The method of integrated data processing for GPS and INS(inertial navigation system) field test over the Rocky Mountains using the adaptive Kalman filtering technique is presented. On the basis of the known GPS output...The method of integrated data processing for GPS and INS(inertial navigation system) field test over the Rocky Mountains using the adaptive Kalman filtering technique is presented. On the basis of the known GPS outputs and the offset of GPS and INS, state equations and observations are designed to perform the calculation and improve the navigation accuracy. An example shows that with the method the reliable navigation parameters have been obtained.展开更多
On April 20, 2013, an M_s 7.0 earthquake struck Lushan County in Sichuan Province, China, and caused serious damage to the source region. We investigated the rupture process of the M_s7.0 Lushan earthquake by jointly ...On April 20, 2013, an M_s 7.0 earthquake struck Lushan County in Sichuan Province, China, and caused serious damage to the source region. We investigated the rupture process of the M_s7.0 Lushan earthquake by jointly inverting waveforms of teleseismic P waveforms and local strong motion records as well as static GPS observations. The inverted results indicate that the rupture of this earthquake was dominated by the failure of an asperity with a triangular shape and that the main shock was dominated by thrust slip. The earthquake released a total seismic moment of 1.01× 10^（19）Nm, with 92% of it being released during the first 11 s. The rupture had an average slip of 0.9 m and produced an average stress drop of 1.8 MPa. Compared with our previous work that was based mainly on a unique dataset, this joint inversion result is more consistent with field observations and the distribution of aftershock zones.展开更多
This paper is part of a research under enhancement since 2001, in which the main objective is to measure small dynamic displacements by analysis of L1 GPS carrier frequency with 1575.42 MHz—wavelength 19.05 cm, under...This paper is part of a research under enhancement since 2001, in which the main objective is to measure small dynamic displacements by analysis of L1 GPS carrier frequency with 1575.42 MHz—wavelength 19.05 cm, under an adaptive method for collecting data and filtering techniques. This method, named Phase Residual Method (PRM) is based on the frequency domain analysis of the phase residuals resulted from the L1 double difference static data processing of two satellites in almost orthogonal elevation angle. In this work it is proposed to obtain the phase residuals directly from the raw phase observable collected in a short baseline during a limited time span, in lieu of obtaining the residual data file from regular GPS processing programs. In order to improve the ability to detect millimetric displacements, two filtering techniques are introduced. The first one is the autocorrelation that reduces the phase noise with random time behavior. The other one is the running mean to separate low frequency from the high frequency phase sources. Two trials are presented to verify the proposed method and filtering techniques applied. One simulates a 2.5 millimeter vertical GPS antenna displacement and the second using the data collected during a bridge dynamic load test. The results show a good consistency to detect millimetric oscillations from L1 frequency and filtering techniques.展开更多
In this work,we combined the model based reinforcement learning(MBRL)and model free reinforcement learning(MFRL)to stabilize a biped robot(NAO robot)on a rotating platform,where the angular velocity of the platform is...In this work,we combined the model based reinforcement learning(MBRL)and model free reinforcement learning(MFRL)to stabilize a biped robot(NAO robot)on a rotating platform,where the angular velocity of the platform is unknown for the proposed learning algorithm and treated as the external disturbance.Nonparametric Gaussian processes normally require a large number of training data points to deal with the discontinuity of the estimated model.Although some improved method such as probabilistic inference for learning control(PILCO)does not require an explicit global model as the actions are obtained by directly searching the policy space,the overfitting and lack of model complexity may still result in a large deviation between the prediction and the real system.Besides,none of these approaches consider the data error and measurement noise during the training process and test process,respectively.We propose a hierarchical Gaussian processes(GP)models,containing two layers of independent GPs,where the physically continuous probability transition model of the robot is obtained.Due to the physically continuous estimation,the algorithm overcomes the overfitting problem with a guaranteed model complexity,and the number of training data is also reduced.The policy for any given initial state is generated automatically by minimizing the expected cost according to the predefined cost function and the obtained probability distribution of the state.Furthermore,a novel Q(λ)based MFRL method scheme is employed to improve the policy.Simulation results show that the proposed RL algorithm is able to balance NAO robot on a rotating platform,and it is capable of adapting to the platform with varying angular velocity.展开更多
