit-stream of FPGA in running of the task.By analyzing the task behavior,this paper builds four basic behaviors,analyzes the critical attributes of each behavior,and summa-rizes the task model suitable for FPGA-based edge nodes.Tasks with specific functions can be created by modifying different attributes of model nodes.Finally,the availability of the model and the task-based collaborative method are verified by simulation exper-iments.The experimental results that the task model proposed in this paper can meet cloud-edge collaboration in the FPGA-based edge computing environment.展开更多
It is important to improve the development efficiency of decoupling a coupling task package according to the information relevancy relation between development tasks in the collaborative development process of complic...It is important to improve the development efficiency of decoupling a coupling task package according to the information relevancy relation between development tasks in the collaborative development process of complicated electronic products.In order to define the task coupling model in the development process,the weighted directed graph based on the information relevancy is established,and the correspondence between weighted directed graph model and numerical design structure matrix model of coupling tasks is introduced.The task coupling model is quantized,thereby the interactivity matrix of task package is built.A multi-goal task decoupling method based on improved genetic algorithm is proposed to decouple the task coupling model,which transforms the decoupling of task package into a multi-goal optimization issue.Then the improved genetic algorithm is used to solve the interactivity matrix of coupling tasks.Finally,the effectiveness of this decomposition method is proved by using the example of task package decoupling of collaborative development of a radar’s phased array antenna.展开更多
Based on the modeling of robot working environment, the shortest distance matrix between points is solved by Floyd algorithm. With the objective of minimizing the sum of the fixed cost of robot and the cost of robot o...Based on the modeling of robot working environment, the shortest distance matrix between points is solved by Floyd algorithm. With the objective of minimizing the sum of the fixed cost of robot and the cost of robot operation, an integer programming model is established and a genetic algorithm for solving the model is designed. In order to make coordination to accomplish their respective tasks for each robot with high efficiency, this paper uses natural number encoding way. The objective function is based on penalty term constructed with the total number of collisions in the running path of robots. The fitness function is constructed by using the objective function with penalty term. Based on elitist retention strategy, a genetic algorithm with collision detection is designed. Using this algorithm for task allocation and path planning of multi-robot, it can effectively avoid or reduce the number of collisions in the process of multi-robot performing tasks. Finally, an example is used to validate the method.展开更多
A theoretical approach of ordered emergency tasks generation is proposed for dealing with a specific emergency event rapidly, exactly and effectively. According to the general principles of an emergency plan developed...A theoretical approach of ordered emergency tasks generation is proposed for dealing with a specific emergency event rapidly, exactly and effectively. According to the general principles of an emergency plan developed to response to an emergency management, a workflow model is employed to complete the formal modeling of concrete emergency plan firstly. Then the HTN planning system SHOP2 is introduced, the transformation method of domain knowledge from emergency domain to SHOP2 domain is studied. At last, the general procedure to solve the emergency decision prob-lems and to generate executive emergency tasks is set up drawing support from SHOP2 planning system, which will combine the principles (or knowledge) of emergency plan and the real emergency situations.展开更多
