As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge device...As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge devices.The execution speed of the deployed model is a key element to ensure service quality.Considering a highly heterogeneous edge deployment scenario,deep learning compiling is a novel approach that aims to solve this problem.It defines models using certain DSLs and generates efficient code implementations on different hardware devices.However,there are still two aspects that are not yet thoroughly investigated yet.The first is the optimization of memory-intensive operations,and the second problem is the heterogeneity of the deployment target.To that end,in this work,we propose a system solution that optimizes memory-intensive operation,optimizes the subgraph distribution,and enables the compiling and deployment of DNN models on multiple targets.The evaluation results show the performance of our proposed system.展开更多
The tether deployment of a tethered satellite system involves the consideration of complex dynamic properties of the tether,such as large deformation,slack,and even rebound,and therefore,the dynamic modelling of the t...The tether deployment of a tethered satellite system involves the consideration of complex dynamic properties of the tether,such as large deformation,slack,and even rebound,and therefore,the dynamic modelling of the tether is necessary for performing a dynamic analysis of the system.For a variablelength tether element,the absolute nodal coordinate formulation(ANCF)in the framework of the arbitrary Lagrange-Euler(ALE)description was used to develop a precise dynamic model of a tethered satellite.The model considered the gravitational gradient force and Coriolis force in the orbital coordinate frame,and it was validated through numerical simulation.In the presence of dynamic constraints,a deployment velocity of the tether was obtained by an optimal procedure.In the simulation,rebound behavior of the tethered satellite system was observed when the ANCF-ALE model was employed.Notably,the rebound behavior cannot be predicted by the traditional dumbbell model.Furthermore,an improved optimal deployment velocity was developed.Simulation results indicated that the rebound phenomenon was eliminated,and smooth deployment as well as a stable state of the station-keeping process were achieved.Additionally,the swing amplitude in the station-keeping phase decreased when a deployment strategy based on the improved optimal deployment velocity was used.展开更多
The dynamic equivalent continuum modeling method of the mast which is based on energy equivalency principle was investigated. And three kinds of mast dynamic model were established, which were equivalent continuum mod...The dynamic equivalent continuum modeling method of the mast which is based on energy equivalency principle was investigated. And three kinds of mast dynamic model were established, which were equivalent continuum model, finite element model and simulation model, respectively. The mast frequencies and mode shapes were calculated by these models and compared with each other. The error between the equivalent continuum model and the finite element model is less than 5% when the mast length is longer. Dynamic responses of the mast with different lengths are tested, the mode frequencies and mode shapes are compared with finite element model. The mode shapes match well with each other, while the frequencies tested by experiments are lower than the results of the finite element model, which reflects the joints lower the mast stiffness. The nonlinear dynamic characteristics are presented in the dynamic responses of the mast under different excitation force levels. The joint nonlinearities in the deployable mast are identified as nonlinear hysteresis contributed by the coulomb friction which soften the mast stiffness and lower the mast frequencies.展开更多
基金supported by the National Natural Science Foundation of China(U21A20519)。
文摘As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge devices.The execution speed of the deployed model is a key element to ensure service quality.Considering a highly heterogeneous edge deployment scenario,deep learning compiling is a novel approach that aims to solve this problem.It defines models using certain DSLs and generates efficient code implementations on different hardware devices.However,there are still two aspects that are not yet thoroughly investigated yet.The first is the optimization of memory-intensive operations,and the second problem is the heterogeneity of the deployment target.To that end,in this work,we propose a system solution that optimizes memory-intensive operation,optimizes the subgraph distribution,and enables the compiling and deployment of DNN models on multiple targets.The evaluation results show the performance of our proposed system.
基金supported by the Natural Science Foundation of Shaanxi Province,China(2020JQ-288)Science and Technology on Space Intelligent Control Laboratory,China(HTKJ2019KL502016)+1 种基金China Scholarship Council(201806120093)National Natural Science Foundation of China(61903289).
文摘The tether deployment of a tethered satellite system involves the consideration of complex dynamic properties of the tether,such as large deformation,slack,and even rebound,and therefore,the dynamic modelling of the tether is necessary for performing a dynamic analysis of the system.For a variablelength tether element,the absolute nodal coordinate formulation(ANCF)in the framework of the arbitrary Lagrange-Euler(ALE)description was used to develop a precise dynamic model of a tethered satellite.The model considered the gravitational gradient force and Coriolis force in the orbital coordinate frame,and it was validated through numerical simulation.In the presence of dynamic constraints,a deployment velocity of the tether was obtained by an optimal procedure.In the simulation,rebound behavior of the tethered satellite system was observed when the ANCF-ALE model was employed.Notably,the rebound behavior cannot be predicted by the traditional dumbbell model.Furthermore,an improved optimal deployment velocity was developed.Simulation results indicated that the rebound phenomenon was eliminated,and smooth deployment as well as a stable state of the station-keeping process were achieved.Additionally,the swing amplitude in the station-keeping phase decreased when a deployment strategy based on the improved optimal deployment velocity was used.
基金Projects(50935002, 11002039) supported by the National Natural Science Foundation of ChinaProject(HIT.KLOF.2009062) supported by Key Laboratory Opening Funding of Aerospace Mechanism and Control Technology,Chinasupport by "111 Project" (Grant No.B07018)
文摘The dynamic equivalent continuum modeling method of the mast which is based on energy equivalency principle was investigated. And three kinds of mast dynamic model were established, which were equivalent continuum model, finite element model and simulation model, respectively. The mast frequencies and mode shapes were calculated by these models and compared with each other. The error between the equivalent continuum model and the finite element model is less than 5% when the mast length is longer. Dynamic responses of the mast with different lengths are tested, the mode frequencies and mode shapes are compared with finite element model. The mode shapes match well with each other, while the frequencies tested by experiments are lower than the results of the finite element model, which reflects the joints lower the mast stiffness. The nonlinear dynamic characteristics are presented in the dynamic responses of the mast under different excitation force levels. The joint nonlinearities in the deployable mast are identified as nonlinear hysteresis contributed by the coulomb friction which soften the mast stiffness and lower the mast frequencies.