Similar to many fields of sciences,recent deep learning advances have been applied extensively in geosciences for both small-and large-scale problems.However,the necessity of using large training data and the’black ...Similar to many fields of sciences,recent deep learning advances have been applied extensively in geosciences for both small-and large-scale problems.However,the necessity of using large training data and the’black box’nature of learning have limited them in practice and difficult to interpret.Furthermore,including the governing equations and physical facts in such methods is also another challenge,which entails either ignoring the physics or simplifying them using unrealistic data.To address such issues,physics informed machine learning methods have been developed which can integrate the governing physics law into the learning process.In this work,a 1-dimensional(1 D)time-dependent seismic wave equation is considered and solved using two methods,namely Gaussian process(GP)and physics informed neural networks.We show that these meshless methods are trained by smaller amount of data and can predict the solution of the equation with even high accuracy.They are also capable of inverting any parameter involved in the governing equation such as wave velocity in our case.Results show that the GP can predict the solution of the seismic wave equation with a lower level of error,while our developed neural network is more accurate for velocity(P-and S-wave)and density inversion.展开更多
Knowledge-Based Engineering （KBE） is introduced into the ship structural design in this paper. From the implementation of KBE, the design solutions for both Rules Design Method （RDM） and Interpolation Design Meth...Knowledge-Based Engineering （KBE） is introduced into the ship structural design in this paper. From the implementation of KBE, the design solutions for both Rules Design Method （RDM） and Interpolation Design Method （IDM） are generated. The corresponding Finite Element （FE） models are generated. Topological design of the longitudinal structures is studied where the Gaussian Process （GP） is employed to build the surrogate model for FE analysis. Multi-objective optimization methods inspired by Pareto Front are used to reduce the design tank weight and outer surface area simultaneously. Additionally, an enhanced Level Set Method （LSM） which employs implicit algorithm is applied to the topological design of typical bracket plate which is used extensively in ship structures. Two different sets of boundary conditions are considered. The proposed methods show satisfactory efficiency and accuracy.展开更多
A novel model named Multi-scale Gaussian Processes (MGP) is proposed. Motivated by the ideas of multi-scale representations in the wavelet theory, in the new model, a Gaussian process is represented at a scale by a li...A novel model named Multi-scale Gaussian Processes (MGP) is proposed. Motivated by the ideas of multi-scale representations in the wavelet theory, in the new model, a Gaussian process is represented at a scale by a linear basis that is composed of a scale function and its different translations. Finally the distribution of the targets of the given samples can be obtained at different scales. Compared with the standard Gaussian Processes (GP) model, the MGP model can control its complexity conveniently just by adjusting the scale pa-rameter. So it can trade-off the generalization ability and the empirical risk rapidly. Experiments verify the fea-sibility of the MGP model, and exhibit that its performance is superior to the GP model if appropriate scales are chosen.展开更多
Joint roughness is one of the most important issues in the hydromechanical behavior of rock mass.Therefore,the joint roughness coefficient(JRC)estimation is of paramount importance in geomechanics engineering applicat...Joint roughness is one of the most important issues in the hydromechanical behavior of rock mass.Therefore,the joint roughness coefficient(JRC)estimation is of paramount importance in geomechanics engineering applications.Studies show that the application of statistical parameters alone may not produce a sufficiently reliable estimation of the JRC values.Therefore,alternative data-driven methods are proposed to assess the JRC values.In this study,Gaussian process(GP),K-star,random forest(RF),and extreme gradient boosting(XGBoost)models are employed,and their performance and accuracy are compared with those of benchmark regression formula(i.e.Z2,Rp,and SDi)for the JRC estimation.To analyze the models’performance,112 rock joint profile datasets having eight common statistical parameters(R_(ave),R_(max),SD_(h),iave,SD_(i),Z_(2),R_(p),and SF)and one output variable(JRC)are utilized,of which 89 and 23 datasets are used for training and validation of models,respectively.The interpretability of the developed XGBoost model is presented in terms of feature importance ranking,partial dependence plots(PDPs),feature interaction,and local interpretable model-agnostic explanations(LIME)techniques.Analyses of results show that machine learning models demonstrate higher accuracy and precision for estimating JRC values compared with the benchmark empirical equations,indicating the generalization ability of the data-driven models in better estimation accuracy.展开更多