In the smart warehousing system adopting cargo-to-person mode, all the items are stored in the movable shelves. There are some warehouse robots transporting the shelves to the working platforms for completing order pi...In the smart warehousing system adopting cargo-to-person mode, all the items are stored in the movable shelves. There are some warehouse robots transporting the shelves to the working platforms for completing order picking or items replenishment tasks. When the number of robots is insufficient, the task allocation problem of robots is an important issue in designing the warehousing system. In this paper, the task allocation problem of insufficient warehouse robots (TAPIR) is investigated. Firstly, the TAPIR problem is decomposed into three sub-problems: task grouping problem, task scheduling problem and task balanced allocation problem. Then three sub-problems are respectively formulated into integer programming models, and the corresponding heuristic algorithms for solving three sub-problems are designed. Finally, the simulation and analysis are done on the real data of online bookstore. Simulation results show that the mathematical models and algorithms of this paper can provide a theoretical basis for solving the TAPIR problem.展开更多
在短文本分类这一热门研究领域,预训练模型虽表现出色,但存在数据表征散布和转移学习受限等问题,导致分类效果降低。尽管诸如ERNIE(enhanced representation through knowledge integration)等大模型提升了分类效果,但由于其速度慢、资...在短文本分类这一热门研究领域,预训练模型虽表现出色,但存在数据表征散布和转移学习受限等问题,导致分类效果降低。尽管诸如ERNIE(enhanced representation through knowledge integration)等大模型提升了分类效果,但由于其速度慢、资源需求大、受硬件限制等问题,难以应用于工业生产。为此提出了基于ERNIE的辅助任务模型EL(ERNIE+LAT(learning with auxiliary tasks)),旨在验证其获取更优类别特征的可行性。针对工业任务落地需求,同时提出知识蒸馏法,对EL与TextCNN模型进行蒸馏训练,以验证其在提升分类效果以及线上推理效率方面的可行性。研究验证现有预训练模型在公共数据集上的分类优势;采用辅助任务结合预训练模型的方法进行实验改进;并基于知识蒸馏法深入推进模型加速研究。实验表明,通过联合训练能够提升模型的泛化能力与特征提取能力,进而增强特定任务下的学习能力;学生模型TextCNN经蒸馏后不仅可与教师模型相媲美,在线上部署时还更具优势。展开更多
Smart cities have different contradicting goals having no apparent solution.The selection of the appropriate solution,which is considered the best compromise among the candidates,is known as complex problem-solving.Sm...Smart cities have different contradicting goals having no apparent solution.The selection of the appropriate solution,which is considered the best compromise among the candidates,is known as complex problem-solving.Smart city administrators face different problems of complex nature,such as optimal energy trading in microgrids and optimal comfort index in smart homes,to mention a few.This paper proposes a novel architecture to offer complex problem solutions as a service(CPSaaS)based on predictive model optimization and optimal task orchestration to offer solutions to different problems in a smart city.Predictive model optimization uses a machine learning module and optimization objective to compute the given problem’s solutions.The task orchestration module helps decompose the complex problem in small tasks and deploy them on real-world physical sensors and actuators.The proposed architecture is hierarchical and modular,making it robust against faults and easy to maintain.The proposed architecture’s evaluation results highlight its strengths in fault tolerance,accuracy,and processing speed.展开更多
在航空航天技术领域,代理模型发挥着关键作用。为平衡代理模型的训练成本与预测精度,需要发展能够有效挖掘多保真度数据间潜在相关性的建模方法。针对现有模型依赖预设核函数、缺乏数据自适应性的问题,提出了一种基于稀疏混合专家神经核...在航空航天技术领域,代理模型发挥着关键作用。为平衡代理模型的训练成本与预测精度,需要发展能够有效挖掘多保真度数据间潜在相关性的建模方法。针对现有模型依赖预设核函数、缺乏数据自适应性的问题,提出了一种基于稀疏混合专家神经核(mixture of experts neural kernel,MoENK)的多保真度代理模型。MoENK通过线性混合和乘积混合基本单元构造新核函数,选择性屏蔽中间结果以过滤噪声,并应用于多任务高斯过程中。将该方法应用于3个函数示例和2个翼型算例中,结果表明该方法的预测精度有较大提升,尤其在NACA0012翼型阻力系数的预测中,相较于次佳方法LR-MFS,RMSE和MAE分别降低了40.42%和44.70%。证实了所提出的MoENK核函数能够不依赖预设核函数进行自适应预测,具有良好的泛化能力和鲁棒性,为工程系统的代理模型构建提供了新的工具。展开更多
基金supported in part by the Major Project for New Generation of AI,China(No.2018AAA0100400)in part by the Natural Science Foundation of China(No.62072457).