It has long been thought that bioprocess, with their inherent measurement difficulties and complex dynamics, posed almost insurmountable problems to engineers. A novel software sensor is proposed to make more effectiv...It has long been thought that bioprocess, with their inherent measurement difficulties and complex dynamics, posed almost insurmountable problems to engineers. A novel software sensor is proposed to make more effective use of those meas- urements that are already available, which enable improvement in fermentation process control. The proposed method is based on mixtures of Gaussian processes (GP) with expectation maximization (EM) algorithm employed for parameter estimation of mixture of models. The mixture model can alleviate computational complexity of GP and also accord with changes of operating condition in fermentation processes, i.e., it would certainly be able to examine what types of process-knowledge would be most relevant for local models’ specific operating points of the process and then combine them into a global one. Demonstrated by on-line estimate of yeast concentration in fermentation industry as an example, it is shown that soft sensor based state estimation is a powerful technique for both enhancing automatic control performance of biological systems and implementing on-line moni- toring and optimization.展开更多
We present our results by using a machine learning(ML)approach for the solution of the Riemann problem for the Euler equations of fluid dynamics.The Riemann problem is an initial-value problem with piecewise-constant ...We present our results by using a machine learning(ML)approach for the solution of the Riemann problem for the Euler equations of fluid dynamics.The Riemann problem is an initial-value problem with piecewise-constant initial data and it represents a mathematical model of the shock tube.The solution of the Riemann problem is the building block for many numerical algorithms in computational fluid dynamics,such as finite-volume or discontinuous Galerkin methods.Therefore,a fast and accurate approximation of the solution of the Riemann problem and construction of the associated numerical fluxes is of crucial importance.The exact solution of the shock tube problem is fully described by the intermediate pressure and mathematically reduces to finding a solution of a nonlinear equation.Prior to delving into the complexities of ML for the Riemann problem,we consider a much simpler formulation,yet very informative,problem of learning roots of quadratic equations based on their coefficients.We compare two approaches:(i)Gaussian process(GP)regressions,and(ii)neural network(NN)approximations.Among these approaches,NNs prove to be more robust and efficient,although GP can be appreciably more accurate(about 30\%).We then use our experience with the quadratic equation to apply the GP and NN approaches to learn the exact solution of the Riemann problem from the initial data or coefficients of the gas equation of state(EOS).We compare GP and NN approximations in both regression and classification analysis and discuss the potential benefits and drawbacks of the ML approach.展开更多
Knowledge-based engineering(KBE) has made success in automobile and molding design industry, and it is introduced into the ship structural design in this paper. From the implementation of KBE, the deterministic design...Knowledge-based engineering(KBE) has made success in automobile and molding design industry, and it is introduced into the ship structural design in this paper. From the implementation of KBE, the deterministic design solutions for both rules design method(RDM) and interpolation design method(IDM) are generated. The corresponding finite element model is generated. Gaussian process(GP) is then employed to build the surrogate model for finite element analysis, in order to increase efficiency and maintain accuracy at the same time, and the multi-modal adaptive importance sampling method is adopted to calculate the corresponding structural reliability.An example is given to validate the proposed method. Finally, the reliabilities of the structures' strength caused by uncertainty lying in water corrosion, static and wave moments are calculated, and the ship structures are optimized to resist the water corrosion by multi-island genetic algorithm. Deterministic design results from the RDM and IDM are compared with each separate robust design result. The proposed method shows great efficiency and accuracy.展开更多
Robot interaction control with variable impedance parameters may conform to task requirements during continuous interaction with dynamic environments.Iterative learning(IL)is effective to learn desired impedance param...Robot interaction control with variable impedance parameters may conform to task requirements during continuous interaction with dynamic environments.Iterative learning(IL)is effective to learn desired impedance parameters for robots under unknown environments,and Gaussian process(GP)is a nonparametric Bayesian approach that models complicated functions with provable confidence using limited data.In this paper,we propose an impedance IL method enhanced by a sparse online Gaussian process(SOGP)to speed up learning convergence and improve generalization.The SOGP for variable impedance modeling is updated in the same iteration by removing similar data points from previous iterations while learning impedance parameters in multiple iterations.The proposed IL-SOGP method is verified by high-fidelity simulations of a collaborative robot with 7 degrees of freedom based on the admittance control framework.It is shown that the proposed method accelerates iterative convergence and improves generalization compared to the classical IL-based impedance learning method.展开更多
Global Navigation Satellite System(GNSS)data is an inexpensive and ubiquitous source of activity data.Global Positioning System(GPS)is an example of such data.Although there have been several studies about inferring d...Global Navigation Satellite System(GNSS)data is an inexpensive and ubiquitous source of activity data.Global Positioning System(GPS)is an example of such data.Although there have been several studies about inferring device activity using GPS data from a consumer device,freight GPS data presents unique challenges for example having low and variable frequency,long transmission gaps,and frequent and unpredictable device ID resetting for preserving privacy.This study aims to provide an end-to-end,generic data analytical framework to infer multiple aspects of truck activity such as stops,trips,and tours.We use popular existing methods to construct the data processing pipeline and provide insights into their practical usage.We also propose improved data filters to different aspects of the data processing pipeline to address challenges found in privacy-preserving freight GPS data.We use freight data across four weeks from the greater Philadelphia region with variable transmission frequency ranging from one second to several hours to perform experiments and validate our methods.Our findings indicate that auxiliary information such as land use can be helpful in fine tuning stop inference,but spatio-temporal information contained in timestamped GPS pings is still the most powerful source of false stop identification.We also find that a combination of simple clustering techniques can provide a way to perform fast and reasonable clustering of the same stop.展开更多
In this paper,the recently developed machine learning(ML)approach to improve orbit prediction accuracy is systematically investigated using three ML algorithms,including support vector machine(SVM),artificial neural n...In this paper,the recently developed machine learning(ML)approach to improve orbit prediction accuracy is systematically investigated using three ML algorithms,including support vector machine(SVM),artificial neural network(ANN),and Gaussian processes(GPs).In a simulation environment consisting of orbit propagation,measurement,estimation,and prediction processes,totally 12 resident space objects(RSOs)in solar-synchronous orbit(SSO),low Earth orbit(LEO),and medium Earth orbit(MEO)are simulated to compare the performance of three ML algorithms.The results in this paper show that ANN usually has the best approximation capability but is easiest to overfit data;SVM is the least likely to overfit but the performance usually cannot surpass ANN and GPs.Additionally,the ML approach with all the three algorithms is observed to be robust with respect to the measurement noise.展开更多
Model-based methods require an accurate dynamic model to design the controller.However,the hydraulic parameters of nonlinear systems,complex friction,or actuator dynamics make it challenging to obtain accurate models....Model-based methods require an accurate dynamic model to design the controller.However,the hydraulic parameters of nonlinear systems,complex friction,or actuator dynamics make it challenging to obtain accurate models.In this case,using the input-output data of the system to learn a dynamic model is an alternative approach.Therefore,we propose a dynamic model based on the Gaussian process(GP)to construct systems with control constraints.Since GP provides a measure of model confidence,it can deal with uncertainty.Unfortunately,most GP-based literature considers model uncertainty but does not consider the effect of constraints on inputs in closed-loop systems.An auxiliary system is developed to deal with the influence of the saturation constraints of input.Meanwhile,we relax the nonsingular assumption of the control coefficients to construct the controller.Some numerical results verify the rationality of the proposed approach and compare it with similar methods.展开更多
基金Supported by the Scientific Research Foundation for ROCS,SEMJiangxi Education Bureau Project(No.200525) .