文摘In this study,relevant work on autonomy evaluation(AE)in recent years was comprehensively reviewed and classified from the perspective of task models,and a closed-loop task models based theoretical framework for AE was developed.The main contributions of this study are as follows.1)A taxonomy for AE based on task models was introduced to classify current theories,methods and standards.2)The limitations of the current autonomous evaluation methods were addressed to provide a theoretical framework for quantitative evaluation based on task models,and evaluation metrics for each stage were proposed based on the AE theoretical framework.3)Qualitative analyses of the superiority of the proposed AE framework based on the closed-loop task models were conducted.This study attempts to provide a reference for researchers and engineers in the autonomous unmanned systems field and inspire future development of AE.
基金This work is supported by the National Key R&D Program of China(Grant No.2018YFB1701600).
文摘With the development of the Internet of Things and devices continuing to scale,using cloud computing resources to process data in real-time is challenging.Edge computing technologies can improve real-time performance in processing data.By introducing the FPGA into the computing node and using the dynamic reconfigurability of the FPGA,the FPGA-based edge node can increase the edge node capability.In this paper,a task-based collaborative method for an FPGA-based edge computing system is proposed in order to meet the collaboration among FPGA-based edge nodes,edge nodes,and the cloud.The modeling of the task includes two parts,task information and task-dependent file.Task information is used to describe the running information and dependency infor-mation required for the task execution.Task-dependent file contains the configuration bit-stream of FPGA in running of the task.By analyzing the task behavior,this paper builds four basic behaviors,analyzes the critical attributes of each behavior,and summa-rizes the task model suitable for FPGA-based edge nodes.Tasks with specific functions can be created by modifying different attributes of model nodes.Finally,the availability of the model and the task-based collaborative method are verified by simulation exper-iments.The experimental results that the task model proposed in this paper can meet cloud-edge collaboration in the FPGA-based edge computing environment.
基金supported by the National Defense Basic Research Program of China (No. A1120131044)
文摘It is important to improve the development efficiency of decoupling a coupling task package according to the information relevancy relation between development tasks in the collaborative development process of complicated electronic products.In order to define the task coupling model in the development process,the weighted directed graph based on the information relevancy is established,and the correspondence between weighted directed graph model and numerical design structure matrix model of coupling tasks is introduced.The task coupling model is quantized,thereby the interactivity matrix of task package is built.A multi-goal task decoupling method based on improved genetic algorithm is proposed to decouple the task coupling model,which transforms the decoupling of task package into a multi-goal optimization issue.Then the improved genetic algorithm is used to solve the interactivity matrix of coupling tasks.Finally,the effectiveness of this decomposition method is proved by using the example of task package decoupling of collaborative development of a radar’s phased array antenna.
文摘Based on the modeling of robot working environment, the shortest distance matrix between points is solved by Floyd algorithm. With the objective of minimizing the sum of the fixed cost of robot and the cost of robot operation, an integer programming model is established and a genetic algorithm for solving the model is designed. In order to make coordination to accomplish their respective tasks for each robot with high efficiency, this paper uses natural number encoding way. The objective function is based on penalty term constructed with the total number of collisions in the running path of robots. The fitness function is constructed by using the objective function with penalty term. Based on elitist retention strategy, a genetic algorithm with collision detection is designed. Using this algorithm for task allocation and path planning of multi-robot, it can effectively avoid or reduce the number of collisions in the process of multi-robot performing tasks. Finally, an example is used to validate the method.
文摘A theoretical approach of ordered emergency tasks generation is proposed for dealing with a specific emergency event rapidly, exactly and effectively. According to the general principles of an emergency plan developed to response to an emergency management, a workflow model is employed to complete the formal modeling of concrete emergency plan firstly. Then the HTN planning system SHOP2 is introduced, the transformation method of domain knowledge from emergency domain to SHOP2 domain is studied. At last, the general procedure to solve the emergency decision prob-lems and to generate executive emergency tasks is set up drawing support from SHOP2 planning system, which will combine the principles (or knowledge) of emergency plan and the real emergency situations.