文摘The method of integrated data processing for GPS and INS(inertial navigation system) field test over the Rocky Mountains using the adaptive Kalman filtering technique is presented. On the basis of the known GPS outputs and the offset of GPS and INS, state equations and observations are designed to perform the calculation and improve the navigation accuracy. An example shows that with the method the reliable navigation parameters have been obtained.
基金supported by a grant from the Chinese Earthquake Administration (No.201308013)the National Natural Science Foundation of China (Nos.41604057, 40974034, and 41021003)a key project from the Institute of Geodesy and Geophysics
文摘On April 20, 2013, an M_s 7.0 earthquake struck Lushan County in Sichuan Province, China, and caused serious damage to the source region. We investigated the rupture process of the M_s7.0 Lushan earthquake by jointly inverting waveforms of teleseismic P waveforms and local strong motion records as well as static GPS observations. The inverted results indicate that the rupture of this earthquake was dominated by the failure of an asperity with a triangular shape and that the main shock was dominated by thrust slip. The earthquake released a total seismic moment of 1.01× 10^（19）Nm, with 92% of it being released during the first 11 s. The rupture had an average slip of 0.9 m and produced an average stress drop of 1.8 MPa. Compared with our previous work that was based mainly on a unique dataset, this joint inversion result is more consistent with field observations and the distribution of aftershock zones.
文摘This paper is part of a research under enhancement since 2001, in which the main objective is to measure small dynamic displacements by analysis of L1 GPS carrier frequency with 1575.42 MHz—wavelength 19.05 cm, under an adaptive method for collecting data and filtering techniques. This method, named Phase Residual Method (PRM) is based on the frequency domain analysis of the phase residuals resulted from the L1 double difference static data processing of two satellites in almost orthogonal elevation angle. In this work it is proposed to obtain the phase residuals directly from the raw phase observable collected in a short baseline during a limited time span, in lieu of obtaining the residual data file from regular GPS processing programs. In order to improve the ability to detect millimetric displacements, two filtering techniques are introduced. The first one is the autocorrelation that reduces the phase noise with random time behavior. The other one is the running mean to separate low frequency from the high frequency phase sources. Two trials are presented to verify the proposed method and filtering techniques applied. One simulates a 2.5 millimeter vertical GPS antenna displacement and the second using the data collected during a bridge dynamic load test. The results show a good consistency to detect millimetric oscillations from L1 frequency and filtering techniques.
文摘In this work,we combined the model based reinforcement learning(MBRL)and model free reinforcement learning(MFRL)to stabilize a biped robot(NAO robot)on a rotating platform,where the angular velocity of the platform is unknown for the proposed learning algorithm and treated as the external disturbance.Nonparametric Gaussian processes normally require a large number of training data points to deal with the discontinuity of the estimated model.Although some improved method such as probabilistic inference for learning control(PILCO)does not require an explicit global model as the actions are obtained by directly searching the policy space,the overfitting and lack of model complexity may still result in a large deviation between the prediction and the real system.Besides,none of these approaches consider the data error and measurement noise during the training process and test process,respectively.We propose a hierarchical Gaussian processes(GP)models,containing two layers of independent GPs,where the physically continuous probability transition model of the robot is obtained.Due to the physically continuous estimation,the algorithm overcomes the overfitting problem with a guaranteed model complexity,and the number of training data is also reduced.The policy for any given initial state is generated automatically by minimizing the expected cost according to the predefined cost function and the obtained probability distribution of the state.Furthermore,a novel Q(λ)based MFRL method scheme is employed to improve the policy.Simulation results show that the proposed RL algorithm is able to balance NAO robot on a rotating platform,and it is capable of adapting to the platform with varying angular velocity.
文摘Similar to many fields of sciences,recent deep learning advances have been applied extensively in geosciences for both small-and large-scale problems.However,the necessity of using large training data and the’black box’nature of learning have limited them in practice and difficult to interpret.Furthermore,including the governing equations and physical facts in such methods is also another challenge,which entails either ignoring the physics or simplifying them using unrealistic data.To address such issues,physics informed machine learning methods have been developed which can integrate the governing physics law into the learning process.In this work,a 1-dimensional(1 D)time-dependent seismic wave equation is considered and solved using two methods,namely Gaussian process(GP)and physics informed neural networks.We show that these meshless methods are trained by smaller amount of data and can predict the solution of the equation with even high accuracy.They are also capable of inverting any parameter involved in the governing equation such as wave velocity in our case.Results show that the GP can predict the solution of the seismic wave equation with a lower level of error,while our developed neural network is more accurate for velocity(P-and S-wave)and density inversion.
基金financially supported by the Project of Ministry of Education and Finance of China(Grant Nos.200512 and 201335)the Project of the State Key Laboratory of Ocean Engineering,Shanghai Jiao Tong University(Grant No.GKZD010053-10)
文摘Knowledge-Based Engineering （KBE） is introduced into the ship structural design in this paper. From the implementation of KBE, the design solutions for both Rules Design Method （RDM） and Interpolation Design Method （IDM） are generated. The corresponding Finite Element （FE） models are generated. Topological design of the longitudinal structures is studied where the Gaussian Process （GP） is employed to build the surrogate model for FE analysis. Multi-objective optimization methods inspired by Pareto Front are used to reduce the design tank weight and outer surface area simultaneously. Additionally, an enhanced Level Set Method （LSM） which employs implicit algorithm is applied to the topological design of typical bracket plate which is used extensively in ship structures. Two different sets of boundary conditions are considered. The proposed methods show satisfactory efficiency and accuracy.
文摘A novel model named Multi-scale Gaussian Processes (MGP) is proposed. Motivated by the ideas of multi-scale representations in the wavelet theory, in the new model, a Gaussian process is represented at a scale by a linear basis that is composed of a scale function and its different translations. Finally the distribution of the targets of the given samples can be obtained at different scales. Compared with the standard Gaussian Processes (GP) model, the MGP model can control its complexity conveniently just by adjusting the scale pa-rameter. So it can trade-off the generalization ability and the empirical risk rapidly. Experiments verify the fea-sibility of the MGP model, and exhibit that its performance is superior to the GP model if appropriate scales are chosen.
文摘Joint roughness is one of the most important issues in the hydromechanical behavior of rock mass.Therefore,the joint roughness coefficient(JRC)estimation is of paramount importance in geomechanics engineering applications.Studies show that the application of statistical parameters alone may not produce a sufficiently reliable estimation of the JRC values.Therefore,alternative data-driven methods are proposed to assess the JRC values.In this study,Gaussian process(GP),K-star,random forest(RF),and extreme gradient boosting(XGBoost)models are employed,and their performance and accuracy are compared with those of benchmark regression formula(i.e.Z2,Rp,and SDi)for the JRC estimation.To analyze the models’performance,112 rock joint profile datasets having eight common statistical parameters(R_(ave),R_(max),SD_(h),iave,SD_(i),Z_(2),R_(p),and SF)and one output variable(JRC)are utilized,of which 89 and 23 datasets are used for training and validation of models,respectively.The interpretability of the developed XGBoost model is presented in terms of feature importance ranking,partial dependence plots(PDPs),feature interaction,and local interpretable model-agnostic explanations(LIME)techniques.Analyses of results show that machine learning models demonstrate higher accuracy and precision for estimating JRC values compared with the benchmark empirical equations,indicating the generalization ability of the data-driven models in better estimation accuracy.