文摘In the smart warehousing system adopting cargo-to-person mode, all the items are stored in the movable shelves. There are some warehouse robots transporting the shelves to the working platforms for completing order picking or items replenishment tasks. When the number of robots is insufficient, the task allocation problem of robots is an important issue in designing the warehousing system. In this paper, the task allocation problem of insufficient warehouse robots (TAPIR) is investigated. Firstly, the TAPIR problem is decomposed into three sub-problems: task grouping problem, task scheduling problem and task balanced allocation problem. Then three sub-problems are respectively formulated into integer programming models, and the corresponding heuristic algorithms for solving three sub-problems are designed. Finally, the simulation and analysis are done on the real data of online bookstore. Simulation results show that the mathematical models and algorithms of this paper can provide a theoretical basis for solving the TAPIR problem.
文摘在短文本分类这一热门研究领域,预训练模型虽表现出色,但存在数据表征散布和转移学习受限等问题,导致分类效果降低。尽管诸如ERNIE(enhanced representation through knowledge integration)等大模型提升了分类效果,但由于其速度慢、资源需求大、受硬件限制等问题,难以应用于工业生产。为此提出了基于ERNIE的辅助任务模型EL(ERNIE+LAT(learning with auxiliary tasks)),旨在验证其获取更优类别特征的可行性。针对工业任务落地需求,同时提出知识蒸馏法,对EL与TextCNN模型进行蒸馏训练,以验证其在提升分类效果以及线上推理效率方面的可行性。研究验证现有预训练模型在公共数据集上的分类优势;采用辅助任务结合预训练模型的方法进行实验改进;并基于知识蒸馏法深入推进模型加速研究。实验表明,通过联合训练能够提升模型的泛化能力与特征提取能力,进而增强特定任务下的学习能力;学生模型TextCNN经蒸馏后不仅可与教师模型相媲美,在线上部署时还更具优势。
基金This research was supported by Energy Cloud R&D Program through the National Research Foundation of Korea(NRF)funded by the Ministry of Science,ICT(2019M3F2A1073387)this research was supported by Basic Science Research Program through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(2018R1D1A1A09082919)this research was supported by Institute for Information&communications Technology Planning&Evaluation(IITP)grant funded by the Korea government(MSIT)(No.2018-0-01456,AutoMaTa:Autonomous Management framework based on artificial intelligent Technology for adaptive and disposable IoT).Any correspondence related to this paper should be addressed to Dohyeun Kim.
文摘Smart cities have different contradicting goals having no apparent solution.The selection of the appropriate solution,which is considered the best compromise among the candidates,is known as complex problem-solving.Smart city administrators face different problems of complex nature,such as optimal energy trading in microgrids and optimal comfort index in smart homes,to mention a few.This paper proposes a novel architecture to offer complex problem solutions as a service(CPSaaS)based on predictive model optimization and optimal task orchestration to offer solutions to different problems in a smart city.Predictive model optimization uses a machine learning module and optimization objective to compute the given problem’s solutions.The task orchestration module helps decompose the complex problem in small tasks and deploy them on real-world physical sensors and actuators.The proposed architecture is hierarchical and modular,making it robust against faults and easy to maintain.The proposed architecture’s evaluation results highlight its strengths in fault tolerance,accuracy,and processing speed.
文摘在航空航天技术领域,代理模型发挥着关键作用。为平衡代理模型的训练成本与预测精度,需要发展能够有效挖掘多保真度数据间潜在相关性的建模方法。针对现有模型依赖预设核函数、缺乏数据自适应性的问题,提出了一种基于稀疏混合专家神经核(mixture of experts neural kernel,MoENK)的多保真度代理模型。MoENK通过线性混合和乘积混合基本单元构造新核函数,选择性屏蔽中间结果以过滤噪声,并应用于多任务高斯过程中。将该方法应用于3个函数示例和2个翼型算例中,结果表明该方法的预测精度有较大提升,尤其在NACA0012翼型阻力系数的预测中,相较于次佳方法LR-MFS,RMSE和MAE分别降低了40.42%和44.70%。证实了所提出的MoENK核函数能够不依赖预设核函数进行自适应预测,具有良好的泛化能力和鲁棒性,为工程系统的代理模型构建提供了新的工具。