基金Project (No. 2002AA412010) supported by the National High-TechResearch and Development Program (863) of China
文摘It has long been thought that bioprocess, with their inherent measurement difficulties and complex dynamics, posed almost insurmountable problems to engineers. A novel software sensor is proposed to make more effective use of those meas- urements that are already available, which enable improvement in fermentation process control. The proposed method is based on mixtures of Gaussian processes (GP) with expectation maximization (EM) algorithm employed for parameter estimation of mixture of models. The mixture model can alleviate computational complexity of GP and also accord with changes of operating condition in fermentation processes, i.e., it would certainly be able to examine what types of process-knowledge would be most relevant for local models’ specific operating points of the process and then combine them into a global one. Demonstrated by on-line estimate of yeast concentration in fermentation industry as an example, it is shown that soft sensor based state estimation is a powerful technique for both enhancing automatic control performance of biological systems and implementing on-line moni- toring and optimization.
基金This work was performed under the auspices of the National Nuclear Security Administration of the US Department of Energy at Los Alamos National Laboratory under Contract No.DE-AC52-06NA25396The authors gratefully acknowledge the support of the US Department of Energy National Nuclear Security Administration Advanced Simulation and Computing Program.The Los Alamos unlimited release number is LA-UR-19-32257.
文摘We present our results by using a machine learning(ML)approach for the solution of the Riemann problem for the Euler equations of fluid dynamics.The Riemann problem is an initial-value problem with piecewise-constant initial data and it represents a mathematical model of the shock tube.The solution of the Riemann problem is the building block for many numerical algorithms in computational fluid dynamics,such as finite-volume or discontinuous Galerkin methods.Therefore,a fast and accurate approximation of the solution of the Riemann problem and construction of the associated numerical fluxes is of crucial importance.The exact solution of the shock tube problem is fully described by the intermediate pressure and mathematically reduces to finding a solution of a nonlinear equation.Prior to delving into the complexities of ML for the Riemann problem,we consider a much simpler formulation,yet very informative,problem of learning roots of quadratic equations based on their coefficients.We compare two approaches:(i)Gaussian process(GP)regressions,and(ii)neural network(NN)approximations.Among these approaches,NNs prove to be more robust and efficient,although GP can be appreciably more accurate(about 30\%).We then use our experience with the quadratic equation to apply the GP and NN approaches to learn the exact solution of the Riemann problem from the initial data or coefficients of the gas equation of state(EOS).We compare GP and NN approximations in both regression and classification analysis and discuss the potential benefits and drawbacks of the ML approach.
基金the Project of Ministry of Finance andMinistry of Education of China(Nos.200512 and201335)the State Key Laboratory of Ocean Engineering Foundation of Shanghai Jiao Tong University(No.GKZD010053-10)
文摘Knowledge-based engineering(KBE) has made success in automobile and molding design industry, and it is introduced into the ship structural design in this paper. From the implementation of KBE, the deterministic design solutions for both rules design method(RDM) and interpolation design method(IDM) are generated. The corresponding finite element model is generated. Gaussian process(GP) is then employed to build the surrogate model for finite element analysis, in order to increase efficiency and maintain accuracy at the same time, and the multi-modal adaptive importance sampling method is adopted to calculate the corresponding structural reliability.An example is given to validate the proposed method. Finally, the reliabilities of the structures' strength caused by uncertainty lying in water corrosion, static and wave moments are calculated, and the ship structures are optimized to resist the water corrosion by multi-island genetic algorithm. Deterministic design results from the RDM and IDM are compared with each separate robust design result. The proposed method shows great efficiency and accuracy.
基金supported in part by the National Research Foundation of Korea(NRF)Grant Funded by the Korea Government(MSIT)(RS-2025-00555064).Recommended by Associate Editor Zengguang Hou.
文摘Robot interaction control with variable impedance parameters may conform to task requirements during continuous interaction with dynamic environments.Iterative learning(IL)is effective to learn desired impedance parameters for robots under unknown environments,and Gaussian process(GP)is a nonparametric Bayesian approach that models complicated functions with provable confidence using limited data.In this paper,we propose an impedance IL method enhanced by a sparse online Gaussian process(SOGP)to speed up learning convergence and improve generalization.The SOGP for variable impedance modeling is updated in the same iteration by removing similar data points from previous iterations while learning impedance parameters in multiple iterations.The proposed IL-SOGP method is verified by high-fidelity simulations of a collaborative robot with 7 degrees of freedom based on the admittance control framework.It is shown that the proposed method accelerates iterative convergence and improves generalization compared to the classical IL-based impedance learning method.
文摘Global Navigation Satellite System(GNSS)data is an inexpensive and ubiquitous source of activity data.Global Positioning System(GPS)is an example of such data.Although there have been several studies about inferring device activity using GPS data from a consumer device,freight GPS data presents unique challenges for example having low and variable frequency,long transmission gaps,and frequent and unpredictable device ID resetting for preserving privacy.This study aims to provide an end-to-end,generic data analytical framework to infer multiple aspects of truck activity such as stops,trips,and tours.We use popular existing methods to construct the data processing pipeline and provide insights into their practical usage.We also propose improved data filters to different aspects of the data processing pipeline to address challenges found in privacy-preserving freight GPS data.We use freight data across four weeks from the greater Philadelphia region with variable transmission frequency ranging from one second to several hours to perform experiments and validate our methods.Our findings indicate that auxiliary information such as land use can be helpful in fine tuning stop inference,but spatio-temporal information contained in timestamped GPS pings is still the most powerful source of false stop identification.We also find that a combination of simple clustering techniques can provide a way to perform fast and reasonable clustering of the same stop.
基金The authors would acknowledge the research support from the Air Force Office of Scientific Research(AFOSR)FA9550-16-1-0184 and the Office of Naval Research(ONR)N00014-16-1-2729.Large amount of simulations of RSOs have been supported by the HPC cluster in School of Engineering,Rutgers University.
文摘In this paper,the recently developed machine learning(ML)approach to improve orbit prediction accuracy is systematically investigated using three ML algorithms,including support vector machine(SVM),artificial neural network(ANN),and Gaussian processes(GPs).In a simulation environment consisting of orbit propagation,measurement,estimation,and prediction processes,totally 12 resident space objects(RSOs)in solar-synchronous orbit(SSO),low Earth orbit(LEO),and medium Earth orbit(MEO)are simulated to compare the performance of three ML algorithms.The results in this paper show that ANN usually has the best approximation capability but is easiest to overfit data;SVM is the least likely to overfit but the performance usually cannot surpass ANN and GPs.Additionally,the ML approach with all the three algorithms is observed to be robust with respect to the measurement noise.
文摘Model-based methods require an accurate dynamic model to design the controller.However,the hydraulic parameters of nonlinear systems,complex friction,or actuator dynamics make it challenging to obtain accurate models.In this case,using the input-output data of the system to learn a dynamic model is an alternative approach.Therefore,we propose a dynamic model based on the Gaussian process(GP)to construct systems with control constraints.Since GP provides a measure of model confidence,it can deal with uncertainty.Unfortunately,most GP-based literature considers model uncertainty but does not consider the effect of constraints on inputs in closed-loop systems.An auxiliary system is developed to deal with the influence of the saturation constraints of input.Meanwhile,we relax the nonsingular assumption of the control coefficients to construct the controller.Some numerical results verify the rationality of the proposed approach and compare it with similar methods.
